diff --git a/docs/common/ai/_cubie_acuity_env.mdx b/docs/common/ai/_cubie_acuity_env.mdx index 06f3e5a5b..6be96d73f 100644 --- a/docs/common/ai/_cubie_acuity_env.mdx +++ b/docs/common/ai/_cubie_acuity_env.mdx @@ -29,10 +29,11 @@ sudo curl -fsSL https://download.docker.com/linux/ubuntu/gpg -o /etc/apt/keyring sudo chmod a+r /etc/apt/keyrings/docker.asc # Add the repository to Apt sources: + echo \ - "deb [arch=$(dpkg --print-architecture) signed-by=/etc/apt/keyrings/docker.asc] https://download.docker.com/linux/ubuntu \ - $(. /etc/os-release && echo "${UBUNTU_CODENAME:-$VERSION_CODENAME}") stable" | \ - sudo tee /etc/apt/sources.list.d/docker.list > /dev/null + "deb [arch=$(dpkg --print-architecture) signed-by=/etc/apt/keyrings/docker.asc] https://download.docker.com/linux/ubuntu \ + $(. /etc/os-release && echo "${UBUNTU_CODENAME:-$VERSION_CODENAME}") stable" | \ + sudo tee /etc/apt/sources.list.d/docker.list > /dev/null sudo apt-get update ``` @@ -54,108 +55,89 @@ sudo apt-get install docker-ce docker-ce-cli containerd.io docker-buildx-plugin 由于SDK版本兼容问题，需要根据NPU进行不同的选择 ::: - - - - - ### 获取 ACUITY 下载包 - - 在 [全志网盘](https://netstorage.allwinnertech.com:5001/sharing/Mh23BhPHq) 下载 ACUITY docker 压缩包并解压 - - ```bash - unzip docker_images_v2.0.x.zip - ``` - - ### 载入镜像 - - - - ```bash - cd docker_images_v2.0.x - unzip ubuntu-npu_v2.0.10.tar.zip - sudo docker load -i ubuntu-npu_v2.0.10.tar - ``` - - - - 当 docker 镜像载入完成后可以在 `docker images` 中看到此镜像， 名字为 `ubuntu‑npu:v2.0.10` - - ### 创建 docker 容器 - - - - ```bash - mkdir docker_data && cd docker_data - sudo docker run --ipc=host -itd -v ${PWD}:/workspace --name allwinner_v2.0.10 ubuntu-npu:v2.0.10 /bin/bash - ``` - - +### 获取 ACUITY 下载包 - 当 docker 容器创建完成后可以在 `docker ps -a` 中看到此容器， 名字为 `allwinner_v2.0.10` +在 [全志网盘](https://netstorage.allwinnertech.com:5001/sharing/Mh23BhPHq) 下载 ACUITY docker 压缩包并解压 - ### 进入 docker 容器 - - 请使用 `docker ps -a` 查看 `allwinner_v2.0.10` 容器 ID + + + - +```bash +unzip docker_images_v2.0.x.zip +``` - ```bash - sudo docker exec -it 容器ID /bin/bash - ``` + + + + - +```bash +unzip docker_images_v1.8.x.zip +``` - + + + - +### 载入镜像 - ### 获取 ACUITY 下载包 + + + - 在 [全志网盘](https://netstorage.allwinnertech.com:5001/sharing/N6TVlZQVZ) 下载 ACUITY docker 压缩包并解压 - - ```bash - unzip docker_images_v1.8.x.zip - ``` - - ### 载入镜像 +```bash +cd docker_images_v2.0.x unzip ubuntu-npu_v2.0.10.1.tar.zip sudo docker load -i ubuntu-npu_v2.0.10.1.tar +``` - + + + + - ```bash - cd docker_images_v1.8.x - unzip ubuntu-npu_v1.8.11.tar.zip - sudo docker load -i ubuntu-npu_v1.8.11.tar - ``` +```bash +cd docker_images_v1.8.x unzip ubuntu-npu_v1.8.11.tar.zip sudo docker load -i ubuntu-npu_v1.8.11.tar +``` - + + + - 当 docker 镜像载入完成后可以在 `docker images` 中看到此镜像， 名字为 `ubuntu-npu:v1.8.11` +当 docker 镜像载入完成后可以在 `docker images` 中看到此镜像， 名字为 `ubuntu‑npu:v2.0.10.1` (A733) 或 `ubuntu-npu:v1.8.11` (T527) - ### 创建 docker 容器 +### 创建 docker 容器 - + + + - ```bash - mkdir docker_data && cd docker_data - sudo docker run --ipc=host -itd -v ${PWD}:/workspace --name allwinner_v1.8.11 ubuntu-npu:v1.8.11 /bin/bash - ``` +```bash +mkdir docker_data && cd docker_data sudo docker run --ipc=host -itd -v ${PWD}:/workspace --name allwinner_v2.0.10.1 ubuntu-npu:v2.0.10.1 /bin/bash +``` - + + + + - 当 docker 容器创建完成后可以在 `docker ps -a` 中看到此容器， 名字为 `allwinner_v1.8.11` +```bash +mkdir docker_data && cd docker_data sudo docker run --ipc=host -itd -v ${PWD}:/workspace --name allwinner_v1.8.11 ubuntu-npu:v1.8.11 /bin/bash +``` - ### 进入 docker 容器 + + + - 请使用 `docker ps -a` 查看 `allwinner_v1.8.11` 容器 ID +当 docker 容器创建完成后可以在 `docker ps -a` 中看到此容器， 名字为 `allwinner_v2.0.10.1` (A733) 或 `allwinner_v1.8.11` (T527) - +### 进入 docker 容器 - ```bash - sudo docker exec -it 容器ID /bin/bash - ``` +请使用 `docker ps -a` 查看容器 ID - + - +```bash +sudo docker exec -it 容器ID /bin/bash +``` - + diff --git a/docs/common/ai/cubie/_model-zoo-densenet121-keras.mdx b/docs/common/ai/cubie/_model-zoo-densenet121-keras.mdx new file mode 100644 index 000000000..a0b648cc8 --- /dev/null +++ b/docs/common/ai/cubie/_model-zoo-densenet121-keras.mdx @@ -0,0 +1,312 @@ +本文档讲述如何在 NPU 上运行 DenseNet121。 + +:::info +参考 [Model Zoo 下载](./model-zoo-download)获取示例。 +::: + +DenseNet121 示例目录结构： + +```bash +$ tree ./ +./ +├── class_post.cpp +├── class_pre.cpp +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ └── convert_model_env.sh +├── label.h +├── main.cpp +├── model +│ └── space_shuttle_224x224.jpg +└── README.md +``` + +## 模型转换 + +### 下载模型 + +点击下载 [densenet121_batch1_224x224.h5](http://netstorage.allwinnertech.com:5000/sharing/d7pP1omn5) 。 + +然后将模型移动到 convert_model/ 目录下。 + +### 创建转换脚本的软链接 + + + +```bash +./convert_model_env.sh +``` + + + +### 模型导入/量化/转换 + +需要先进入容器开发环境。可以参考 Model Zoo 下载中[创建容器](./model-zoo-download#创建并启动容器)这一部分。 + +:::info +不同平台请使用对应的 Docker 镜像： + +- A733：ubuntu-npu:v2.0.10.1 +- T527：ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +进入容器对应目录之后运行脚本。 + + + +```bash +cd /workspace/examples/densenet121_keras/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh densenet121_batch1_224x224 +./pegasus_quantize.sh densenet121_batch1_224x224 uint8 10 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh densenet121_batch1_224x224 uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh densenet121_batch1_224x224 uint8 t527 +``` + + + + + + +导出的模型文件存放在../model目录。 + +### 编译示例 + +接下来可以编译示例，**先 exit 退出容器**，然后执行下面的命令编译示例。 + +首先需要配置第三方库和交叉编译工具链。 + +:::info +如果你已经在其他示例中配置过第三方库和交叉编译工具链则可以跳过这一步。 +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +需要先手动[点击链接](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4)下载之后放到 0-toolchains/ 再执行下面的命令： + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/densenet121_keras/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## 模型部署 + +编译示例完成之后，示例会安装到 install 目录，可以使用 scp 传输到板端。 + +### 配置 NPU 驱动 + +:::info +如果你已经在其他示例中配置过 NPU 驱动则可以跳过这一步。 +::: + +将驱动库 scp 传输到板端的 lib 目录。 + +- A733 对应 common/lib_linux_aarch64/A733 目录 +- T527 对应 common/lib_linux_aarch64/T527 目录 + +然后执行下面的命令导出到环境变量。 + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### 运行示例 + +配置好驱动之后就可以运行示例了。 + +:::tip +对于 T527 平台，你还需要参考 A5E 的`板端启用 NPU`文档先启用 NPU ，然后使用下面的命令增加当前用户使用 /dev/vipcore 的权限。 +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd densenet121_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./densenet121_demo_a733 +./densenet121_demo_a733 -nb model/densenet121_batch1_224x224_uint8_a733.nb -i model/space_shuttle_224x224.jpg +``` + + + +运行结果如下： + +```bash +$ ./densenet121_demo_a733 -nb model/densenet121_batch1_224x224_uint8_a733.nb -i model/space_shuttle_224x224.jpg +model_file=model/densenet121_batch1_224x224_uint8_a733.nb, input=model/space_shuttle_224x224.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 224 224 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 1000 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/densenet121_batch1_224x224_uint8_a733.nb, size: 6397400. +create network 0: 10448 us. +prepare network: 3582 us. +network: 0, loop count: 1 +run time for this network 0: 8278 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 812, prob: 0.973633, label: space shuttle +class id: 569, prob: 0.005028, label: garbage truck, dustcart +class id: 403, prob: 0.003510, label: aircraft carrier, carrier, flattop, attack aircraft carrier +class id: 408, prob: 0.002764, label: amphibian, amphibious vehicle +class id: 895, prob: 0.002176, label: warplane, military plane +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :---------- | :--------- | :----------- | :----------- | :----------- | :--------- | :------ | :------- | +| 全志 A733 | Vivante VIP9000 | densenet121 | 224×224 | 10.4 ms | 3.6 ms | 8.3 ms | | 22.3 ms | 44.8 FPS | + + + + + + + +```bash +cd densenet121_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./densenet121_demo_t527 +./densenet121_demo_t527 -nb model/densenet121_batch1_224x224_uint8_t527.nb -i model/space_shuttle_224x224.jpg +``` + + + +运行结果如下： + +```bash +$ ./densenet121_demo_t527 -nb model/densenet121_batch1_224x224_uint8_t527.nb -i model/space_shuttle_224x224.jpg +model_file=model/densenet121_batch1_224x224_uint8_t527.nb, input=model/space_shuttle_224x224.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 224 224 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 1000 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/densenet121_batch1_224x224_uint8_t527.nb, size: 5857536. +create network 0: 12870 us. +prepare network: 2366 us. +network: 0, loop count: 1 +run time for this network 0: 10573 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 812, prob: 0.961914, label: space shuttle +class id: 569, prob: 0.008011, label: garbage truck, dustcart +class id: 408, prob: 0.004963, label: amphibian, amphibious vehicle +class id: 403, prob: 0.003466, label: aircraft carrier, carrier, flattop, attack aircraft carrier +class id: 895, prob: 0.003466, label: warplane, military plane +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :---------- | :--------- | :----------- | :----------- | :----------- | :--------- | :------ | :------- | +| 全志 T527 | Vivante VIP9000 | densenet121 | 224×224 | 12.9 ms | 2.4 ms | 10.6 ms | | 25.9 ms | 38.6 FPS | + + + diff --git a/docs/common/ai/cubie/_model-zoo-download.mdx b/docs/common/ai/cubie/_model-zoo-download.mdx new file mode 100644 index 000000000..dbd801893 --- /dev/null +++ b/docs/common/ai/cubie/_model-zoo-download.mdx @@ -0,0 +1,112 @@ +我们可以从[全志客户服务平台](https://open.allwinnertech.com/)获取 Model Zoo 。 + +点击立即注册： + +![全志客户服务平台登陆](/img/cubie/allwinnertech-login.webp) + +成功注册并登陆之后会进入工作台界面，点击资源下载： + +![点击资源下载](/img/cubie/allwinnertech-model-zoo-download-guide1.webp) + +点击左侧工具查询下的 AI 开发 SDK，然后找到 v0.9.0 版本的 Model Zoo 下载到本地即可。 + +![下载 Model Zoo](/img/cubie/allwinnertech-model-zoo-download-guide2.webp) + +或者你也可以通过 dl.radxa 下载： + + + +```bash +wget https://dl.radxa.com/cubie/allwinner-model-zoo.tar.gz +``` + + + +运行 Model Zoo 中的案例之前，你需要下载好容器开发包，你可以在上面的界面找到对应版本下载，或者参考我们的[指引文档](../cubie-acuity-env)。 + +:::info +注意 T527 和 A733 使用不同版本的容器开发包。 +::: + +容器开发包和 Model Zoo 都下载好之后，可以参考下面的步骤： + +## 解压 Model Zoo + + + +```bash +tar -xvf 1768567762439_awnpu_model_zoo-v0.9.0-20260116-83a67d4b.tar.gz +cd awnpu_model_zoo-v0.9.0-20260116-83a67d4b/ +``` + + + +## 创建并启动容器 + + + + + + +```bash +sudo docker run --ipc=host -d -v ${PWD}:/workspace --name model-zoo ubuntu-npu:v2.0.10.1 tail -f /dev/null +``` + + + + + + + + + +```bash +sudo docker run --ipc=host -d -v ${PWD}:/workspace --name model-zoo ubuntu-npu:v1.8.11 tail -f /dev/null +``` + + + + + + +当容器创建完成后可以在 `docker ps` 中看到此容器，名字为 `model-zoo`。 + +## 进入容器 + + + +```bash +sudo docker exec -it model-zoo /bin/bash +``` + + + +完成之后 Model Zoo 目录已经挂载到容器的 `/workspace` 目录下，在容器内执行： + + + +```bash +cd /workspace +ls -al +``` + + + +可以看到 Model Zoo 目录结构： + +```bash +/workspace# ls -al +total 52 +drwxrwxr-x 10 1000 1000 4096 Apr 2 19:09 . +drwxr-xr-x 1 root root 4096 Apr 2 18:48 .. +drwxrwxr-x 4 1000 1000 4096 Apr 3 12:03 0-toolchains # 交叉编译工具链 +drwxrwxr-x 3 1000 1000 4096 Jan 16 20:44 3rdparty # 第三方工具链 +-rwxrwxr-x 1 1000 1000 4976 Jan 16 20:44 README.md # README文档 +drwxrwxr-x 2 1000 1000 4096 Apr 3 12:06 cmake_toolchain # 编译工具链配置 +drwxrwxr-x 3 1000 1000 4096 Jan 16 20:44 common # 通用库 +drwxrwxr-x 2 1000 1000 4096 Jan 16 20:44 docs # 文档 +drwxrwxr-x 26 1000 1000 4096 Jan 16 20:44 examples # 示例目录 +drwxrwxr-x 6 1000 1000 4096 Jan 16 20:44 functions # 特色功能 +drwxrwxr-x 2 1000 1000 4096 Jan 16 20:44 scripts_model_convert # 转换脚本 +-rwxrwxr-x 1 1000 1000 6 Jan 16 20:44 version # 版本号 +``` diff --git a/docs/common/ai/cubie/_model-zoo-lenet-caffe.mdx b/docs/common/ai/cubie/_model-zoo-lenet-caffe.mdx new file mode 100644 index 000000000..5dd324a79 --- /dev/null +++ b/docs/common/ai/cubie/_model-zoo-lenet-caffe.mdx @@ -0,0 +1,305 @@ +本文档讲述如何在 NPU 上运行 LeNet。 + +:::info +参考 [Model Zoo 下载](./model-zoo-download)获取示例。 +::: + +LeNet 示例目录结构： + +```bash +$ tree ./ +./ +├── class_pre.cpp +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ └── convert_model_env.sh +├── main.cpp +├── model +│ ├── 3.jpg +│ ├── 4.jpg +│ └── 5.jpg +└── README.md +``` + +## 模型转换 + +### 下载模型 + +点击下载 [lenet.caffemodel](http://netstorage.allwinnertech.com:5000/sharing/JG3iqvFvH) 。 + +点击下载 [lenet.prototxt](http://netstorage.allwinnertech.com:5000/sharing/Wf4J4DsTR) 。 + +然后将模型移动到 convert_model/ 目录下。 + +### 创建转换脚本的软链接 + + + +```bash +cd convert_model/ +./convert_model_env.sh +``` + + + +### 模型导入/量化/转换 + +需要先进入容器开发环境。可以参考 Model Zoo 下载中[创建容器](./model-zoo-download#创建并启动容器)这一部分。 + +:::info +不同平台请使用对应的 Docker 镜像： + +- A733：ubuntu-npu:v2.0.10.1 +- T527：ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +进入容器对应目录之后运行脚本。 + + + +```bash +cd /workspace/examples/lenet_caffe/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh lenet +./pegasus_quantize.sh lenet uint8 12 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh lenet uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh lenet uint8 t527 +``` + + + + + + +导出的模型文件存放在../model目录。 + +### 编译示例 + +接下来可以编译示例，**先 exit 退出容器**，然后执行下面的命令编译示例。 + +首先需要配置第三方库和交叉编译工具链。 + +:::info +如果你已经在其他示例中配置过第三方库和交叉编译工具链则可以跳过这一步。 +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +需要先手动[点击链接](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4)下载之后放到 0-toolchains/ 再执行下面的命令： + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/lenet_caffe/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## 模型部署 + +编译示例完成之后，示例会安装到 install 目录，可以使用 scp 传输到板端。 + +### 配置 NPU 驱动 + +:::info +如果你已经在其他示例中配置过 NPU 驱动则可以跳过这一步。 +::: + +将驱动库 scp 传输到板端的 lib 目录。 + +- A733 对应 common/lib_linux_aarch64/A733 目录 +- T527 对应 common/lib_linux_aarch64/T527 目录 + +然后执行下面的命令导出到环境变量。 + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### 运行示例 + +配置好驱动之后就可以运行示例了。 + +:::tip +对于 T527 平台，你还需要参考 A5E 的`板端启用 NPU`文档先启用 NPU ，然后使用下面的命令增加当前用户使用 /dev/vipcore 的权限。 +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd lenet_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./lenet_demo_a733 +./lenet_demo_a733 -nb model/lenet_uint8_a733.nb -i model/3.jpg +``` + + + +运行结果如下： + +```bash +$ ./lenet_demo_a733 -nb model/lenet_uint8_a733.nb -i model/3.jpg +model_file=model/lenet_uint8_a733.nb, input=model/3.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 28 28 1 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 10 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/lenet_uint8_a733.nb, size: 407776. +create network 0: 1486 us. +prepare network: 176 us. +network: 0, loop count: 1 +run time for this network 0: 281 us. +Image: model/3.jpg, Predicted digit: 3, Probability: 1.000000 +Class probabilities: 0 : 0.0000 1 : 0.0000 2 : 0.0000 3 : 1.0000 4 : 0.0000 5 : 0.0000 6 : 0.0000 7 : 0.0000 8 : 0.0000 9 : 0.0000 +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :---- | :--------- | :----------- | :----------- | :----------- | :--------- | :----- | :-------- | +| 全志 A733 | Vivante VIP9000 | lenet | 28×28 | 1.5 ms | 0.2 ms | 0.3 ms | | 2.0 ms | 500.0 FPS | + + + + + + + +```bash +cd lenet_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./lenet_demo_t527 +./lenet_demo_t527 -nb model/lenet_uint8_t527.nb -i model/3.jpg +``` + + + +运行结果如下： + +```bash +$ ./lenet_demo_t527 -nb model/lenet_uint8_t527.nb -i model/3.jpg +model_file=model/lenet_uint8_t527.nb, input=model/3.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 28 28 1 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 10 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/lenet_uint8_t527.nb, size: 403520. +create network 0: 744 us. +prepare network: 123 us. +network: 0, loop count: 1 +run time for this network 0: 219 us. +Image: model/3.jpg, Predicted digit: 1, Probability: 0.607422 +Class probabilities: 0 : 0.0523 1 : 0.6074 2 : 0.0154 3 : 0.0132 4 : 0.0243 5 : 0.0610 6 : 0.0449 7 : 0.1530 8 : 0.0045 9 : 0.0243 +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :---- | :--------- | :----------- | :----------- | :----------- | :--------- | :----- | :--------- | +| 全志 T527 | Vivante VIP9000 | lenet | 28×28 | 0.7 ms | 0.1 ms | 0.2 ms | | 1.0 ms | 1000.0 FPS | + + + diff --git a/docs/common/ai/cubie/_model-zoo-mobilenetv1-tensorflow.mdx b/docs/common/ai/cubie/_model-zoo-mobilenetv1-tensorflow.mdx new file mode 100644 index 000000000..debad39ce --- /dev/null +++ b/docs/common/ai/cubie/_model-zoo-mobilenetv1-tensorflow.mdx @@ -0,0 +1,314 @@ +本文档讲述如何在 NPU 上运行 MobileNetV1。 + +:::info +参考 [Model Zoo 下载](./model-zoo-download)获取示例。 +::: + +MobileNetV1 示例目录结构： + +```bash +$ tree ./ +./ +├── class_post.cpp +├── class_pre.cpp +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ └── inputs_outputs.txt +├── label.h +├── main.cpp +├── model +│ └── space_shuttle_224x224.jpg +└── README.md +``` + +## 模型转换 + +### 下载模型 + +点击下载 [mobilenet_v1_1.0_224_frozen.pb](http://netstorage.allwinnertech.com:5000/sharing/YtgpJ76vp) 。 + +然后将模型移动到 convert_model/ 目录下。 + +### 创建转换脚本的软链接 + + + +```bash +cd convert_model/ +./convert_model_env.sh +``` + + + +### 模型导入/量化/转换 + +需要先进入容器开发环境。可以参考 Model Zoo 下载中[创建容器](./model-zoo-download#创建并启动容器)这一部分。 + +:::info +不同平台请使用对应的 Docker 镜像： + +- A733：ubuntu-npu:v2.0.10.1 +- T527：ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +进入容器对应目录之后运行脚本。 + + + +```bash +cd /workspace/examples/mobilenetv1_tensorflow/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh mobilenet_v1_1.0_224_frozen +./pegasus_quantize.sh mobilenet_v1_1.0_224_frozen uint8 10 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh mobilenet_v1_1.0_224_frozen uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh mobilenet_v1_1.0_224_frozen uint8 t527 +``` + + + + + + +导出的模型文件存放在../model目录。 + +### 编译示例 + +接下来可以编译示例，**先 exit 退出容器**，然后执行下面的命令编译示例。 + +首先需要配置第三方库和交叉编译工具链。 + +:::info +如果你已经在其他示例中配置过第三方库和交叉编译工具链则可以跳过这一步。 +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +需要先手动[点击链接](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4)下载之后放到 0-toolchains/ 再执行下面的命令： + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/mobilenetv1_tensorflow/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## 模型部署 + +编译示例完成之后，示例会安装到 install 目录，可以使用 scp 传输到板端。 + +### 配置 NPU 驱动 + +:::info +如果你已经在其他示例中配置过 NPU 驱动则可以跳过这一步。 +::: + +将驱动库 scp 传输到板端的 lib 目录。 + +- A733 对应 common/lib_linux_aarch64/A733 目录 +- T527 对应 common/lib_linux_aarch64/T527 目录 + +然后执行下面的命令导出到环境变量。 + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### 运行示例 + +配置好驱动之后就可以运行示例了。 + +:::tip +对于 T527 平台，你还需要参考 A5E 的`板端启用 NPU`文档先启用 NPU ，然后使用下面的命令增加当前用户使用 /dev/vipcore 的权限。 +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd mobilenetv1_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./mobilenetv1_demo_a733 +./mobilenetv1_demo_a733 -nb model/mobilenet_v1_1.0_224_frozen_uint8_a733.nb -i model/space_shuttle_224x224.jpg +``` + + + +运行结果如下： + +```bash +$ ./mobilenetv1_demo_a733 -nb model/mobilenet_v1_1.0_224_frozen_uint8_a733.nb -i model/space_shuttle_224x224.jpg +model_file=model/mobilenet_v1_1.0_224_frozen_uint8_a733.nb, input=model/space_shuttle_224x224.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 224 224 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 1001 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/mobilenet_v1_1.0_224_frozen_uint8_a733.nb, size: 3182272. +create network 0: 5661 us. +prepare network: 1085 us. +network: 0, loop count: 1 +run time for this network 0: 1845 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 813, prob: 0.995117, label: space shuttle +class id: 868, prob: 0.001483, label: trailer truck, tractor trailer, trucking rig, rig, articulated lorry, semi +class id: 405, prob: 0.000979, label: airliner +class id: 406, prob: 0.000563, label: airship, dirigible +class id: 628, prob: 0.000371, label: limousine, limo +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :---------- | :--------- | :----------- | :----------- | :----------- | :--------- | :----- | :-------- | +| 全志 A733 | Vivante VIP9000 | mobilenetv1 | 224×224 | 5.7 ms | 1.1 ms | 1.8 ms | | 8.6 ms | 116.3 FPS | + + + + + + + +```bash +cd mobilenetv1_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./mobilenetv1_demo_t527 +./mobilenetv1_demo_t527 -nb model/mobilenet_v1_1.0_224_frozen_uint8_t527.nb -i model/space_shuttle_224x224.jpg +``` + + + +运行结果如下： + +```bash +$ ./mobilenetv1_demo_t527 -nb model/mobilenet_v1_1.0_224_frozen_uint8_t527.nb -i model/space_shuttle_224x224.jpg +model_file=model/mobilenet_v1_1.0_224_frozen_uint8_t527.nb, input=model/space_shuttle_224x224.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 224 224 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 1001 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/mobilenet_v1_1.0_224_frozen_uint8_t527.nb, size: 3097728. +create network 0: 5639 us. +prepare network: 980 us. +network: 0, loop count: 1 +run time for this network 0: 3809 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 813, prob: 0.993652, label: space shuttle +class id: 868, prob: 0.001479, label: trailer truck, tractor trailer, trucking rig, rig, articulated lorry, semi +class id: 405, prob: 0.000976, label: airliner +class id: 406, prob: 0.000561, label: airship, dirigible +class id: 628, prob: 0.000370, label: limousine, limo +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :---------- | :--------- | :----------- | :----------- | :----------- | :--------- | :------ | :------- | +| 全志 T527 | Vivante VIP9000 | mobilenetv1 | 224×224 | 5.6 ms | 1.0 ms | 3.8 ms | | 10.4 ms | 96.2 FPS | + + + diff --git a/docs/common/ai/cubie/_model-zoo-mobilenetv2.mdx b/docs/common/ai/cubie/_model-zoo-mobilenetv2.mdx new file mode 100644 index 000000000..e536698ef --- /dev/null +++ b/docs/common/ai/cubie/_model-zoo-mobilenetv2.mdx @@ -0,0 +1,313 @@ +本文档讲述如何在 NPU 上运行 MobileNetV2。 + +:::info +参考 [Model Zoo 下载](./model-zoo-download)获取示例。 +::: + +MobileNetV2 示例目录结构： + +```bash +$ tree ./ +./ +├── class_post.cpp +├── class_pre.cpp +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ └── mobilenetv2-12.onnx +├── label.h +├── main.cpp +├── model +│ ├── 1.jpg +│ └── mobilenetv2-12_pcq_t527.nb +└── README.md +``` + +## 模型转换 + +### 下载模型 + +目录中已有模型，无需下载。 + +### 创建转换脚本的软链接 + + + +```bash +cd convert_model/ +./convert_model_env.sh +``` + + + +### 模型导入/量化/转换 + +需要先进入容器开发环境。可以参考 Model Zoo 下载中[创建容器](./model-zoo-download#创建并启动容器)这一部分。 + +:::info +不同平台请使用对应的 Docker 镜像： + +- A733：ubuntu-npu:v2.0.10.1 +- T527：ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +进入容器对应目录之后运行脚本。 + + + +```bash +cd /workspace/examples/mobilenetv2/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh mobilenetv2-12 +./pegasus_quantize.sh mobilenetv2-12 pcq 10 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh mobilenetv2-12 pcq a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh mobilenetv2-12 pcq t527 +``` + + + + + + +导出的模型文件存放在../model目录。 + +### 编译示例 + +接下来可以编译示例，**先 exit 退出容器**，然后执行下面的命令编译示例。 + +首先需要配置第三方库和交叉编译工具链。 + +:::info +如果你已经在其他示例中配置过第三方库和交叉编译工具链则可以跳过这一步。 +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +需要先手动[点击链接](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4)下载之后放到 0-toolchains/ 再执行下面的命令： + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/mobilenetv2/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## 模型部署 + +编译示例完成之后，示例会安装到 install 目录，可以使用 scp 传输到板端。 + +### 配置 NPU 驱动 + +:::info +如果你已经在其他示例中配置过 NPU 驱动则可以跳过这一步。 +::: + +将驱动库 scp 传输到板端的 lib 目录。 + +- A733 对应 common/lib_linux_aarch64/A733 目录 +- T527 对应 common/lib_linux_aarch64/T527 目录 + +然后执行下面的命令导出到环境变量。 + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### 运行示例 + +配置好驱动之后就可以运行示例了。 + +:::tip +对于 T527 平台，你还需要参考 A5E 的`板端启用 NPU`文档先启用 NPU ，然后使用下面的命令增加当前用户使用 /dev/vipcore 的权限。 +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd mobilenetv2_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./mobilenetv2_demo_a733 +./mobilenetv2_demo_a733 -nb model/mobilenetv2-12_pcq_a733.nb -i model/1.jpg +``` + + + +运行结果如下： + +```bash +$ ./mobilenetv2_demo_a733 -nb model/mobilenetv2-12_pcq_a733.nb -i model/1.jpg +model_file=model/mobilenetv2-12_pcq_a733.nb, input=model/1.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 224 224 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 1000 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/mobilenetv2-12_pcq_a733.nb, size: 3617080. +create network 0: 6194 us. +prepare network: 1502 us. +network: 0, loop count: 1 +run time for this network 0: 2028 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 281, prob: 14.801300, label: tabby, tabby cat +class id: 282, prob: 13.230242, label: tiger cat +class id: 285, prob: 12.491362, label: Egyptian cat +class id: 287, prob: 8.243347, label: lynx, catamount +class id: 478, prob: 8.116148, label: carton +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :---------- | :--------- | :----------- | :----------- | :----------- | :--------- | :----- | :-------- | +| 全志 A733 | Vivante VIP9000 | mobilenetv2 | 224×224 | 6.2 ms | 1.5 ms | 2.0 ms | | 9.7 ms | 103.1 FPS | + + + + + + + +```bash +cd mobilenetv2_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./mobilenetv2_demo_t527 +./mobilenetv2_demo_t527 -nb model/mobilenetv2-12_pcq_t527.nb -i model/1.jpg +``` + + + +运行结果如下： + +```bash +$ ./mobilenetv2_demo_t527 -nb model/mobilenetv2-12_pcq_t527.nb -i model/1.jpg +model_file=model/mobilenetv2-12_pcq_t527.nb, input=model/1.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 224 224 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 1000 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/mobilenetv2-12_pcq_t527.nb, size: 3515008. +create network 0: 6479 us. +prepare network: 1872 us. +network: 0, loop count: 1 +run time for this network 0: 3127 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 281, prob: 14.506104, label: tabby, tabby cat +class id: 282, prob: 13.106445, label: tiger cat +class id: 285, prob: 12.215820, label: Egyptian cat +class id: 287, prob: 8.143799, label: lynx, catamount +class id: 478, prob: 7.889282, label: carton +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :---------- | :--------- | :----------- | :----------- | :----------- | :--------- | :------ | :------- | +| 全志 T527 | Vivante VIP9000 | mobilenetv2 | 224×224 | 6.5 ms | 1.9 ms | 3.1 ms | | 11.5 ms | 87.0 FPS | + + + diff --git a/docs/common/ai/cubie/_model-zoo-ppseg.mdx b/docs/common/ai/cubie/_model-zoo-ppseg.mdx new file mode 100644 index 000000000..7f66e98fd --- /dev/null +++ b/docs/common/ai/cubie/_model-zoo-ppseg.mdx @@ -0,0 +1,311 @@ +本文档讲述如何在 NPU 上运行 PPSeg。 + +:::info +参考 [Model Zoo 下载](./model-zoo-download)获取示例。 +::: + +PPSeg 示例目录结构： + +```bash +$ tree ./ +./ +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ ├── model.pdparams +│ └── pp_liteseg_cityscapes.txt +├── figures +│ └── out_ppseg.png +├── main.cpp +├── model +│ └── munster_000022_000019_leftImg8bit.png +├── model_config.h +├── ppseg_post.cpp +├── ppseg_pre.cpp +└── README.md +``` + +## 模型转换 + +### 下载模型 + +点击下载 [pp_liteseg_cityscapes](http://netstorage.allwinnertech.com:5000/sharing/H2GNL3AuC) 。 + +然后将模型移动到 convert_model/ 目录下。 + +### 创建转换脚本的软链接 + + + +```bash +cd convert_model/ +./convert_model_env.sh +``` + + + +### 模型导入/量化/转换 + +需要先进入容器开发环境。可以参考 Model Zoo 下载中[创建容器](./model-zoo-download#创建并启动容器)这一部分。 + +:::info +不同平台请使用对应的 Docker 镜像： + +- A733：ubuntu-npu:v2.0.10.1 +- T527：ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +进入容器对应目录之后运行脚本。 + + + +```bash +cd /workspace/examples/ppseg/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh pp_liteseg_cityscapes +./pegasus_quantize.sh pp_liteseg_cityscapes pcq 10 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh pp_liteseg_cityscapes pcq a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh pp_liteseg_cityscapes pcq t527 +``` + + + + + + +导出的模型文件存放在../model目录。 + +### 编译示例 + +接下来可以编译示例，**先 exit 退出容器**，然后执行下面的命令编译示例。 + +首先需要配置第三方库和交叉编译工具链。 + +:::info +如果你已经在其他示例中配置过第三方库和交叉编译工具链则可以跳过这一步。 +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +需要先手动[点击链接](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4)下载之后放到 0-toolchains/ 再执行下面的命令： + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/ppseg/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## 模型部署 + +编译示例完成之后，示例会安装到 install 目录，可以使用 scp 传输到板端。 + +### 配置 NPU 驱动 + +:::info +如果你已经在其他示例中配置过 NPU 驱动则可以跳过这一步。 +::: + +将驱动库 scp 传输到板端的 lib 目录。 + +- A733 对应 common/lib_linux_aarch64/A733 目录 +- T527 对应 common/lib_linux_aarch64/T527 目录 + +然后执行下面的命令导出到环境变量。 + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### 运行示例 + +配置好驱动之后就可以运行示例了。 + +:::tip +对于 T527 平台，你还需要参考 A5E 的`板端启用 NPU`文档先启用 NPU ，然后使用下面的命令增加当前用户使用 /dev/vipcore 的权限。 +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd ppseg_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./ppseg_demo_a733 +./ppseg_demo_a733 -nb model/pp_liteseg_cityscapes_pcq_a733.nb -i model/munster_000022_000019_leftImg8bit.png +``` + + + +运行结果如下： + +```bash +$ ./ppseg_demo_a733 -nb model/pp_liteseg_cityscapes_pcq_a733.nb -i model/munster_000022_000019_leftImg8bit.png +model_file=model/pp_liteseg_cityscapes_pcq_a733.nb, input=model/munster_000022_000019_leftImg8bit.png, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 512 512 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 512 512 19 1, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/pp_liteseg_cityscapes_pcq_a733.nb, size: 10615496. +create network 0: 30474 us. +prepare network: 1578 us. +buffer ptr: 0x190c0300, buffer size: 786432 +network: 0, loop count: 1 +run time for this network 0: 81842 us. +output 0, ptr 0xffffa2a2b040, size 4980736. +post process time : 146 ms +ppseg_postprocess finished. +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :-------------------- | :--------- | :----------- | :----------- | :----------- | :--------- | :----- | :------ | +| 全志 A733 | Vivante VIP9000 | pp_liteseg_cityscapes | 512×512 | 30.5 ms | 1.6 ms | 81.8 ms | 146.0 ms | 260 ms | 3.8 FPS | + + + + + + + +```bash +cd ppseg_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./ppseg_demo_t527 +./ppseg_demo_t527 -nb model/pp_liteseg_cityscapes_pcq_t527.nb -i model/munster_000022_000019_leftImg8bit.png +``` + + + +运行结果如下： + +```bash +$ ./ppseg_demo_t527 -nb model/pp_liteseg_cityscapes_pcq_t527.nb -i model/munster_000022_000019_leftImg8bit.png +model_file=model/pp_liteseg_cityscapes_pcq_t527.nb, input=model/munster_000022_000019_leftImg8bit.png, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 512 512 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 512 512 19 1, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/pp_liteseg_cityscapes_pcq_t527.nb, size: 10894912. +create network 0: 39396 us. +prepare network: 6386 us. +buffer ptr: 0x2fa182c0, buffer size: 786432 +network: 0, loop count: 1 +run time for this network 0: 122524 us. +output 0, ptr 0xffff8f7d9040, size 4980736. +post process time : 346 ms +ppseg_postprocess finished. +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :-------------------- | :--------- | :----------- | :----------- | :----------- | :--------- | :------- | :------ | +| 全志 T527 | Vivante VIP9000 | pp_liteseg_cityscapes | 512×512 | 39.4 ms | 6.4 ms | 122.5 ms | 346.0 ms | 514.3 ms | 1.9 FPS | + + + diff --git a/docs/common/ai/cubie/_model-zoo-resnet50-tflite.mdx b/docs/common/ai/cubie/_model-zoo-resnet50-tflite.mdx new file mode 100644 index 000000000..90842c607 --- /dev/null +++ b/docs/common/ai/cubie/_model-zoo-resnet50-tflite.mdx @@ -0,0 +1,317 @@ +本文档讲述如何在 NPU 上运行 ResNet50 TFLite。 + +:::info +参考 [Model Zoo 下载](./model-zoo-download)获取示例。 +::: + +ResNet50 TFLite 示例目录结构： + +```bash +$ tree ./ +./ +├── class_post.cpp +├── class_pre.cpp +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ └── convert_model_env.sh +├── label.h +├── main.cpp +├── model +│ └── space_shuttle_224x224.jpg +└── README.md +``` + +## 模型转换 + +### 下载模型 + + + +```bash +cd convert_model/ +wget https://huggingface.co/qualcomm/ResNet50/resolve/18ab0a0ae3c14bc3ee7006c017f12802ab89cdf2/ResNet50.tflite +``` + + + +### 创建转换脚本的软链接 + + + +```bash +./convert_model_env.sh +``` + + + +### 模型导入/量化/转换 + +需要先进入容器开发环境。可以参考 Model Zoo 下载中[创建容器](./model-zoo-download#创建并启动容器)这一部分。 + +:::info +不同平台请使用对应的 Docker 镜像： + +- A733：ubuntu-npu:v2.0.10.1 +- T527：ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +进入容器对应目录之后运行脚本。 + + + +```bash +cd /workspace/examples/resnet50/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh ResNet50 +./pegasus_quantize.sh ResNet50 uint8 10 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh ResNet50 uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh ResNet50 uint8 t527 +``` + + + + + + +导出的模型文件存放在../model目录。 + +### 编译示例 + +接下来可以编译示例，**先 exit 退出容器**，然后执行下面的命令编译示例。 + +首先需要配置第三方库和交叉编译工具链。 + +:::info +如果你已经在其他示例中配置过第三方库和交叉编译工具链则可以跳过这一步。 +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +需要先手动[点击链接](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4)下载之后放到 0-toolchains/ 再执行下面的命令： + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/resnet50/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## 模型部署 + +编译示例完成之后，示例会安装到 install 目录，可以使用 scp 传输到板端。 + +### 配置 NPU 驱动 + +:::info +如果你已经在其他示例中配置过 NPU 驱动则可以跳过这一步。 +::: + +将驱动库 scp 传输到板端的 lib 目录。 + +- A733 对应 common/lib_linux_aarch64/A733 目录 +- T527 对应 common/lib_linux_aarch64/T527 目录 + +然后执行下面的命令导出到环境变量。 + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### 运行示例 + +配置好驱动之后就可以运行示例了。 + +:::tip +对于 T527 平台，你还需要参考 A5E 的`板端启用 NPU`文档先启用 NPU ，然后使用下面的命令增加当前用户使用 /dev/vipcore 的权限。 +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd resnet50_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./resnet50_demo_a733 +./resnet50_demo_a733 -nb model/ResNet50_uint8_a733.nb -i model/space_shuttle_224x224.jpg +``` + + + +运行结果如下： + +```bash +$ ./resnet50_demo_a733 -nb model/ResNet50_uint8_a733.nb -i model/space_shuttle_224x224.jpg +model_file=model/ResNet50_uint8_a733.nb, input=model/space_shuttle_224x224.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 224 224 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 1000 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/ResNet50_uint8_a733.nb, size: 16737832. +create network 0: 19026 us. +prepare network: 646 us. +network: 0, loop count: 1 +run time for this network 0: 7255 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 812, prob: 26.843445, label: space shuttle +class id: 404, prob: 12.652412, label: airliner +class id: 867, prob: 11.439104, label: trailer truck, tractor trailer, trucking rig, rig, articulated lorry, semi +class id: 833, prob: 10.994865, label: submarine, pigboat, sub, U-boat +class id: 675, prob: 9.653847, label: moving van +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :------- | :--------- | :----------- | :----------- | :----------- | :--------- | :------ | :------- | +| 全志 A733 | Vivante VIP9000 | resnet50 | 224×224 | 19.0 ms | 0.6 ms | 7.3 ms | | 26.9 ms | 37.2 FPS | + + + + + + + +```bash +cd resnet50_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./resnet50_demo_t527 +./resnet50_demo_t527 -nb model/ResNet50_uint8_t527.nb -i model/space_shuttle_224x224.jpg +``` + + + +运行结果如下： + +```bash +$ ./resnet50_demo_t527 -nb model/ResNet50_uint8_t527.nb -i model/space_shuttle_224x224.jpg +model_file=model/ResNet50_uint8_t527.nb, input=model/space_shuttle_224x224.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 224 224 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 1000 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/ResNet50_uint8_t527.nb, size: 16724480. +create network 0: 21618 us. +prepare network: 2776 us. +network: 0, loop count: 1 +run time for this network 0: 14320 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 812, prob: 26.991699, label: space shuttle +class id: 404, prob: 12.526367, label: airliner +class id: 867, prob: 11.482666, label: trailer truck, tractor trailer, trucking rig, rig, articulated lorry, semi +class id: 833, prob: 11.184326, label: submarine, pigboat, sub, U-boat +class id: 675, prob: 9.693115, label: moving van +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :------- | :--------- | :----------- | :----------- | :----------- | :--------- | :------ | :------- | +| 全志 T527 | Vivante VIP9000 | resnet50 | 224×224 | 21.6 ms | 2.8 ms | 14.3 ms | | 38.7 ms | 25.8 FPS | + + + diff --git a/docs/common/ai/cubie/_model-zoo-resnet50v2.mdx b/docs/common/ai/cubie/_model-zoo-resnet50v2.mdx new file mode 100644 index 000000000..7338ce57a --- /dev/null +++ b/docs/common/ai/cubie/_model-zoo-resnet50v2.mdx @@ -0,0 +1,313 @@ +本文档讲述如何在 NPU 上运行 ResNet50 V2。 + +:::info +参考 [Model Zoo 下载](./model-zoo-download)获取示例。 +::: + +ResNet50 V2 示例目录结构： + +```bash +$ tree ./ +./ +├── class_post.cpp +├── class_pre.cpp +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ └── convert_model_env.sh +├── label.h +├── main.cpp +├── model +│ └── 1.jpg +└── README.md +``` + +## 模型转换 + +### 下载模型 + +点击下载 [resnet50v2.onnx](http://netstorage.allwinnertech.com:5000/sharing/WQ6FcVgs5) 。 + +然后将模型移动到 convert_model/ 目录下。 + +### 创建转换脚本的软链接 + + + +```bash +cd convert_model/ +./convert_model_env.sh +``` + + + +### 模型导入/量化/转换 + +需要先进入容器开发环境。可以参考 Model Zoo 下载中[创建容器](./model-zoo-download#创建并启动容器)这一部分。 + +:::info +不同平台请使用对应的 Docker 镜像： + +- A733：ubuntu-npu:v2.0.10.1 +- T527：ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +进入容器对应目录之后运行脚本。 + + + +```bash +cd /workspace/examples/resnet50v2/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh resnet50v2 +./pegasus_quantize.sh resnet50v2 uint8 10 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh resnet50v2 uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh resnet50v2 uint8 t527 +``` + + + + + + +导出的模型文件存放在../model目录。 + +### 编译示例 + +接下来可以编译示例，**先 exit 退出容器**，然后执行下面的命令编译示例。 + +首先需要配置第三方库和交叉编译工具链。 + +:::info +如果你已经在其他示例中配置过第三方库和交叉编译工具链则可以跳过这一步。 +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +需要先手动[点击链接](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4)下载之后放到 0-toolchains/ 再执行下面的命令： + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/resnet50v2/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## 模型部署 + +编译示例完成之后，示例会安装到 install 目录，可以使用 scp 传输到板端。 + +### 配置 NPU 驱动 + +:::info +如果你已经在其他示例中配置过 NPU 驱动则可以跳过这一步。 +::: + +将驱动库 scp 传输到板端的 lib 目录。 + +- A733 对应 common/lib_linux_aarch64/A733 目录 +- T527 对应 common/lib_linux_aarch64/T527 目录 + +然后执行下面的命令导出到环境变量。 + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### 运行示例 + +配置好驱动之后就可以运行示例了。 + +:::tip +对于 T527 平台，你还需要参考 A5E 的`板端启用 NPU`文档先启用 NPU ，然后使用下面的命令增加当前用户使用 /dev/vipcore 的权限。 +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd resnet50v2_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./resnet50v2_demo_a733 +./resnet50v2_demo_a733 -nb model/resnet50v2_uint8_a733.nb -i model/1.jpg +``` + + + +运行结果如下： + +```bash +$ ./resnet50v2_demo_a733 -nb model/resnet50v2_uint8_a733.nb -i model/1.jpg +model_file=model/resnet50v2_uint8_a733.nb, input=model/1.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 224 224 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 1000 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/resnet50v2_uint8_a733.nb, size: 17593328. +create network 0: 15664 us. +prepare network: 1734 us. +network: 0, loop count: 1 +run time for this network 0: 8900 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 281, prob: 11.564558, label: tabby, tabby cat +class id: 285, prob: 10.988928, label: Egyptian cat +class id: 282, prob: 9.769331, label: tiger cat +class id: 287, prob: 5.913674, label: lynx, catamount +class id: 292, prob: 4.939133, label: tiger, Panthera tigris +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :--------- | :--------- | :----------- | :----------- | :----------- | :--------- | :------ | :------- | +| 全志 A733 | Vivante VIP9000 | resnet50v2 | 224×224 | 15.7 ms | 1.7 ms | 8.9 ms | | 26.3 ms | 38.0 FPS | + + + + + + + +```bash +cd resnet50v2_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./resnet50v2_demo_t527 +./resnet50v2_demo_t527 -nb model/resnet50v2_uint8_t527.nb -i model/1.jpg +``` + + + +运行结果如下： + +```bash +$ ./resnet50v2_demo_t527 -nb model/resnet50v2_uint8_t527.nb -i model/1.jpg +model_file=model/resnet50v2_uint8_t527.nb, input=model/1.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 224 224 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 1000 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/resnet50v2_uint8_t527.nb, size: 17309120. +create network 0: 27251 us. +prepare network: 3129 us. +network: 0, loop count: 1 +run time for this network 0: 16298 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 281, prob: 11.682373, label: tabby, tabby cat +class id: 285, prob: 11.270020, label: Egyptian cat +class id: 282, prob: 10.033203, label: tiger cat +class id: 287, prob: 6.047363, label: lynx, catamount +class id: 292, prob: 5.085327, label: tiger, Panthera tigris +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :--------- | :--------- | :----------- | :----------- | :----------- | :--------- | :------ | :------- | +| 全志 T527 | Vivante VIP9000 | resnet50v2 | 224×224 | 27.3 ms | 3.1 ms | 16.3 ms | | 46.7 ms | 21.4 FPS | + + + diff --git a/docs/common/ai/cubie/_model-zoo-retinaface.mdx b/docs/common/ai/cubie/_model-zoo-retinaface.mdx new file mode 100644 index 000000000..613dcfa5b --- /dev/null +++ b/docs/common/ai/cubie/_model-zoo-retinaface.mdx @@ -0,0 +1,337 @@ +本文档讲述如何在 NPU 上运行 RetinaFace。 + +:::info +参考 [Model Zoo 下载](./model-zoo-download)获取示例。 +::: + +RetinaFace 示例目录结构： + +```bash +$ tree ./ +./ +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ └── Retinaface_resnet50_320.txt +├── figures +│ └── out_retinaface.png +├── main.cpp +├── model +│ └── test.jpg +├── model_config.h +├── README.md +├── retinaface_post.cpp +└── retinaface_pre.cpp +``` + +## 模型转换 + +### 导出 onnx 模型 + +点击下载 [Resnet50_Final.pth](https://drive.google.com/drive/folders/1oZRSG0ZegbVkVwUd8wUIQx8W7yfZ_ki1) 。 + +### 下载 onnx 模型 + +可以下载修改好的模型。 + +点击下载 [Retinaface_resnet50_320.onnx](http://netstorage.allwinnertech.com:5000/sharing/YObunQV8S) 。 + +点击下载 [Retinaface_mobilenet0.25_320.onnx](http://netstorage.allwinnertech.com:5000/sharing/xt9rDVXzI) 。 + +然后移动到 convert_model/ 目录下。 + +### 创建转换脚本的软链接 + + + +```bash +./convert_model_env.sh +``` + + + +### 模型导入/量化/转换 + +需要先进入容器开发环境。可以参考 Model Zoo 下载中[创建容器](./model-zoo-download#创建并启动容器)这一部分。 + +:::info +不同平台请使用对应的 Docker 镜像： + +- A733：ubuntu-npu:v2.0.10.1 +- T527：ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +进入容器对应目录之后运行脚本。 + + + +```bash +cd /workspace/examples/retinaface/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh Retinaface_resnet50_320 +./pegasus_quantize.sh Retinaface_resnet50_320 uint8 10 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh Retinaface_resnet50_320 uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh Retinaface_resnet50_320 uint8 t527 +``` + + + + + + +导出的模型文件存放在../model目录。 + +### 编译示例 + +接下来可以编译示例，**先 exit 退出容器**，然后执行下面的命令编译示例。 + +首先需要配置第三方库和交叉编译工具链。 + +:::info +如果你已经在其他示例中配置过第三方库和交叉编译工具链则可以跳过这一步。 +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +需要先手动[点击链接](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4)下载之后放到 0-toolchains/ 再执行下面的命令： + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/retinaface/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## 模型部署 + +编译示例完成之后，示例会安装到 install 目录，可以使用 scp 传输到板端。 + +### 配置 NPU 驱动 + +:::info +如果你已经在其他示例中配置过 NPU 驱动则可以跳过这一步。 +::: + +将驱动库 scp 传输到板端的 lib 目录。 + +- A733 对应 common/lib_linux_aarch64/A733 目录 +- T527 对应 common/lib_linux_aarch64/T527 目录 + +然后执行下面的命令导出到环境变量。 + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### 运行示例 + +配置好驱动之后就可以运行示例了。 + +:::tip +对于 T527 平台，你还需要参考 A5E 的`板端启用 NPU`文档先启用 NPU ，然后使用下面的命令增加当前用户使用 /dev/vipcore 的权限。 +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd retinaface_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./retinaface_demo_a733 +./retinaface_demo_a733 -nb model/Retinaface_resnet50_320_uint8_a733.nb -i model/test.jpg +``` + + + +运行结果如下： + +```bash +$ ./retinaface_demo_a733 -nb model/Retinaface_resnet50_320_uint8_a733.nb -i model/test.jpg +model_file=model/Retinaface_resnet50_320_uint8_a733.nb, input=model/test.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 320 320 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 4 4200 1 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 2 4200 1 0, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 10 4200 1 0, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +nbg name=model/Retinaface_resnet50_320_uint8_a733.nb, size: 19056048. +create network 0: 20781 us. +prepare network: 2285 us. +buffer ptr: 0x25971380, buffer size: 307200 +network: 0, loop count: 1 +run time for this network 0: 15703 us. +output 0, ptr 0x259bc480, size 16800. +output 1, ptr 0x259ccb80, size 8400. +output 2, ptr 0x259d4f40, size 42000. +post process time : 0 ms +detection num: 1 +100%, [ 244, 46, 363, 209], face +275.10 113.49 +328.03 112.26 +300.95 147.94 +277.57 165.17 +326.80 165.17 +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :------------------ | :--------- | :----------- | :----------- | :----------- | :--------- | :------ | :------- | +| 全志 A733 | Vivante VIP9000 | Retinaface_resnet50 | 320×320 | 20.8 ms | 2.3 ms | 15.7 ms | 0.0 ms | 38.8 ms | 25.8 FPS | + + + + + + + +```bash +cd retinaface_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./retinaface_demo_t527 +./retinaface_demo_t527 -nb model/Retinaface_resnet50_320_uint8_t527.nb -i model/test.jpg +``` + + + +运行结果如下： + +```bash +$ ./retinaface_demo_t527 -nb model/Retinaface_resnet50_320_uint8_t527.nb -i model/test.jpg +model_file=model/Retinaface_resnet50_320_uint8_t527.nb, input=model/test.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 320 320 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 4 4200 1 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 2 4200 1 0, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 10 4200 1 0, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +nbg name=model/Retinaface_resnet50_320_uint8_t527.nb, size: 18714688. +create network 0: 27602 us. +prepare network: 5276 us. +buffer ptr: 0x23c57380, buffer size: 307200 +network: 0, loop count: 1 +run time for this network 0: 30483 us. +output 0, ptr 0x23ca2440, size 16800. +output 1, ptr 0x23cb2b40, size 8400. +output 2, ptr 0x23cbaf40, size 42000. +post process time : 1 ms +detection num: 1 +100%, [ 244, 45, 363, 208], face +275.10 113.49 +328.03 112.26 +300.95 147.94 +277.57 166.40 +326.80 165.17 +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :------------------ | :--------- | :----------- | :----------- | :----------- | :--------- | :------ | :------- | +| 全志 T527 | Vivante VIP9000 | Retinaface_resnet50 | 320×320 | 27.6 ms | 5.3 ms | 30.5 ms | 1.0 ms | 64.4 ms | 15.5 FPS | + + + diff --git a/docs/common/ai/cubie/_model-zoo-squeezenet-pytorch.mdx b/docs/common/ai/cubie/_model-zoo-squeezenet-pytorch.mdx new file mode 100644 index 000000000..7583be5e8 --- /dev/null +++ b/docs/common/ai/cubie/_model-zoo-squeezenet-pytorch.mdx @@ -0,0 +1,314 @@ +本文档讲述如何在 NPU 上运行 SqueezeNet。 + +:::info +参考 [Model Zoo 下载](./model-zoo-download)获取示例。 +::: + +SqueezeNet 示例目录结构： + +```bash +$ tree ./ +./ +├── class_post.cpp +├── class_pre.cpp +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ └── inputs_outputs.txt +├── label.h +├── main.cpp +├── model +│ └── space_shuttle_227x227.jpg +└── README.md +``` + +## 模型转换 + +### 下载模型 + +点击下载 [squeezenet1_0.pt](http://netstorage.allwinnertech.com:5000/sharing/bSFCrqSqC) 。 + +然后将模型移动到 convert_model/ 目录下。 + +### 创建转换脚本的软链接 + + + +```bash +cd convert_model/ +./convert_model_env.sh +``` + + + +### 模型导入/量化/转换 + +需要先进入容器开发环境。可以参考 Model Zoo 下载中[创建容器](./model-zoo-download#创建并启动容器)这一部分。 + +:::info +不同平台请使用对应的 Docker 镜像： + +- A733：ubuntu-npu:v2.0.10.1 +- T527：ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +进入容器对应目录之后运行脚本。 + + + +```bash +cd /workspace/examples/squeezenet_pytorch/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh squeezenet1_0 +./pegasus_quantize.sh squeezenet1_0 uint8 10 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh squeezenet1_0 uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh squeezenet1_0 uint8 t527 +``` + + + + + + +导出的模型文件存放在../model目录。 + +### 编译示例 + +接下来可以编译示例，**先 exit 退出容器**，然后执行下面的命令编译示例。 + +首先需要配置第三方库和交叉编译工具链。 + +:::info +如果你已经在其他示例中配置过第三方库和交叉编译工具链则可以跳过这一步。 +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +需要先手动[点击链接](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4)下载之后放到 0-toolchains/ 再执行下面的命令： + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/squeezenet_pytorch/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## 模型部署 + +编译示例完成之后，示例会安装到 install 目录，可以使用 scp 传输到板端。 + +### 配置 NPU 驱动 + +:::info +如果你已经在其他示例中配置过 NPU 驱动则可以跳过这一步。 +::: + +将驱动库 scp 传输到板端的 lib 目录。 + +- A733 对应 common/lib_linux_aarch64/A733 目录 +- T527 对应 common/lib_linux_aarch64/T527 目录 + +然后执行下面的命令导出到环境变量。 + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### 运行示例 + +配置好驱动之后就可以运行示例了。 + +:::tip +对于 T527 平台，你还需要参考 A5E 的`板端启用 NPU`文档先启用 NPU ，然后使用下面的命令增加当前用户使用 /dev/vipcore 的权限。 +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd squeezenet_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./squeezenet_demo_a733 +./squeezenet_demo_a733 -nb model/squeezenet1_0_uint8_a733.nb -i model/space_shuttle_227x227.jpg +``` + + + +运行结果如下： + +```bash +$ ./squeezenet_demo_a733 -nb model/squeezenet1_0_uint8_a733.nb -i model/space_shuttle_227x227.jpg +model_file=model/squeezenet1_0_uint8_a733.nb, input=model/space_shuttle_227x227.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 227 227 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 1000 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/squeezenet1_0_uint8_a733.nb, size: 1066072. +create network 0: 2628 us. +prepare network: 844 us. +network: 0, loop count: 1 +run time for this network 0: 2455 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 812, prob: 16.893208, label: space shuttle +class id: 404, prob: 15.714762, label: airliner +class id: 833, prob: 14.979437, label: submarine, pigboat, sub, U-boat +class id: 554, prob: 14.664755, label: fireboat +class id: 895, prob: 14.632645, label: warplane, military plane +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :--------- | :--------- | :----------- | :----------- | :----------- | :--------- | :----- | :-------- | +| 全志 A733 | Vivante VIP9000 | squeezenet | 227×227 | 2.6 ms | 0.8 ms | 2.5 ms | | 5.9 ms | 169.5 FPS | + + + + + + + +```bash +cd squeezenet_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./squeezenet_demo_t527 +./squeezenet_demo_t527 -nb model/squeezenet1_0_uint8_t527.nb -i model/space_shuttle_227x227.jpg +``` + + + +运行结果如下： + +```bash +$ ./squeezenet_demo_t527 -nb model/squeezenet1_0_uint8_t527.nb -i model/space_shuttle_227x227.jpg +model_file=model/squeezenet1_0_uint8_t527.nb, input=model/space_shuttle_227x227.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 227 227 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 1000 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/squeezenet1_0_uint8_t527.nb, size: 1078912. +create network 0: 2536 us. +prepare network: 1301 us. +network: 0, loop count: 1 +run time for this network 0: 3103 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 812, prob: 16.922852, label: space shuttle +class id: 404, prob: 15.730225, label: airliner +class id: 833, prob: 14.984619, label: submarine, pigboat, sub, U-boat +class id: 554, prob: 14.686523, label: fireboat +class id: 895, prob: 14.611816, label: warplane, military plane +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :--------- | :--------- | :----------- | :----------- | :----------- | :--------- | :----- | :-------- | +| 全志 T527 | Vivante VIP9000 | squeezenet | 227×227 | 2.5 ms | 1.3 ms | 3.1 ms | | 6.9 ms | 144.9 FPS | + + + diff --git a/docs/common/ai/cubie/_model-zoo-yolo11-pose.mdx b/docs/common/ai/cubie/_model-zoo-yolo11-pose.mdx new file mode 100644 index 000000000..f7b19ac13 --- /dev/null +++ b/docs/common/ai/cubie/_model-zoo-yolo11-pose.mdx @@ -0,0 +1,481 @@ +本文档讲述如何在 NPU 上运行 YOLO11 Pose。 + +:::info +参考 [Model Zoo 下载](./model-zoo-download)获取示例。 +::: + +YOLO11 Pose 示例目录结构： + +```bash +$ tree ./ +./ +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ ├── python +│ │ ├── onnx_extract.py +│ │ └── yolo11s-pose_640.txt +│ └── yolo11s-pose_9.txt +├── figures +│ ├── diff_img.png +│ └── out_yolo11_pose_pcq.png +├── main.cpp +├── model +│ └── COCO_train2014_000000500390.jpg +├── model_config.h +├── README.md +├── yolo11_pose_9_post.cpp +└── yolo11_pose_9_pre.cpp +``` + +## 模型转换 + +### 配置虚拟环境 + + + +```bash +python -m venv .venv && source .venv/bin/activate +pip install ultralytics +``` + + + +### 导出 onnx 模型 + + + +```bash +cd convert_model/python/ +yolo export model=yolo11s-pose.pt format=onnx simplify=True dynamic=False opset=11 nms=False batch=1 device=cpu +``` + + + +### 裁剪模型 + + + +```bash +python onnx_extract.py +mv ./yolo11s-pose_9.onnx ../ +cd .. +``` + + + +### 创建转换脚本的软链接 + + + +```bash +./convert_model_env.sh +``` + + + +### 模型导入/量化/转换 + +需要先进入容器开发环境。可以参考 Model Zoo 下载中[创建容器](./model-zoo-download#创建并启动容器)这一部分。 + +:::info +不同平台请使用对应的 Docker 镜像： + +- A733：ubuntu-npu:v2.0.10.1 +- T527：ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +进入容器对应目录之后运行脚本。 + + + +```bash +cd /workspace/examples/yolo11_pose/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh yolo11s-pose_9 +./pegasus_quantize.sh yolo11s-pose_9 uint8 12 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolo11s-pose_9 uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolo11s-pose_9 uint8 t527 +``` + + + + + + +导出的模型文件存放在../model目录。 + +### 编译示例 + +接下来可以编译示例，**先 exit 退出容器**，然后执行下面的命令编译示例。 + +首先需要配置第三方库和交叉编译工具链。 + +:::info +如果你已经在其他示例中配置过第三方库和交叉编译工具链则可以跳过这一步。 +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +需要先手动[点击链接](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4)下载之后放到 0-toolchains/ 再执行下面的命令： + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/yolo11_pose/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## 模型部署 + +编译示例完成之后，示例会安装到 install 目录，可以使用 scp 传输到板端。 + +### 配置 NPU 驱动 + +:::info +如果你已经在其他示例中配置过 NPU 驱动则可以跳过这一步。 +::: + +将驱动库 scp 传输到板端的 lib 目录。 + +- A733 对应 common/lib_linux_aarch64/A733 目录 +- T527 对应 common/lib_linux_aarch64/T527 目录 + +然后执行下面的命令导出到环境变量。 + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### 运行示例 + +配置好驱动之后就可以运行示例了。 + +:::tip +对于 T527 平台，你还需要参考 A5E 的`板端启用 NPU`文档先启用 NPU ，然后使用下面的命令增加当前用户使用 /dev/vipcore 的权限。 +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd yolo11_pose_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./yolo11_pose_demo_a733 +./yolo11_pose_demo_a733 -nb model/yolo11s-pose_9_uint8_a733.nb -i model/COCO_train2014_000000500390.jpg +``` + + + +运行结果如下： + +```bash +$ ./yolo11_pose_demo_a733 -nb model/yolo11s-pose_9_uint8_a733.nb -i model/COCO_train2014_000000500390.jpg +model_file=model/yolo11s-pose_9_uint8_a733.nb, input=model/COCO_train2014_000000500390.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 80 80 64 1, data_format=0, name=uid_17_out_0b_uid_1_out_0, none-quant +output 1 dim 80 80 1 1, data_format=0, name=uid_16_out_0b_uid_1_out_0, none-quant +output 2 dim 80 80 51 1, data_format=0, name=uid_15_out_0b_uid_1_out_0, none-quant +output 3 dim 40 40 64 1, data_format=0, name=uid_14_out_0b_uid_1_out_0, none-quant +output 4 dim 40 40 1 1, data_format=0, name=uid_13_out_0b_uid_1_out_0, none-quant +output 5 dim 40 40 51 1, data_format=0, name=uid_12_out_0b_uid_1_out_0, none-quant +output 6 dim 20 20 64 1, data_format=0, name=uid_11_out_0b_uid_1_out_0, none-quant +output 7 dim 20 20 1 1, data_format=0, name=uid_10_out_0b_uid_1_out_0, none-quant +output 8 dim 20 20 51 1, data_format=0, name=uid_9_out_0ub_uid_1_out_0, none-quant +nbg name=model/yolo11s-pose_9_uint8_a733.nb, size: 7284048. +create network 0: 16110 us. +prepare network: 3977 us. +buffer ptr: 0x202f5380, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 32374 us. +output 0, ptr 0x20421480, size 409600. +output 1, ptr 0x205b1500, size 6400. +output 2, ptr 0x205b7980, size 326400. +output 3, ptr 0x206f6640, size 102400. +output 4, ptr 0x2075a6c0, size 1600. +output 5, ptr 0x2075c040, size 81600. +output 6, ptr 0x207abbc0, size 25600. +output 7, ptr 0x207c4c80, size 400. +output 8, ptr 0x207c5340, size 20400. +post process time : 4 ms +detection num: 3 + 0: 94%, [ 370, 0, 589, 346], person +405.75 26.20 = 0.96988 +419.11 23.03 = 0.96501 +405.65 21.63 = 0.29929 +441.04 31.18 = 0.99146 +421.11 22.33 = 0.04379 +455.76 67.51 = 0.99977 +430.35 62.14 = 0.99950 +466.39 121.18 = 0.99797 +405.08 109.99 = 0.98330 +447.50 96.32 = 0.98985 +382.14 70.42 = 0.94582 +466.06 166.69 = 0.99986 +455.44 165.19 = 0.99974 +411.43 242.60 = 0.99939 +497.02 230.87 = 0.99880 +408.66 307.99 = 0.98213 +562.98 301.11 = 0.97806 + 0: 88%, [ 86, 27, 292, 389], person +146.77 66.48 = 0.99659 +157.56 60.56 = 0.99517 +138.54 62.15 = 0.93738 +177.10 58.15 = 0.97191 +136.75 58.00 = 0.22714 +182.16 88.95 = 0.99876 +146.17 100.58 = 0.99755 +210.99 144.46 = 0.99757 +161.48 152.14 = 0.98186 +171.08 179.70 = 0.99453 +131.07 188.60 = 0.97326 +222.98 197.17 = 0.99975 +178.42 204.61 = 0.99950 +250.05 264.09 = 0.99831 +151.41 290.25 = 0.99650 +287.21 296.16 = 0.97016 +127.74 355.67 = 0.95755 + 0: 92%, [ 228, 39, 399, 407], person +275.86 94.61 = 0.99351 +286.44 88.28 = 0.98999 +267.42 87.73 = 0.88035 +308.03 73.30 = 0.97833 +265.48 74.83 = 0.23741 +339.54 98.91 = 0.99963 +280.47 109.74 = 0.99938 +372.16 125.90 = 0.99505 +272.82 170.12 = 0.98157 +380.93 163.22 = 0.98073 +243.21 204.51 = 0.94730 +339.07 225.60 = 0.99986 +302.82 223.45 = 0.99980 +294.02 310.02 = 0.99952 +314.44 286.69 = 0.99926 +270.90 355.43 = 0.99344 +374.00 318.30 = 0.99277 +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :----------- | :--------- | :----------- | :----------- | :----------- | :--------- | :------ | :------- | +| 全志 A733 | Vivante VIP9000 | yolo11s-pose | 640×640 | 16.1 ms | 4.0 ms | 32.4 ms | 4.0 ms | 56.5 ms | 17.7 FPS | + + + + + + + +```bash +cd yolo11_pose_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./yolo11_pose_demo_t527 +./yolo11_pose_demo_t527 -nb model/yolo11s-pose_9_uint8_t527.nb -i model/COCO_train2014_000000500390.jpg +``` + + + +运行结果如下： + +```bash +$ ./yolo11_pose_demo_t527 -nb model/yolo11s-pose_9_uint8_t527.nb -i model/COCO_train2014_000000500390.jpg +model_file=model/yolo11s-pose_9_uint8_t527.nb, input=model/COCO_train2014_000000500390.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 80 80 64 1, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 80 80 1 1, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 80 80 51 1, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +output 3 dim 40 40 64 1, data_format=0, name=uid_20003_sub_uid_1_out_0, none-quant +output 4 dim 40 40 1 1, data_format=0, name=uid_20004_sub_uid_1_out_0, none-quant +output 5 dim 40 40 51 1, data_format=0, name=uid_20005_sub_uid_1_out_0, none-quant +output 6 dim 20 20 64 1, data_format=0, name=uid_20006_sub_uid_1_out_0, none-quant +output 7 dim 20 20 1 1, data_format=0, name=uid_20007_sub_uid_1_out_0, none-quant +output 8 dim 20 20 51 1, data_format=0, name=uid_20008_sub_uid_1_out_0, none-quant +nbg name=model/yolo11s-pose_9_uint8_t527.nb, size: 8148288. +create network 0: 23417 us. +prepare network: 10280 us. +buffer ptr: 0x22f74380, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 75989 us. +output 0, ptr 0x230a0440, size 409600. +output 1, ptr 0x23230500, size 6400. +output 2, ptr 0x23236980, size 326400. +output 3, ptr 0x23375600, size 102400. +output 4, ptr 0x233d9680, size 1600. +output 5, ptr 0x233db040, size 81600. +output 6, ptr 0x2342abc0, size 25600. +output 7, ptr 0x23443c40, size 400. +output 8, ptr 0x23444300, size 20400. +post process time : 11 ms +detection num: 3 + 0: 94%, [ 371, 0, 587, 346], person +406.01 30.36 = 0.96783 +418.34 26.25 = 0.95618 +406.01 26.25 = 0.45198 +434.78 30.36 = 0.98485 +418.34 26.25 = 0.12014 +455.34 71.46 = 0.99981 +426.56 67.35 = 0.99938 +467.67 120.79 = 0.99790 +406.01 104.35 = 0.98050 +451.23 96.12 = 0.98748 +389.57 75.57 = 0.93298 +475.89 174.23 = 0.99984 +459.45 170.12 = 0.99967 +414.23 244.11 = 0.99955 +484.11 235.88 = 0.99903 +410.12 301.65 = 0.99145 +558.10 297.54 = 0.98825 + 0: 87%, [ 86, 27, 292, 389], person +147.67 66.46 = 0.99650 +160.00 58.25 = 0.99518 +139.45 58.25 = 0.94058 +180.55 58.25 = 0.96977 +135.34 54.13 = 0.22788 +180.55 87.02 = 0.99866 +147.67 99.35 = 0.99729 +213.44 144.57 = 0.99729 +164.11 148.68 = 0.98050 +172.33 177.45 = 0.99417 +131.23 189.78 = 0.97160 +221.66 198.00 = 0.99971 +176.44 202.12 = 0.99946 +250.43 263.77 = 0.99816 +151.78 288.44 = 0.99627 +283.32 292.55 = 0.96783 +123.00 354.21 = 0.95618 + 0: 92%, [ 228, 38, 399, 408], person +275.67 96.12 = 0.99247 +288.00 87.90 = 0.98897 +267.45 87.90 = 0.86560 +308.55 75.57 = 0.97646 +267.45 75.57 = 0.22788 +341.44 100.23 = 0.99960 +279.78 108.46 = 0.99930 +374.32 124.90 = 0.99487 +275.67 170.12 = 0.97924 +382.54 161.89 = 0.98050 +242.78 203.00 = 0.94407 +341.44 227.66 = 0.99985 +304.44 223.55 = 0.99978 +292.11 309.88 = 0.99949 +312.66 285.21 = 0.99920 +271.56 355.09 = 0.99294 +374.32 318.10 = 0.99198 +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :----------- | :--------- | :----------- | :----------- | :----------- | :--------- | :------- | :------ | +| 全志 T527 | Vivante VIP9000 | yolo11s-pose | 640×640 | 23.4 ms | 10.3 ms | 76.0 ms | 11.0 ms | 120.7 ms | 8.3 FPS | + + + diff --git a/docs/common/ai/cubie/_model-zoo-yolo11-seg.mdx b/docs/common/ai/cubie/_model-zoo-yolo11-seg.mdx new file mode 100644 index 000000000..03cb32adb --- /dev/null +++ b/docs/common/ai/cubie/_model-zoo-yolo11-seg.mdx @@ -0,0 +1,383 @@ +本文档讲述如何在 NPU 上运行 YOLO11 Seg。 + +:::info +参考 [Model Zoo 下载](./model-zoo-download)获取示例。 +::: + +YOLO11 Seg 示例目录结构： + +```bash +$ tree ./ +./ +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ ├── python +│ │ ├── onnx_extract.py +│ │ └── yolo11s-seg_640.txt +│ └── yolo11s-seg_10.txt +├── figures +│ ├── diff_img.png +│ └── out_yolo11_seg_pcq.png +├── main.cpp +├── model +│ └── dog.jpg +├── model_config.h +├── README.md +├── yolo11_seg_10_post.cpp +└── yolo11_seg_10_pre.cpp +``` + +## 模型转换 + +### 配置虚拟环境 + + + +```bash +python -m venv .venv && source .venv/bin/activate +pip install ultralytics +``` + + + +### 导出 onnx 模型 + + + +```bash +cd convert_model/python/ +yolo export model=yolo11s-seg.pt format=onnx imgsz=640 dynamic=False simplify=True opset=11 nms=False batch=1 device=cpu +``` + + + +### 裁剪模型 + + + +```bash +python onnx_extract.py +mv yolo11s-seg_10.onnx ../ +cd .. +``` + + + +### 创建转换脚本的软链接 + + + +```bash +./convert_model_env.sh +``` + + + +### 模型导入/量化/转换 + +需要先进入容器开发环境。可以参考 Model Zoo 下载中[创建容器](./model-zoo-download#创建并启动容器)这一部分。 + +:::info +不同平台请使用对应的 Docker 镜像： + +- A733：ubuntu-npu:v2.0.10.1 +- T527：ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +进入容器对应目录之后运行脚本。 + + + +```bash +cd /workspace/examples/yolo11_seg/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh yolo11s-seg_10 +./pegasus_quantize.sh yolo11s-seg_10 uint8 12 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolo11s-seg_10 uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolo11s-seg_10 uint8 t527 +``` + + + + + + +导出的模型文件存放在../model目录。 + +### 编译示例 + +接下来可以编译示例，**先 exit 退出容器**，然后执行下面的命令编译示例。 + +首先需要配置第三方库和交叉编译工具链。 + +:::info +如果你已经在其他示例中配置过第三方库和交叉编译工具链则可以跳过这一步。 +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +需要先手动[点击链接](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4)下载之后放到 0-toolchains/ 再执行下面的命令： + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/yolo11_seg/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## 模型部署 + +编译示例完成之后，示例会安装到 install 目录，可以使用 scp 传输到板端。 + +### 配置 NPU 驱动 + +:::info +如果你已经在其他示例中配置过 NPU 驱动则可以跳过这一步。 +::: + +将驱动库 scp 传输到板端的 lib 目录。 + +- A733 对应 common/lib_linux_aarch64/A733 目录 +- T527 对应 common/lib_linux_aarch64/T527 目录 + +然后执行下面的命令导出到环境变量。 + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### 运行示例 + +配置好驱动之后就可以运行示例了。 + +:::tip +对于 T527 平台，你还需要参考 A5E 的`板端启用 NPU`文档先启用 NPU ，然后使用下面的命令增加当前用户使用 /dev/vipcore 的权限。 +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd yolo11_seg_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./yolo11_seg_demo_a733 +./yolo11_seg_demo_a733 -nb model/yolo11s-seg_10_uint8_a733.nb -i model/dog.jpg +``` + + + +运行结果如下： + +```bash +$ ./yolo11_seg_demo_a733 -nb model/yolo11s-seg_10_uint8_a733.nb -i model/dog.jpg +model_file=model/yolo11s-seg_10_uint8_a733.nb, input=model/dog.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 80 80 64 1, data_format=0, name=uid_19_out_0b_uid_1_out_0, none-quant +output 1 dim 80 80 80 1, data_format=0, name=uid_18_out_0b_uid_1_out_0, none-quant +output 2 dim 80 80 32 1, data_format=0, name=uid_17_out_0b_uid_1_out_0, none-quant +output 3 dim 40 40 64 1, data_format=0, name=uid_16_out_0b_uid_1_out_0, none-quant +output 4 dim 40 40 80 1, data_format=0, name=uid_15_out_0b_uid_1_out_0, none-quant +output 5 dim 40 40 32 1, data_format=0, name=uid_14_out_0b_uid_1_out_0, none-quant +output 6 dim 20 20 64 1, data_format=0, name=uid_13_out_0b_uid_1_out_0, none-quant +output 7 dim 20 20 80 1, data_format=0, name=uid_12_out_0b_uid_1_out_0, none-quant +output 8 dim 20 20 32 1, data_format=0, name=uid_11_out_0ub_uid_1_out_0, none-quant +output 9 dim 160 160 32 1, data_format=0, name=uid_20009_sub_uid_1_out_0, none-quant +nbg name=model/yolo11s-seg_10_uint8_a733.nb, size: 7326672. +create network 0: 24693 us. +prepare network: 2986 us. +buffer ptr: 0x226f1600, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 37744 us. +output 0, ptr 0x2281d780, size 409600. +output 1, ptr 0x229ad800, size 512000. +output 2, ptr 0x22ba1880, size 204800. +output 3, ptr 0x22c69900, size 102400. +output 4, ptr 0x22ccd9c0, size 128000. +output 5, ptr 0x22d4aa40, size 51200. +output 6, ptr 0x22d7cac0, size 25600. +output 7, ptr 0x22d95b40, size 32000. +output 8, ptr 0x22db5000, size 12800. +output 9, ptr 0x22dc1880, size 819200. +post process time : 11 ms +detection num: 3 + 1: 95%, [ 126, 126, 568, 420], bicycle +16: 95%, [ 131, 221, 311, 541], dog + 2: 86%, [ 467, 75, 691, 172], car +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :---------- | :--------- | :----------- | :----------- | :----------- | :--------- | :------ | :------- | +| 全志 A733 | Vivante VIP9000 | yolo11s-seg | 640×640 | 24.7 ms | 3.0 ms | 37.7 ms | 11.0 ms | 76.4 ms | 13.1 FPS | + + + + + + + +```bash +cd yolo11_seg_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./yolo11_seg_demo_t527 +./yolo11_seg_demo_t527 -nb model/yolo11s-seg_10_uint8_t527.nb -i model/dog.jpg +``` + + + +运行结果如下： + +```bash +$ ./yolo11_seg_demo_t527 -nb model/yolo11s-seg_10_uint8_t527.nb -i model/dog.jpg +model_file=model/yolo11s-seg_10_uint8_t527.nb, input=model/dog.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 80 80 64 1, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 80 80 80 1, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 80 80 32 1, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +output 3 dim 40 40 64 1, data_format=0, name=uid_20003_sub_uid_1_out_0, none-quant +output 4 dim 40 40 80 1, data_format=0, name=uid_20004_sub_uid_1_out_0, none-quant +output 5 dim 40 40 32 1, data_format=0, name=uid_20005_sub_uid_1_out_0, none-quant +output 6 dim 20 20 64 1, data_format=0, name=uid_20006_sub_uid_1_out_0, none-quant +output 7 dim 20 20 80 1, data_format=0, name=uid_20007_sub_uid_1_out_0, none-quant +output 8 dim 20 20 32 1, data_format=0, name=uid_20008_sub_uid_1_out_0, none-quant +output 9 dim 160 160 32 1, data_format=0, name=uid_20009_sub_uid_1_out_0, none-quant +nbg name=model/yolo11s-seg_10_uint8_t527.nb, size: 8522240. +create network 0: 26153 us. +prepare network: 11813 us. +buffer ptr: 0x38e48600, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 94147 us. +output 0, ptr 0x38f74740, size 409600. +output 1, ptr 0x391047c0, size 512000. +output 2, ptr 0x392f8880, size 204800. +output 3, ptr 0x393c0900, size 102400. +output 4, ptr 0x39424980, size 128000. +output 5, ptr 0x394a1a00, size 51200. +output 6, ptr 0x394d3ac0, size 25600. +output 7, ptr 0x394ecb40, size 32000. +output 8, ptr 0x3950bfc0, size 12800. +output 9, ptr 0x39518840, size 819200. +post process time : 51 ms +detection num: 3 + 1: 94%, [ 126, 124, 568, 420], bicycle +16: 95%, [ 132, 222, 311, 541], dog + 2: 82%, [ 467, 76, 692, 172], car +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :---------- | :--------- | :----------- | :----------- | :----------- | :--------- | :------- | :------ | +| 全志 T527 | Vivante VIP9000 | yolo11s-seg | 640×640 | 26.2 ms | 11.8 ms | 94.1 ms | 51.0 ms | 183.1 ms | 5.5 FPS | + + + diff --git a/docs/common/ai/cubie/_model-zoo-yolo11.mdx b/docs/common/ai/cubie/_model-zoo-yolo11.mdx new file mode 100644 index 000000000..0a5f7401d --- /dev/null +++ b/docs/common/ai/cubie/_model-zoo-yolo11.mdx @@ -0,0 +1,365 @@ +本文档讲述如何在 NPU 上运行 YOLO11。 + +:::info +参考 [Model Zoo 下载](./model-zoo-download)获取示例。 +::: + +YOLO11 示例目录结构： + +```bash +$ tree ./ +./ +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ ├── python +│ │ ├── onnx_extract.py +│ │ └── yolo11s_640.txt +│ └── yolo11s_6.txt +├── figures +│ ├── diff_img.png +│ └── out_yolo11.png +├── main.cpp +├── model +│ └── dog.jpg +├── model_config.h +├── README.md +├── yolo11_6_post.cpp +└── yolo11_6_pre.cpp +``` + +## 模型转换 + +### 配置虚拟环境 + + + +```bash +python -m venv .venv && source .venv/bin/activate +pip install ultralytics +``` + + + +### 导出 onnx 模型 + + + +```bash +cd convert_model/python +yolo export model=yolo11s.pt format=onnx imgsz=640 simplify=True dynamic=False opset=11 nms=False batch=1 device=cpu +``` + + + +### 裁剪模型 + + + +```bash +python onnx_extract.py +mv yolo11s_6.onnx ../ +cd ../ +``` + + + +### 创建转换脚本的软链接 + + + +```bash +./convert_model_env.sh +``` + + + +### 模型导入/量化/转换 + +需要先进入容器开发环境。可以参考 Model Zoo 下载中[创建容器](./model-zoo-download#创建并启动容器)这一部分。 + +:::info +不同平台请使用对应的 Docker 镜像： + +- A733：ubuntu-npu:v2.0.10.1 +- T527：ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +进入容器对应目录之后运行脚本。 + + + +```bash +cd /workspace/examples/yolo11/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh yolo11s_6 +./pegasus_quantize.sh yolo11s_6 uint8 12 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolo11s_6 uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolo11s_6 uint8 t527 +``` + + + + + + +导出的模型文件存放在../model目录。 + +### 编译示例 + +接下来可以编译示例，**先 exit 退出容器**，然后执行下面的命令编译示例。 + +首先需要配置第三方库和交叉编译工具链。 + +:::info +如果你已经在其他示例中配置过第三方库和交叉编译工具链则可以跳过这一步。 +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +需要先手动[点击链接](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4)下载之后放到 0-toolchains/ 再执行下面的命令： + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/yolo11/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## 模型部署 + +编译示例完成之后，示例会安装到 install 目录，可以使用 scp 传输到板端。 + +### 配置 NPU 驱动 + +:::info +如果你已经在其他示例中配置过 NPU 驱动则可以跳过这一步。 +::: + +将驱动库 scp 传输到板端的 lib 目录。 + +- A733 对应 common/lib_linux_aarch64/A733 目录 +- T527 对应 common/lib_linux_aarch64/T527 目录 + +然后执行下面的命令导出到环境变量。 + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### 运行示例 + +配置好驱动之后就可以运行示例了。 + +:::tip +对于 T527 平台，你还需要参考 A5E 的`板端启用 NPU`文档先启用 NPU ，然后使用下面的命令增加当前用户使用 /dev/vipcore 的权限。 +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd yolo11_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./yolo11_demo_a733 +./yolo11_demo_a733 -nb model/yolo11s_6_uint8_a733.nb -i model/dog.jpg +``` + + + +运行结果如下： + +```bash +$ ./yolo11_demo_a733 -nb model/yolo11s_6_uint8_a733.nb -i model/dog.jpg +model_file=model/yolo11s_6_uint8_a733.nb, input=model/dog.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 80 80 64 1, data_format=0, name=uid_11_out_0b_uid_1_out_0, none-quant +output 1 dim 80 80 80 1, data_format=0, name=uid_10_out_0b_uid_1_out_0, none-quant +output 2 dim 40 40 64 1, data_format=0, name=uid_9_out_0ub_uid_1_out_0, none-quant +output 3 dim 40 40 80 1, data_format=0, name=uid_8_out_0ub_uid_1_out_0, none-quant +output 4 dim 20 20 64 1, data_format=0, name=uid_7_out_0ub_uid_1_out_0, none-quant +output 5 dim 20 20 80 1, data_format=0, name=uid_6_out_0ub_uid_1_out_0, none-quant +nbg name=model/yolo11s_6_uint8_a733.nb, size: 6850432. +create network 0: 18253 us. +prepare network: 4469 us. +buffer ptr: 0x2baf600, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 31930 us. +output 0, ptr 0x2cdb740, size 409600. +output 1, ptr 0x2e6b7c0, size 512000. +output 2, ptr 0x305f840, size 102400. +output 3, ptr 0x30c38c0, size 128000. +output 4, ptr 0x3140980, size 25600. +output 5, ptr 0x3159a00, size 32000. +detection num: 3 + 1: 95%, [ 126, 130, 568, 419], bicycle +16: 93%, [ 132, 220, 311, 541], dog + 7: 50%, [ 465, 74, 692, 170], truck +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :------ | :--------- | :----------- | :----------- | :----------- | :--------- | :------ | :------- | +| 全志 A733 | Vivante VIP9000 | yolo11s | 640×640 | 18.3 ms | 4.5 ms | 31.9 ms | 5 ms | 59.7 ms | 16.8 FPS | + + + + + + + +```bash +cd yolo11_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./yolo11_demo_t527 +./yolo11_demo_t527 -nb model/yolo11s_6_uint8_t527.nb -i model/dog.jpg +``` + + + +运行结果如下： + +```bash +$ ./yolo11_demo_t527 -nb model/yolo11s_6_uint8_t527.nb -i model/dog.jpg +model_file=model/yolo11s_6_uint8_t527.nb, input=model/dog.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 80 80 64 1, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 80 80 80 1, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 40 40 64 1, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +output 3 dim 40 40 80 1, data_format=0, name=uid_20003_sub_uid_1_out_0, none-quant +output 4 dim 20 20 64 1, data_format=0, name=uid_20004_sub_uid_1_out_0, none-quant +output 5 dim 20 20 80 1, data_format=0, name=uid_20005_sub_uid_1_out_0, none-quant +nbg name=model/yolo11s_6_uint8_t527.nb, size: 7783808. +create network 0: 21201 us. +prepare network: 10246 us. +buffer ptr: 0xbf9e600, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 75191 us. +output 0, ptr 0xc0ca700, size 409600. +output 1, ptr 0xc25a780, size 512000. +output 2, ptr 0xc44e840, size 102400. +output 3, ptr 0xc4b28c0, size 128000. +output 4, ptr 0xc52f940, size 25600. +output 5, ptr 0xc5489c0, size 32000. +detection num: 3 + 1: 94%, [ 127, 129, 567, 419], bicycle +16: 93%, [ 132, 220, 312, 541], dog + 2: 46%, [ 465, 75, 693, 171], car +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :------ | :--------- | :----------- | :----------- | :----------- | :--------- | :------- | :------ | +| 全志 T527 | Vivante VIP9000 | yolo11s | 640×640 | 21.2 ms | 10.2 ms | 75.2 ms | | 106.6 ms | 9.4 FPS | + + + diff --git a/docs/common/ai/cubie/_model-zoo-yolo26.mdx b/docs/common/ai/cubie/_model-zoo-yolo26.mdx new file mode 100644 index 000000000..5059c2842 --- /dev/null +++ b/docs/common/ai/cubie/_model-zoo-yolo26.mdx @@ -0,0 +1,370 @@ +本文档讲述如何在 NPU 上运行 YOLO26 。 + +:::info +参考 [Model Zoo 下载](./model-zoo-download)获取示例。 +::: + +YOLO26示例目录结构： + +```bash +$ tree ./ +./ +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ ├── python +│ │ ├── onnx_extract.py +│ │ └── yolo26s_640.txt +│ └── yolo26s_6.txt +├── figures +│ ├── banner-yolo26.png +│ ├── bus.jpg +│ ├── out_yolo26_6_pcq.png +│ └── performance-comparison.png +├── main.cpp +├── model +│ └── dog.jpg +├── model_config.h +├── README.md +├── yolov26_6_post.cpp +└── yolov26_6_pre.cpp +``` + +## 模型转换 + +### 配置虚拟环境 + + + +```bash +python -m venv .venv && source .venv/bin/activate +pip install ultralytics +``` + + + +### 导出 onnx 模型 + +ultralytics 会自动下载模型和缺失的依赖，耐心等待。 + + + +```bash +cd convert_model/python/ +yolo export model=yolo26s.pt format=onnx simplify=True dynamic=False opset=16 +``` + + + +### 裁剪模型 + + + +```bash +python onnx_extract.py +cd .. +``` + + + +### 创建转换脚本的软链接 + + + +```bash +./convert_model_env.sh +``` + + + +### 模型导入/量化/转换 + +需要先进入容器开发环境。可以参考 Model Zoo 下载中[创建容器](./model-zoo-download#创建并启动容器)这一部分。 + +:::info +不同平台请使用对应的 Docker 镜像： + +- A733：ubuntu-npu:v2.0.10.1 +- T527：ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +进入容器对应目录之后运行脚本。 + + + +```bash +# 进入容器对应目录 +cd /workspace/examples/yolo26/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh yolo26s_6 # 去掉后缀的模型名 +./pegasus_quantize.sh yolo26s_6 pcq 12 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolo26s_6 pcq a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolo26s_6 pcq t527 +``` + + + + + + +导出的模型文件存放在../model目录。 + +### 编译示例 + +接下来可以编译示例，**先 exit 退出容器**，然后执行下面的命令编译示例。 + +首先需要配置第三方库和交叉编译工具链。 + +:::info +如果你已经在其他示例中配置过第三方库和交叉编译工具链则可以跳过这一步。 +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +需要先手动[点击链接](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4)下载之后放到 0-toolchains/ 再执行下面的命令： + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/yolo26/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## 模型部署 + +编译示例完成之后，示例会安装到 install 目录，可以使用 scp 传输到板端。 + +### 配置 NPU 驱动 + +:::info +如果你已经在其他示例中配置过 NPU 驱动则可以跳过这一步。 +::: + +将驱动库 scp 传输到板端的 lib 目录。 + +- A733 对应 common/lib_linux_aarch64/A733 目录 +- T527 对应 common/lib_linux_aarch64/T527 目录 + +然后执行下面的命令导出到环境变量。 + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### 运行示例 + +配置好驱动之后就可以运行示例了。 + +:::tip +对于 T527 平台，你还需要参考 A5E 的`板端启用 NPU`文档先启用 NPU ，然后使用下面的命令增加当前用户使用 /dev/vipcore 的权限。 +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd yolo26_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./yolo26_demo_a733 +./yolo26_demo_a733 -nb model/yolo26s_6_pcq_a733.nb -i model/dog.jpg +``` + + + +运行结果如下： + +```bash +radxa@radxa-cubie-a7a:~/Project/yolo26_demo_linux_a733$ ./yolo26_demo_a733 -nb model/yolo26s_6_pcq_a733.nb -i model/dog.jpg +model_file=model/yolo26s_6_pcq_a733.nb, input=model/dog.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 6400 4 1 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 1600 4 1 0, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 400 4 1 0, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +output 3 dim 6400 80 1 0, data_format=0, name=uid_20003_sub_uid_1_out_0, none-quant +output 4 dim 1600 80 1 0, data_format=0, name=uid_20004_sub_uid_1_out_0, none-quant +output 5 dim 400 80 1 0, data_format=0, name=uid_20005_sub_uid_1_out_0, none-quant +nbg name=model/yolo26s_6_pcq_a733.nb, size: 9362920. +create network 0: 18319 us. +prepare network: 8871 us. +buffer ptr: 0x5cb2600, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 34765 us. +output 0, ptr 0x5dde740, size 25600. +output 1, ptr 0x5df77c0, size 6400. +output 2, ptr 0x5dfdc40, size 1600. +output 3, ptr 0x5dff5c0, size 512000. +output 4, ptr 0x5ff3680, size 128000. +output 5, ptr 0x6070700, size 32000. +postprocess time : 6 ms +detection num: 3 + 7: 68%, [ 466, 74, 690, 171], truck + 1: 89%, [ 130, 136, 566, 420], bicycle +16: 90%, [ 133, 221, 310, 540], dog +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :------ | :--------- | :----------- | :----------- | :----------- | :--------- | :------ | :------- | +| 全志 A733 | Vivante VIP9000 | yolo26s | 640×640 | 18.3 ms | 8.9 ms | 34.8 ms | 6 ms | 68.0 ms | 14.7 FPS | + + + + + + + +```bash +cd yolo26_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./yolo26_demo_t527 +./yolo26_demo_t527 -nb model/yolo26s_6_pcq_t527.nb -i model/dog.jpg +``` + + + +运行结果如下： + +```bash +$ ./yolo26_demo_t527 -nb model/yolo26s_6_pcq_t527.nb -i model/dog.jpg +model_file=model/yolo26s_6_pcq_t527.nb, input=model/dog.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 6400 4 1 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 1600 4 1 0, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 400 4 1 0, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +output 3 dim 6400 80 1 0, data_format=0, name=uid_20003_sub_uid_1_out_0, none-quant +output 4 dim 1600 80 1 0, data_format=0, name=uid_20004_sub_uid_1_out_0, none-quant +output 5 dim 400 80 1 0, data_format=0, name=uid_20005_sub_uid_1_out_0, none-quant +nbg name=model/yolo26s_6_pcq_t527.nb, size: 9920576. +create network 0: 20397 us. +prepare network: 11311 us. +buffer ptr: 0x6fe6600, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 87270 us. +output 0, ptr 0x7112700, size 25600. +output 1, ptr 0x712b780, size 6400. +output 2, ptr 0x7131c40, size 1600. +output 3, ptr 0x71335c0, size 512000. +output 4, ptr 0x7327640, size 128000. +output 5, ptr 0x73a46c0, size 32000. +postprocess time : 20 ms +detection num: 2 + 1: 94%, [ 128, 135, 566, 418], bicycle +16: 87%, [ 133, 221, 309, 542], dog +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :------ | :--------- | :----------- | :----------- | :----------- | :--------- | :------- | :------ | +| 全志 T527 | Vivante VIP9000 | yolo26s | 640×640 | 20.4 ms | 11.3 ms | 87.3 ms | 20 ms | 139.0 ms | 7.2 FPS | + + + diff --git a/docs/common/ai/cubie/_model-zoo-yolov3-darknet.mdx b/docs/common/ai/cubie/_model-zoo-yolov3-darknet.mdx new file mode 100644 index 000000000..c69fc681a --- /dev/null +++ b/docs/common/ai/cubie/_model-zoo-yolov3-darknet.mdx @@ -0,0 +1,322 @@ +本文档讲述如何在 NPU 上运行 YOLOv3。 + +:::info +参考 [Model Zoo 下载](./model-zoo-download)获取示例。 +::: + +YOLOv3 示例目录结构： + +```bash +$ tree ./ +./ +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ └── yolov3.cfg +├── main.cpp +├── model +│ └── horses_416x416.jpg +├── model_config.h +├── README.md +├── yolov3_post.cpp +└── yolov3_pre.cpp +``` + +## 模型转换 + +### 下载模型文件 + + + +```bash +cd convert_model/ +wget https://pjreddie.com/media/files/yolov3.weights +``` + + + +### 创建转换脚本的软链接 + + + +```bash +./convert_model_env.sh +``` + + + +### 模型导入/量化/转换 + +需要先进入容器开发环境。可以参考 Model Zoo 下载中[创建容器](./model-zoo-download#创建并启动容器)这一部分。 + +:::info +不同平台请使用对应的 Docker 镜像： + +- A733：ubuntu-npu:v2.0.10.1 +- T527：ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +进入容器对应目录之后运行脚本。 + + + +```bash +cd /workspace/examples/yolov3_darknet/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh yolov3 +./pegasus_quantize.sh yolov3 uint8 12 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolov3 uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolov3 uint8 t527 +``` + + + + + + +导出的模型文件存放在../model目录。 + +### 编译示例 + +接下来可以编译示例，**先 exit 退出容器**，然后执行下面的命令编译示例。 + +首先需要配置第三方库和交叉编译工具链。 + +:::info +如果你已经在其他示例中配置过第三方库和交叉编译工具链则可以跳过这一步。 +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +需要先手动[点击链接](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4)下载之后放到 0-toolchains/ 再执行下面的命令： + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/yolov3/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## 模型部署 + +编译示例完成之后，示例会安装到 install 目录，可以使用 scp 传输到板端。 + +### 配置 NPU 驱动 + +:::info +如果你已经在其他示例中配置过 NPU 驱动则可以跳过这一步。 +::: + +将驱动库 scp 传输到板端的 lib 目录。 + +- A733 对应 common/lib_linux_aarch64/A733 目录 +- T527 对应 common/lib_linux_aarch64/T527 目录 + +然后执行下面的命令导出到环境变量。 + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### 运行示例 + +配置好驱动之后就可以运行示例了。 + +:::tip +对于 T527 平台，你还需要参考 A5E 的`板端启用 NPU`文档先启用 NPU ，然后使用下面的命令增加当前用户使用 /dev/vipcore 的权限。 +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd yolov3_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./yolov3_demo_a733 +./yolov3_demo_a733 -nb model/yolov3_uint8_a733.nb -i model/horses_416x416.jpg +``` + + + +运行结果如下： + +```bash +$ ./yolov3_demo_a733 -nb model/yolov3_uint8_a733.nb -i model/horses_416x416.jpg +model_file=model/yolov3_uint8_a733.nb, input=model/horses_416x416.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 416 416 3 1, data_format=2, quant_format=2, name=input/output[0], scale=0.003906, zero_point=0 +output 0 dim 13 13 255 1, data_format=2, name=uid_198_out_0, scale=0.191153, zero_point=189 +output 1 dim 26 26 255 1, data_format=2, name=uid_224_out_0, scale=0.213471, zero_point=198 +output 2 dim 52 52 255 1, data_format=2, name=uid_250_out_0, scale=0.323550, zero_point=184 +nbg name=model/yolov3_uint8_a733.nb, size: 38927048. +create network 0: 22932 us. +prepare network: 3310 us. +buffer ptr: 0xffffab170040, buffer size: 519168 +feed input cost: 19560 us. +network: 0, loop count: 1 +run time for this network 0: 35442 us. +detection num: 3 +17: 98%, [ 234, 169, 324, 288], horse +17: 88%, [ 137, 149, 233, 281], horse +17: 88%, [ 8, 165, 197, 329], horse +draw objects time : 13 ms +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :----- | :--------- | :----------- | :----------- | :----------- | :--------- | :------ | :------- | +| 全志 A733 | Vivante VIP9000 | yolov3 | 416×416 | 22.9 ms | 3.3 ms | 35.4 ms | 13 ms | 74.6 ms | 13.4 FPS | + + + + + + + +```bash +cd yolov3_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./yolov3_demo_t527 +./yolov3_demo_t527 -nb model/yolov3_uint8_t527.nb -i model/horses_416x416.jpg +``` + + + +运行结果如下： + +```bash +$ ./yolov3_demo_t527 -nb model/yolov3_uint8_t527.nb -i model/horses_416x416.jpg +model_file=model/yolov3_uint8_t527.nb, input=model/horses_416x416.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 416 416 3 1, data_format=2, quant_format=2, name=input[0], scale=0.003906, zero_point=0 +output 0 dim 13 13 255 1, data_format=2, name=uid_198_out_0, scale=0.191153, zero_point=189 +output 1 dim 26 26 255 1, data_format=2, name=uid_224_out_0, scale=0.213471, zero_point=198 +output 2 dim 52 52 255 1, data_format=2, name=uid_250_out_0, scale=0.323550, zero_point=184 +nbg name=model/yolov3_uint8_t527.nb, size: 39252928. +create network 0: 54280 us. +prepare network: 10412 us. +buffer ptr: 0xffff8b764040, buffer size: 519168 +feed input cost: 97856 us. +network: 0, loop count: 1 +run time for this network 0: 63627 us. +detection num: 3 +17: 98%, [ 234, 169, 324, 288], horse +17: 91%, [ 8, 165, 197, 329], horse +17: 88%, [ 127, 148, 243, 280], horse +draw objects time : 44 ms +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :----- | :--------- | :----------- | :----------- | :----------- | :--------- | :------- | :------ | +| 全志 T527 | Vivante VIP9000 | yolov3 | 416×416 | 54.3 ms | 10.4 ms | 63.6 ms | 44.0 ms | 172.3 ms | 5.8 FPS | + + + diff --git a/docs/common/ai/cubie/_model-zoo-yolov5.mdx b/docs/common/ai/cubie/_model-zoo-yolov5.mdx new file mode 100644 index 000000000..73e527eec --- /dev/null +++ b/docs/common/ai/cubie/_model-zoo-yolov5.mdx @@ -0,0 +1,325 @@ +本文档讲述如何在 NPU 上运行 YOLOv5。 + +:::info +参考 [Model Zoo 下载](./model-zoo-download)获取示例。 +::: + +YOLOv5 示例目录结构： + +```bash +$ tree ./ +./ +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ ├── python +│ │ ├── onnx_extract.py +│ │ └── yolov5s-sim.onnx +│ └── yolov5s_rt.onnx +├── figures +│ ├── dog.jpg +│ ├── onnx.jpg +│ ├── onnx-sim.jpg +│ ├── output.jpg +│ └── output_yolov5_uint8.png +├── main.cpp +├── model +│ ├── dog.jpg +│ ├── yolov5s_rt_uint8_a733.nb +│ └── yolov5s_rt_uint8_t527.nb +├── model_config.h +├── README.md +├── yolov5_post.cpp +└── yolov5_pre.cpp +``` + +## 模型转换 + +无需导出到 onnx 模型，直接使用仓库中提供的 yolov5s_rt.onnx 。 + +### 创建转换脚本的软链接 + + + +```bash +cd convert_model/ +./convert_model_env.sh +``` + + + +### 模型导入/量化/转换 + +需要先进入容器开发环境。可以参考 Model Zoo 下载中[创建容器](./model-zoo-download#创建并启动容器)这一部分。 + +:::info +不同平台请使用对应的 Docker 镜像： + +- A733：ubuntu-npu:v2.0.10.1 +- T527：ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +进入容器对应目录之后运行脚本。 + + + +```bash +cd /workspace/examples/yolov5/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh yolov5s_rt +./pegasus_quantize.sh yolov5s_rt uint8 12 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolov5s_rt uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolov5s_rt uint8 t527 +``` + + + + + + +导出的模型文件存放在../model目录。 + +### 编译示例 + +接下来可以编译示例，**先 exit 退出容器**，然后执行下面的命令编译示例。 + +首先需要配置第三方库和交叉编译工具链。 + +:::info +如果你已经在其他示例中配置过第三方库和交叉编译工具链则可以跳过这一步。 +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +需要先手动[点击链接](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4)下载之后放到 0-toolchains/ 再执行下面的命令： + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/yolov5/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## 模型部署 + +编译示例完成之后，示例会安装到 install 目录，可以使用 scp 传输到板端。 + +### 配置 NPU 驱动 + +:::info +如果你已经在其他示例中配置过 NPU 驱动则可以跳过这一步。 +::: + +将驱动库 scp 传输到板端的 lib 目录。 + +- A733 对应 common/lib_linux_aarch64/A733 目录 +- T527 对应 common/lib_linux_aarch64/T527 目录 + +然后执行下面的命令导出到环境变量。 + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### 运行示例 + +配置好驱动之后就可以运行示例了。 + +:::tip +对于 T527 平台，你还需要参考 A5E 的`板端启用 NPU`文档先启用 NPU ，然后使用下面的命令增加当前用户使用 /dev/vipcore 的权限。 +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd yolov5_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./yolov5_demo_a733 +./yolov5_demo_a733 -nb model/yolov5s_rt_uint8_a733.nb -i model/dog.jpg +``` + + + +运行结果如下： + +```bash +$ ./yolov5_demo_a733 -nb model/yolov5s_rt_uint8_a733.nb -i model/dog.jpg +model_file=model/yolov5s_rt_uint8_a733.nb, input=model/dog.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 80 80 85 3, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 40 40 85 3, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 20 20 85 3, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +nbg name=model/yolov5s_rt_uint8_a733.nb, size: 5564152. +create network 0: 11618 us. +prepare network: 2756 us. +buffer ptr: 0x3f0327c0, buffer size: 1228800 +feed input cost: 19421 us. +network: 0, loop count: 1 +run time for this network 0: 20142 us. +detection num: 3 +16: 92%, [ 134, 226, 307, 545], dog + 7: 69%, [ 471, 77, 689, 173], truck + 1: 52%, [ 162, 125, 559, 423], bicycle +draw objects time : 32 ms +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :------ | :--------- | :----------- | :----------- | :----------- | :--------- | :------ | :------- | +| 全志 A733 | Vivante VIP9000 | yolov5s | 640×640 | 11.6 ms | 2.8 ms | 20.1 ms | 32.0 ms | 66.5 ms | 15.0 FPS | + + + + + + + +```bash +cd yolov5_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./yolov5_demo_t527 +./yolov5_demo_t527 -nb model/yolov5s_rt_uint8_t527.nb -i model/dog.jpg +``` + + + +运行结果如下： + +```bash +$ ./yolov5_demo_t527 -nb model/yolov5s_rt_uint8_t527.nb -i model/dog.jpg +model_file=model/yolov5s_rt_uint8_t527.nb, input=model/dog.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 80 80 85 3, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 40 40 85 3, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 20 20 85 3, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +nbg name=model/yolov5s_rt_uint8_t527.nb, size: 6300672. +create network 0: 24270 us. +prepare network: 10842 us. +buffer ptr: 0x18296780, buffer size: 1228800 +feed input cost: 90901 us. +network: 0, loop count: 1 +run time for this network 0: 49985 us. +detection num: 3 +16: 92%, [ 137, 229, 303, 539], dog + 7: 69%, [ 471, 79, 687, 171], truck + 1: 44%, [ 151, 120, 560, 429], bicycle +draw objects time : 81 ms +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :------ | :--------- | :----------- | :----------- | :----------- | :--------- | :------- | :------ | +| 全志 T527 | Vivante VIP9000 | yolov5s | 640×640 | 24.3 ms | 10.8 ms | 50.0 ms | 81.0 ms | 166.1 ms | 6.0 FPS | + + + diff --git a/docs/common/ai/cubie/_model-zoo-yolov8-pose.mdx b/docs/common/ai/cubie/_model-zoo-yolov8-pose.mdx new file mode 100644 index 000000000..cf5197a31 --- /dev/null +++ b/docs/common/ai/cubie/_model-zoo-yolov8-pose.mdx @@ -0,0 +1,490 @@ +本文档讲述如何在 NPU 上运行 YOLOv8 Pose。 + +:::info +参考 [Model Zoo 下载](./model-zoo-download)获取示例。 +::: + +YOLOv8 Pose 示例目录结构： + +```bash +$ tree ./ +./ +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ ├── python +│ │ ├── onnx_extract.py +│ │ └── yolov8s-pose_640.txt +│ └── yolov8s-pose_9.txt +├── figures +│ ├── diff_img.png +│ └── out_yolov8_pose_pcq.png +├── main.cpp +├── model +│ └── COCO_train2014_000000500390.jpg +├── model_config.h +├── README.md +├── yolov8_pose_9_post.cpp +└── yolov8_pose_9_pre.cpp +``` + +## 模型转换 + +### 配置虚拟环境 + + + +```bash +python -m venv .venv && source .venv/bin/activate +pip install ultralytics==8.1.0 onnxsim +``` + + + +### 导出 onnx 模型 + + + +```bash +cd convert_model/python/ +yolo export model=yolov8s-pose.pt format=onnx dynamic=True opset=11 +``` + + + +### 固定 shape + + + +```bash +python3 -m onnxsim yolov8s-pose.onnx yolov8s-pose_640.onnx --input-shape=1,3,640,640 +``` + + + +### 裁剪模型 + + + +```bash +python3 onnx_extract.py +cd .. +``` + + + +### 创建转换脚本的软链接 + + + +```bash +./convert_model_env.sh +``` + + + +### 模型导入/量化/转换 + +需要先进入容器开发环境。可以参考 Model Zoo 下载中[创建容器](./model-zoo-download#创建并启动容器)这一部分。 + +:::info +不同平台请使用对应的 Docker 镜像： + +- A733：ubuntu-npu:v2.0.10.1 +- T527：ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +进入容器对应目录之后运行脚本。 + + + +```bash +cd /workspace/examples/yolov8_pose/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh yolov8s-pose_9 +./pegasus_quantize.sh yolov8s-pose_9 uint8 12 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolov8s-pose_9 uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolov8s-pose_9 uint8 t527 +``` + + + + + + +导出的模型文件存放在../model目录。 + +### 编译示例 + +接下来可以编译示例，**先 exit 退出容器**，然后执行下面的命令编译示例。 + +首先需要配置第三方库和交叉编译工具链。 + +:::info +如果你已经在其他示例中配置过第三方库和交叉编译工具链则可以跳过这一步。 +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +需要先手动[点击链接](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4)下载之后放到 0-toolchains/ 再执行下面的命令： + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/yolov8_pose/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## 模型部署 + +编译示例完成之后，示例会安装到 install 目录，可以使用 scp 传输到板端。 + +### 配置 NPU 驱动 + +:::info +如果你已经在其他示例中配置过 NPU 驱动则可以跳过这一步。 +::: + +将驱动库 scp 传输到板端的 lib 目录。 + +- A733 对应 common/lib_linux_aarch64/A733 目录 +- T527 对应 common/lib_linux_aarch64/T527 目录 + +然后执行下面的命令导出到环境变量。 + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### 运行示例 + +配置好驱动之后就可以运行示例了。 + +:::tip +对于 T527 平台，你还需要参考 A5E 的`板端启用 NPU`文档先启用 NPU ，然后使用下面的命令增加当前用户使用 /dev/vipcore 的权限。 +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd yolov8_pose_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./yolov8_pose_demo_a733 +./yolov8_pose_demo_a733 -nb model/yolov8s-pose_9_uint8_a733.nb -i model/COCO_train2014_000000500390.jpg +``` + + + +运行结果如下： + +```bash +$ ./yolov8_pose_demo_a733 -nb model/yolov8s-pose_9_uint8_a733.nb -i model/COCO_train2014_000000500390.jpg +model_file=model/yolov8s-pose_9_uint8_a733.nb, input=model/COCO_train2014_000000500390.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 80 80 64 1, data_format=0, name=uid_17_out_0b_uid_1_out_0, none-quant +output 1 dim 80 80 1 1, data_format=0, name=uid_16_out_0b_uid_1_out_0, none-quant +output 2 dim 80 80 51 1, data_format=0, name=uid_15_out_0b_uid_1_out_0, none-quant +output 3 dim 40 40 64 1, data_format=0, name=uid_14_out_0b_uid_1_out_0, none-quant +output 4 dim 40 40 1 1, data_format=0, name=uid_13_out_0b_uid_1_out_0, none-quant +output 5 dim 40 40 51 1, data_format=0, name=uid_12_out_0b_uid_1_out_0, none-quant +output 6 dim 20 20 64 1, data_format=0, name=uid_11_out_0b_uid_1_out_0, none-quant +output 7 dim 20 20 1 1, data_format=0, name=uid_10_out_0b_uid_1_out_0, none-quant +output 8 dim 20 20 51 1, data_format=0, name=uid_9_out_0ub_uid_1_out_0, none-quant +nbg name=model/yolov8s-pose_9_uint8_a733.nb, size: 7768344. +create network 0: 18985 us. +prepare network: 5711 us. +buffer ptr: 0x28344380, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 32958 us. +output 0, ptr 0x28470480, size 409600. +output 1, ptr 0x28600500, size 6400. +output 2, ptr 0x28606980, size 326400. +output 3, ptr 0x28745640, size 102400. +output 4, ptr 0x287a96c0, size 1600. +output 5, ptr 0x287ab040, size 81600. +output 6, ptr 0x287fabc0, size 25600. +output 7, ptr 0x28813c80, size 400. +output 8, ptr 0x28814340, size 20400. +post process time : 4 ms +detection num: 3 + 0: 93%, [ 373, 1, 587, 346], person +411.58 41.32 = 0.98922 +419.64 35.78 = 0.98396 +416.36 37.19 = 0.76423 +440.57 37.33 = 0.97060 +422.52 38.08 = 0.12822 +450.85 69.58 = 0.99924 +422.50 75.59 = 0.99804 +473.26 121.09 = 0.99354 +405.71 108.13 = 0.95213 +449.46 97.35 = 0.98640 +389.93 81.26 = 0.92587 +461.08 161.27 = 0.99969 +461.53 162.40 = 0.99945 +405.04 226.47 = 0.99954 +489.77 240.76 = 0.99892 +415.75 320.01 = 0.99481 +555.85 276.33 = 0.99307 + 0: 93%, [ 86, 28, 288, 390], person +155.68 76.87 = 0.99271 +162.21 68.34 = 0.98739 +145.10 65.45 = 0.95864 +175.03 64.92 = 0.91619 +141.19 64.45 = 0.68796 +199.98 93.83 = 0.99730 +160.28 98.94 = 0.99395 +214.27 138.31 = 0.99138 +164.10 156.53 = 0.98026 +175.57 174.47 = 0.98414 +136.65 193.82 = 0.97464 +216.19 199.03 = 0.99952 +180.07 198.95 = 0.99935 +240.79 270.84 = 0.99790 +150.18 279.74 = 0.99727 +293.96 281.26 = 0.98766 +128.72 359.77 = 0.98534 + 0: 91%, [ 227, 36, 398, 405], person +281.36 106.56 = 0.99230 +287.73 97.59 = 0.98680 +279.19 104.41 = 0.88281 +308.31 83.30 = 0.95046 +275.74 96.24 = 0.19450 +328.64 102.08 = 0.99900 +275.67 126.89 = 0.99741 +373.27 126.41 = 0.99145 +278.76 161.70 = 0.95360 +382.76 163.57 = 0.98179 +249.65 205.81 = 0.91721 +332.66 214.56 = 0.99969 +309.45 218.78 = 0.99948 +293.53 304.20 = 0.99923 +310.05 306.35 = 0.99817 +279.65 380.53 = 0.99397 +363.50 304.91 = 0.99248 +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :----------- | :--------- | :----------- | :----------- | :----------- | :--------- | :------ | :------- | +| 全志 A733 | Vivante VIP9000 | yolov8s-pose | 640×640 | 19.0 ms | 5.7 ms | 33.0 ms | 4.0 ms | 61.7 ms | 16.2 FPS | + + + + + + + +```bash +cd yolov8_pose_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./yolov8_pose_demo_t527 +./yolov8_pose_demo_t527 -nb model/yolov8s-pose_9_uint8_t527.nb -i model/COCO_train2014_000000500390.jpg +``` + + + +运行结果如下： + +```bash +$ ./yolov8_pose_demo_t527 -nb model/yolov8s-pose_9_uint8_t527.nb -i model/COCO_train2014_000000500390.jpg +model_file=model/yolov8s-pose_9_uint8_t527.nb, input=model/COCO_train2014_000000500390.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 80 80 64 1, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 80 80 1 1, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 80 80 51 1, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +output 3 dim 40 40 64 1, data_format=0, name=uid_20003_sub_uid_1_out_0, none-quant +output 4 dim 40 40 1 1, data_format=0, name=uid_20004_sub_uid_1_out_0, none-quant +output 5 dim 40 40 51 1, data_format=0, name=uid_20005_sub_uid_1_out_0, none-quant +output 6 dim 20 20 64 1, data_format=0, name=uid_20006_sub_uid_1_out_0, none-quant +output 7 dim 20 20 1 1, data_format=0, name=uid_20007_sub_uid_1_out_0, none-quant +output 8 dim 20 20 51 1, data_format=0, name=uid_20008_sub_uid_1_out_0, none-quant +nbg name=model/yolov8s-pose_9_uint8_t527.nb, size: 10369024. +create network 0: 25460 us. +prepare network: 20276 us. +buffer ptr: 0x2a796380, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 71053 us. +output 0, ptr 0x2a8c2440, size 409600. +output 1, ptr 0x2aa52500, size 6400. +output 2, ptr 0x2aa58980, size 326400. +output 3, ptr 0x2ab97600, size 102400. +output 4, ptr 0x2abfb680, size 1600. +output 5, ptr 0x2abfd040, size 81600. +output 6, ptr 0x2ac4cbc0, size 25600. +output 7, ptr 0x2ac65c40, size 400. +output 8, ptr 0x2ac66300, size 20400. +post process time : 11 ms +detection num: 3 + 0: 93%, [ 373, 1, 587, 347], person +411.64 37.38 = 0.98651 +419.72 33.34 = 0.98042 +415.68 33.34 = 0.72047 +435.88 33.34 = 0.97493 +423.76 37.38 = 0.13819 +452.04 65.66 = 0.99938 +423.76 73.73 = 0.99807 +472.24 118.17 = 0.99438 +407.60 102.01 = 0.94482 +452.04 93.93 = 0.98564 +391.45 81.81 = 0.90655 +460.12 158.56 = 0.99969 +460.12 158.56 = 0.99945 +403.56 227.24 = 0.99963 +492.44 243.40 = 0.99904 +411.64 320.15 = 0.99614 +557.07 283.79 = 0.99438 + 0: 93%, [ 86, 28, 288, 389], person +155.96 77.46 = 0.99278 +164.04 69.38 = 0.98809 +143.84 65.34 = 0.95914 +176.16 65.34 = 0.92141 +139.80 65.34 = 0.68079 +200.40 93.62 = 0.99736 +160.00 97.66 = 0.99363 +212.51 138.05 = 0.99128 +164.04 158.25 = 0.97784 +176.16 174.41 = 0.98374 +135.76 194.60 = 0.97166 +216.55 198.64 = 0.99949 +180.20 198.64 = 0.99930 +240.79 271.36 = 0.99781 +151.92 279.44 = 0.99700 +293.30 283.47 = 0.98732 +127.68 356.19 = 0.98472 + 0: 92%, [ 227, 36, 398, 406], person +279.60 105.73 = 0.99278 +287.68 97.66 = 0.98809 +279.60 105.73 = 0.88286 +307.88 81.50 = 0.95390 +275.56 97.66 = 0.18974 +328.08 101.70 = 0.99904 +275.56 125.93 = 0.99736 +372.51 125.93 = 0.99181 +279.60 162.29 = 0.94802 +384.63 162.29 = 0.98270 +251.33 206.72 = 0.91177 +332.12 214.80 = 0.99969 +307.88 218.84 = 0.99945 +291.72 303.67 = 0.99925 +307.88 307.71 = 0.99807 +279.60 380.42 = 0.99438 +364.44 303.67 = 0.99231 +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :----------- | :--------- | :----------- | :----------- | :----------- | :--------- | :------- | :------ | +| 全志 T527 | Vivante VIP9000 | yolov8s-pose | 640×640 | 25.5 ms | 20.3 ms | 71.1 ms | 11.0 ms | 127.9 ms | 7.8 FPS | + + + diff --git a/docs/common/ai/cubie/_model-zoo-yolov8-seg.mdx b/docs/common/ai/cubie/_model-zoo-yolov8-seg.mdx new file mode 100644 index 000000000..070cd0ba5 --- /dev/null +++ b/docs/common/ai/cubie/_model-zoo-yolov8-seg.mdx @@ -0,0 +1,393 @@ +本文档讲述如何在 NPU 上运行 YOLOv8 Seg。 + +:::info +参考 [Model Zoo 下载](./model-zoo-download)获取示例。 +::: + +YOLOv8 Seg 示例目录结构： + +```bash +$ tree ./ +./ +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ ├── python +│ │ ├── onnx_extract.py +│ │ └── yolov8s-seg_640.txt +│ └── yolov8s-seg_10.txt +├── figures +│ ├── diff_img.png +│ └── out_yolov8_seg.png +├── main.cpp +├── model +│ ├── bus.jpg +│ └── dog.jpg +├── model_config.h +├── README.md +├── yolov8_seg_10_post.cpp +└── yolov8_seg_10_pre.cpp +``` + +## 模型转换 + +### 配置虚拟环境 + + + +```bash +python -m venv .venv && source .venv/bin/activate +pip install ultralytics onnxsim +``` + + + +### 导出 onnx 模型 + + + +```bash +cd convert_model/python/ +yolo export model=yolov8s-seg.pt format=onnx dynamic=True opset=11 +``` + + + +### 固定 shape + + + +```bash +python3 -m onnxsim yolov8s-seg.onnx yolov8s-seg_640.onnx --input-shape=1,3,640,640 +``` + + + +### 裁剪模型 + + + +```bash +python3 onnx_extract.py +cd .. +``` + + + +### 创建转换脚本的软链接 + + + +```bash +./convert_model_env.sh +``` + + + +### 模型导入/量化/转换 + +需要先进入容器开发环境。可以参考 Model Zoo 下载中[创建容器](./model-zoo-download#创建并启动容器)这一部分。 + +:::info +不同平台请使用对应的 Docker 镜像： + +- A733：ubuntu-npu:v2.0.10.1 +- T527：ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +进入容器对应目录之后运行脚本。 + + + +```bash +cd /workspace/examples/yolov8_seg/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh yolov8s-seg_10 +./pegasus_quantize.sh yolov8s-seg_10 uint8 12 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolov8s-seg_10 uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolov8s-seg_10 uint8 t527 +``` + + + + + + +导出的模型文件存放在../model目录。 + +### 编译示例 + +接下来可以编译示例，**先 exit 退出容器**，然后执行下面的命令编译示例。 + +首先需要配置第三方库和交叉编译工具链。 + +:::info +如果你已经在其他示例中配置过第三方库和交叉编译工具链则可以跳过这一步。 +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +需要先手动[点击链接](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4)下载之后放到 0-toolchains/ 再执行下面的命令： + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/yolov8_seg/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## 模型部署 + +编译示例完成之后，示例会安装到 install 目录，可以使用 scp 传输到板端。 + +### 配置 NPU 驱动 + +:::info +如果你已经在其他示例中配置过 NPU 驱动则可以跳过这一步。 +::: + +将驱动库 scp 传输到板端的 lib 目录。 + +- A733 对应 common/lib_linux_aarch64/A733 目录 +- T527 对应 common/lib_linux_aarch64/T527 目录 + +然后执行下面的命令导出到环境变量。 + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### 运行示例 + +配置好驱动之后就可以运行示例了。 + +:::tip +对于 T527 平台，你还需要参考 A5E 的`板端启用 NPU`文档先启用 NPU ，然后使用下面的命令增加当前用户使用 /dev/vipcore 的权限。 +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd yolov8_seg_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./yolov8_seg_demo_a733 +./yolov8_seg_demo_a733 -nb model/yolov8s-seg_10_uint8_a733.nb -i model/dog.jpg +``` + + + +运行结果如下： + +```bash +$ ./yolov8_seg_demo_a733 -nb model/yolov8s-seg_10_uint8_a733.nb -i model/dog.jpg +model_file=model/yolov8s-seg_10_uint8_a733.nb, input=model/dog.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 80 80 64 1, data_format=0, name=uid_19_out_0b_uid_1_out_0, none-quant +output 1 dim 80 80 80 1, data_format=0, name=uid_18_out_0b_uid_1_out_0, none-quant +output 2 dim 80 80 32 1, data_format=0, name=uid_17_out_0b_uid_1_out_0, none-quant +output 3 dim 40 40 64 1, data_format=0, name=uid_16_out_0b_uid_1_out_0, none-quant +output 4 dim 40 40 80 1, data_format=0, name=uid_15_out_0b_uid_1_out_0, none-quant +output 5 dim 40 40 32 1, data_format=0, name=uid_14_out_0b_uid_1_out_0, none-quant +output 6 dim 20 20 64 1, data_format=0, name=uid_13_out_0b_uid_1_out_0, none-quant +output 7 dim 20 20 80 1, data_format=0, name=uid_12_out_0b_uid_1_out_0, none-quant +output 8 dim 20 20 32 1, data_format=0, name=uid_11_out_0b_uid_1_out_0, none-quant +output 9 dim 160 160 32 1, data_format=0, name=uid_20009_sub_uid_1_out_0, none-quant +nbg name=model/yolov8s-seg_10_uint8_a733.nb, size: 8089336. +create network 0: 21684 us. +prepare network: 4881 us. +buffer ptr: 0x127ba600, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 38144 us. +output 0, ptr 0x128e6780, size 409600. +output 1, ptr 0x12a76800, size 512000. +output 2, ptr 0x12c6a880, size 204800. +output 3, ptr 0x12d32900, size 102400. +output 4, ptr 0x12d969c0, size 128000. +output 5, ptr 0x12e13a40, size 51200. +output 6, ptr 0x12e45ac0, size 25600. +output 7, ptr 0x12e5eb40, size 32000. +output 8, ptr 0x12e7e000, size 12800. +output 9, ptr 0x12e8a880, size 819200. +post process time : 12 ms +detection num: 3 + 1: 89%, [ 126, 133, 568, 425], bicycle +16: 96%, [ 131, 220, 310, 541], dog + 7: 65%, [ 470, 73, 688, 171], truck +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :---------- | :--------- | :----------- | :----------- | :----------- | :--------- | :------ | :------- | +| 全志 A733 | Vivante VIP9000 | yolov8s-seg | 640×640 | 21.7 ms | 4.9 ms | 38.1 ms | 12 ms | 76.7 ms | 13.0 FPS | + + + + + + + +```bash +cd yolov8_seg_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./yolov8_seg_demo_t527 +./yolov8_seg_demo_t527 -nb model/yolov8s-seg_10_uint8_t527.nb -i model/dog.jpg +``` + + + +运行结果如下： + +```bash +$ ./yolov8_seg_demo_t527 -nb model/yolov8s-seg_10_uint8_t527.nb -i model/dog.jpg +model_file=model/yolov8s-seg_10_uint8_t527.nb, input=model/dog.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 80 80 64 1, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 80 80 80 1, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 80 80 32 1, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +output 3 dim 40 40 64 1, data_format=0, name=uid_20003_sub_uid_1_out_0, none-quant +output 4 dim 40 40 80 1, data_format=0, name=uid_20004_sub_uid_1_out_0, none-quant +output 5 dim 40 40 32 1, data_format=0, name=uid_20005_sub_uid_1_out_0, none-quant +output 6 dim 20 20 64 1, data_format=0, name=uid_20006_sub_uid_1_out_0, none-quant +output 7 dim 20 20 80 1, data_format=0, name=uid_20007_sub_uid_1_out_0, none-quant +output 8 dim 20 20 32 1, data_format=0, name=uid_20008_sub_uid_1_out_0, none-quant +output 9 dim 160 160 32 1, data_format=0, name=uid_20009_sub_uid_1_out_0, none-quant +nbg name=model/yolov8s-seg_10_uint8_t527.nb, size: 11076992. +create network 0: 34337 us. +prepare network: 22250 us. +buffer ptr: 0x10312600, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 88575 us. +output 0, ptr 0x1043e740, size 409600. +output 1, ptr 0x105ce7c0, size 512000. +output 2, ptr 0x107c2880, size 204800. +output 3, ptr 0x1088a900, size 102400. +output 4, ptr 0x108ee980, size 128000. +output 5, ptr 0x1096ba00, size 51200. +output 6, ptr 0x1099dac0, size 25600. +output 7, ptr 0x109b6b40, size 32000. +output 8, ptr 0x109d5fc0, size 12800. +output 9, ptr 0x109e2840, size 819200. +post process time : 54 ms +detection num: 3 + 1: 89%, [ 126, 133, 568, 426], bicycle +16: 95%, [ 131, 220, 310, 541], dog + 7: 66%, [ 470, 73, 688, 170], truck +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :---------- | :--------- | :----------- | :----------- | :----------- | :--------- | :------- | :------ | +| 全志 T527 | Vivante VIP9000 | yolov8s-seg | 640×640 | 34.3 ms | 22.3 ms | 88.6 ms | 54.0 ms | 199.2 ms | 5.0 FPS | + + + diff --git a/docs/common/ai/cubie/_model-zoo-yolov8.mdx b/docs/common/ai/cubie/_model-zoo-yolov8.mdx new file mode 100644 index 000000000..d1c472b47 --- /dev/null +++ b/docs/common/ai/cubie/_model-zoo-yolov8.mdx @@ -0,0 +1,375 @@ +本文档讲述如何在 NPU 上运行 YOLOv8。 + +:::info +参考 [Model Zoo 下载](./model-zoo-download)获取示例。 +::: + +YOLOv8 示例目录结构： + +```bash +$ tree ./ +./ +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ ├── python +│ │ └── onnx_extract.py +│ └── yolov8n_6.onnx +├── figures +│ ├── banner-yolo-vision-2023.png +│ ├── bus.jpg +│ ├── out_yolov8_640.png +│ └── yolo-comparison-plots.png +├── main.cpp +├── model +│ └── dog.jpg +├── model_config.h +├── README.md +├── yolov8_6_post.cpp +└── yolov8_6_pre.cpp +``` + +## 模型转换 + +### 配置虚拟环境 + + + +```bash +python -m venv .venv && source .venv/bin/activate +pip install ultralytics==8.1.0 onnxsim +``` + + + +### 导出 onnx 模型 + + + +```bash +cd convert_model/python/ +yolo export model=yolov8n.pt format=onnx dynamic=True opset=11 +``` + + + +### 固定 shape + + + +```bash +python3 -m onnxsim yolov8n.onnx yolov8n_640_sim.onnx --input-shape=1,3,640,640 +``` + + + +### 裁剪模型 + + + +```bash +python3 onnx_extract.py +cd .. +``` + + + +### 创建转换脚本的软链接 + + + +```bash +./convert_model_env.sh +``` + + + +### 模型导入/量化/转换 + +需要先进入容器开发环境。可以参考 Model Zoo 下载中[创建容器](./model-zoo-download#创建并启动容器)这一部分。 + +:::info +不同平台请使用对应的 Docker 镜像： + +- A733：ubuntu-npu:v2.0.10.1 +- T527：ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +进入容器对应目录之后运行脚本。 + + + +```bash +cd /workspace/examples/yolov8/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh yolov8n_6 +./pegasus_quantize.sh yolov8n_6 uint8 12 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolov8n_6 uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolov8n_6 uint8 t527 +``` + + + + + + +导出的模型文件存放在../model目录。 + +### 编译示例 + +接下来可以编译示例，**先 exit 退出容器**，然后执行下面的命令编译示例。 + +首先需要配置第三方库和交叉编译工具链。 + +:::info +如果你已经在其他示例中配置过第三方库和交叉编译工具链则可以跳过这一步。 +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +需要先手动[点击链接](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4)下载之后放到 0-toolchains/ 再执行下面的命令： + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/yolov8/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## 模型部署 + +编译示例完成之后，示例会安装到 install 目录，可以使用 scp 传输到板端。 + +### 配置 NPU 驱动 + +:::info +如果你已经在其他示例中配置过 NPU 驱动则可以跳过这一步。 +::: + +将驱动库 scp 传输到板端的 lib 目录。 + +- A733 对应 common/lib_linux_aarch64/A733 目录 +- T527 对应 common/lib_linux_aarch64/T527 目录 + +然后执行下面的命令导出到环境变量。 + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### 运行示例 + +配置好驱动之后就可以运行示例了。 + +:::tip +对于 T527 平台，你还需要参考 A5E 的`板端启用 NPU`文档先启用 NPU ，然后使用下面的命令增加当前用户使用 /dev/vipcore 的权限。 +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd yolov8_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./yolov8_demo_a733 +./yolov8_demo_a733 -nb model/yolov8n_6_uint8_a733.nb -i model/dog.jpg +``` + + + +运行结果如下： + +```bash +$ ./yolov8_demo_a733 -nb model/yolov8n_6_uint8_a733.nb -i model/dog.jpg +model_file=model/yolov8n_6_uint8_a733.nb, input=model/dog.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 80 80 64 1, data_format=0, name=uid_11_out_0b_uid_1_out_0, none-quant +output 1 dim 80 80 80 1, data_format=0, name=uid_10_out_0b_uid_1_out_0, none-quant +output 2 dim 40 40 64 1, data_format=0, name=uid_9_out_0ub_uid_1_out_0, none-quant +output 3 dim 40 40 80 1, data_format=0, name=uid_8_out_0ub_uid_1_out_0, none-quant +output 4 dim 20 20 64 1, data_format=0, name=uid_7_out_0ub_uid_1_out_0, none-quant +output 5 dim 20 20 80 1, data_format=0, name=uid_6_out_0ub_uid_1_out_0, none-quant +nbg name=model/yolov8n_6_uint8_a733.nb, size: 2452448. +create network 0: 11517 us. +prepare network: 1821 us. +buffer ptr: 0x10cdb600, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 12567 us. +output 0, ptr 0x10e07740, size 409600. +output 1, ptr 0x10f977c0, size 512000. +output 2, ptr 0x1118b840, size 102400. +output 3, ptr 0x111ef8c0, size 128000. +output 4, ptr 0x1126c980, size 25600. +output 5, ptr 0x11285a00, size 32000. +detection num: 3 + 1: 87%, [ 130, 136, 568, 419], bicycle +16: 95%, [ 131, 220, 308, 541], dog + 2: 68%, [ 467, 74, 695, 171], car +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :------ | :--------- | :----------- | :----------- | :----------- | :--------- | :------ | :------- | +| 全志 A733 | Vivante VIP9000 | yolov8n | 640×640 | 11.5 ms | 1.8 ms | 12.6 ms | | 25.9 ms | 38.6 FPS | + + + + + + + +```bash +cd yolov8_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./yolov8_demo_t527 +./yolov8_demo_t527 -nb model/yolov8n_6_uint8_t527.nb -i model/dog.jpg +``` + + + +运行结果如下： + +```bash +$ ./yolov8_demo_t527 -nb model/yolov8n_6_uint8_t527.nb -i model/dog.jpg +model_file=model/yolov8n_6_uint8_t527.nb, input=model/dog.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 80 80 64 1, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 80 80 80 1, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 40 40 64 1, data_format=0, name=uid_20003_sub_uid_1_out_0, none-quant +output 3 dim 40 40 80 1, data_format=0, name=uid_20004_sub_uid_1_out_0, none-quant +output 4 dim 20 20 64 1, data_format=0, name=uid_20005_sub_uid_1_out_0, none-quant +output 5 dim 20 20 80 1, data_format=0, name=uid_20006_sub_uid_1_out_0, none-quant +nbg name=model/yolov8n_6_uint8_t527.nb, size: 2915520. +create network 0: 16603 us. +prepare network: 5775 us. +buffer ptr: 0x3e36c600, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 30967 us. +output 0, ptr 0x3e498700, size 409600. +output 1, ptr 0x3e628780, size 512000. +output 2, ptr 0x3e81c840, size 102400. +output 3, ptr 0x3e8808c0, size 128000. +output 4, ptr 0x3e8fd940, size 25600. +output 5, ptr 0x3e9169c0, size 32000. +detection num: 3 + 1: 88%, [ 130, 135, 568, 419], bicycle +16: 93%, [ 131, 219, 307, 540], dog + 2: 68%, [ 466, 74, 695, 171], car +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :------ | :--------- | :----------- | :----------- | :----------- | :--------- | :------ | :------- | +| 全志 T527 | Vivante VIP9000 | yolov8n | 640×640 | 16.6 ms | 5.8 ms | 31.0 ms | | 53.4 ms | 18.7 FPS | + + + diff --git a/docs/common/ai/cubie/_model-zoo-yolox.mdx b/docs/common/ai/cubie/_model-zoo-yolox.mdx new file mode 100644 index 000000000..a2a2dbd45 --- /dev/null +++ b/docs/common/ai/cubie/_model-zoo-yolox.mdx @@ -0,0 +1,354 @@ +本文档讲述如何在 NPU 上运行 YOLOX。 + +:::info +参考 [Model Zoo 下载](./model-zoo-download)获取示例。 +::: + +YOLOX 示例目录结构： + +```bash +$ tree ./ +./ +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ └── python +│ ├── coco_classes.py +│ ├── demo_utils.py +│ ├── sub_model.py +│ ├── visualize.py +│ └── yolox_sim.py +├── figures +│ ├── output_yolox.png +│ └── yolox_rt.png +├── main.cpp +├── model +│ └── bus.jpg +├── model_config.h +├── README.md +├── yolox_postprocess.cpp +└── yolox_preprocess.cpp +``` + +## 模型转换 + +### 下载模型 + + + +```bash +cd convert_model/ +wget https://github.com/Megvii-BaseDetection/YOLOX/releases/download/0.1.1rc0/yolox_s.onnx +``` + + + +或者下载修改好的模型，点击下载 [yolox_s_sim.onnx](http://netstorage.allwinnertech.com:5000/sharing/G84PP2KvG) 。 + +然后移动到 convert_model/ 目录下。 + +### 裁剪模型 + +如果下载的是转换好的模型就可以跳过模型裁剪。 + + + +```bash +cd python/ +python3 sub_model.py +cd ../ +``` + + + +### 创建转换脚本的软链接 + + + +```bash +./convert_model_env.sh +``` + + + +### 模型导入/量化/转换 + +需要先进入容器开发环境。可以参考 Model Zoo 下载中[创建容器](./model-zoo-download#创建并启动容器)这一部分。 + +:::info +不同平台请使用对应的 Docker 镜像： + +- A733：ubuntu-npu:v2.0.10.1 +- T527：ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +进入容器对应目录之后运行脚本。 + + + +```bash +cd /workspace/examples/yolox/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh yolox_s_sim +./pegasus_quantize.sh yolox_s_sim uint8 12 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolox_s_sim uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolox_s_sim uint8 t527 +``` + + + + + + +导出的模型文件存放在../model目录。 + +### 编译示例 + +接下来可以编译示例，**先 exit 退出容器**，然后执行下面的命令编译示例。 + +首先需要配置第三方库和交叉编译工具链。 + +:::info +如果你已经在其他示例中配置过第三方库和交叉编译工具链则可以跳过这一步。 +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +需要先手动[点击链接](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4)下载之后放到 0-toolchains/ 再执行下面的命令： + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/yolox/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## 模型部署 + +编译示例完成之后，示例会安装到 install 目录，可以使用 scp 传输到板端。 + +### 配置 NPU 驱动 + +:::info +如果你已经在其他示例中配置过 NPU 驱动则可以跳过这一步。 +::: + +将驱动库 scp 传输到板端的 lib 目录。 + +- A733 对应 common/lib_linux_aarch64/A733 目录 +- T527 对应 common/lib_linux_aarch64/T527 目录 + +然后执行下面的命令导出到环境变量。 + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### 运行示例 + +配置好驱动之后就可以运行示例了。 + +:::tip +对于 T527 平台，你还需要参考 A5E 的`板端启用 NPU`文档先启用 NPU ，然后使用下面的命令增加当前用户使用 /dev/vipcore 的权限。 +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd yolox_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./yolox_demo_a733 +./yolox_demo_a733 -nb model/yolox_s_sim_uint8_a733.nb -i model/bus.jpg +``` + + + +运行结果如下： + +```bash +$ ./yolox_demo_a733 -nb model/yolox_s_sim_uint8_a733.nb -i model/bus.jpg +model_file=model/yolox_s_sim_uint8_a733.nb, input=model/bus.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 80 80 85 1, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 40 40 85 1, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 20 20 85 1, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +nbg name=model/yolox_s_sim_uint8_a733.nb, size: 7058512. +create network 0: 17292 us. +prepare network: 7783 us. +buffer ptr: 0x24831600, buffer size: 1228800 +Original image size: 640x640 +YOLOX preprocess completed: model/bus.jpg -> 640x640, buffer size: 1228800 +feed input cost: 11464 us. +network: 0, loop count: 1 +run time for this network 0: 30120 us. +detection num: 5 + 5: 93%, [ 85, 136, 555, 433], bus + 0: 89%, [ 113, 243, 199, 524], person + 0: 86%, [ 475, 239, 560, 520], person + 0: 89%, [ 213, 243, 283, 506], person + 0: 56%, [ 79, 328, 121, 515], person +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :------ | :--------- | :----------- | :----------- | :----------- | :--------- | :------ | :------- | +| 全志 A733 | Vivante VIP9000 | yolox_s | 640×640 | 17.3 ms | 7.8 ms | 30.1 ms | | 55.2 ms | 18.1 FPS | + + + + + + + +```bash +cd yolox_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./yolox_demo_t527 +./yolox_demo_t527 -nb model/yolox_s_sim_uint8_t527.nb -i model/bus.jpg +``` + + + +运行结果如下： + +```bash +$ ./yolox_demo_t527 -nb model/yolox_s_sim_uint8_t527.nb -i model/bus.jpg +model_file=model/yolox_s_sim_uint8_t527.nb, input=model/bus.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 80 80 85 1, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 40 40 85 1, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 20 20 85 1, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +nbg name=model/yolox_s_sim_uint8_t527.nb, size: 9132672. +create network 0: 25385 us. +prepare network: 18164 us. +buffer ptr: 0x116c6600, buffer size: 1228800 +Original image size: 640x640 +YOLOX preprocess completed: model/bus.jpg -> 640x640, buffer size: 1228800 +feed input cost: 62033 us. +network: 0, loop count: 1 +run time for this network 0: 73697 us. +detection num: 5 + 5: 93%, [ 98, 137, 550, 435], bus + 0: 89%, [ 107, 239, 210, 533], person + 0: 87%, [ 477, 239, 560, 519], person + 0: 89%, [ 214, 243, 283, 506], person + 0: 58%, [ 79, 329, 120, 516], person +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | 模型 | 输入分辨率 | 网络创建耗时 | 网络准备耗时 | 单帧推理耗时 | 后处理耗时 | 总耗时 | 帧率 | +| :-------- | :-------------- | :------ | :--------- | :----------- | :----------- | :----------- | :--------- | :------- | :------ | +| 全志 T527 | Vivante VIP9000 | yolox_s | 640×640 | 25.4 ms | 18.2 ms | 73.7 ms | | 117.3 ms | 8.5 FPS | + + + diff --git a/docs/cubie/a5e/app-dev/npu-dev/cubie-lenet.md b/docs/cubie/a5e/app-dev/npu-dev/_cubie-lenet.md similarity index 100% rename from docs/cubie/a5e/app-dev/npu-dev/cubie-lenet.md rename to docs/cubie/a5e/app-dev/npu-dev/_cubie-lenet.md diff --git a/docs/cubie/a5e/app-dev/npu-dev/cubie-resnet50.md b/docs/cubie/a5e/app-dev/npu-dev/_cubie-resnet50.md similarity index 100% rename from docs/cubie/a5e/app-dev/npu-dev/cubie-resnet50.md rename to docs/cubie/a5e/app-dev/npu-dev/_cubie-resnet50.md diff --git a/docs/cubie/a5e/app-dev/npu-dev/cubie-yolact.md b/docs/cubie/a5e/app-dev/npu-dev/_cubie-yolact.md similarity index 100% rename from docs/cubie/a5e/app-dev/npu-dev/cubie-yolact.md rename to docs/cubie/a5e/app-dev/npu-dev/_cubie-yolact.md diff --git a/docs/cubie/a5e/app-dev/npu-dev/cubie-yolov5.md b/docs/cubie/a5e/app-dev/npu-dev/_cubie-yolov5.md similarity index 100% rename from docs/cubie/a5e/app-dev/npu-dev/cubie-yolov5.md rename to docs/cubie/a5e/app-dev/npu-dev/_cubie-yolov5.md diff --git a/docs/cubie/a5e/app-dev/npu-dev/model-zoo/README.md b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/README.md new file mode 100644 index 000000000..231911c91 --- /dev/null +++ b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/README.md @@ -0,0 +1,9 @@ +--- +sidebar_position: 11 +--- + +# Model Zoo + +这里有一些预先准备好的模型部署案例可供参考。 + + diff --git a/docs/cubie/a5e/app-dev/npu-dev/model-zoo/_LPRNet.md b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/_LPRNet.md new file mode 100644 index 000000000..1a2a23cdd --- /dev/null +++ b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/_LPRNet.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 25 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-LPRNet.mdx +--- + +# LPRNet + +import LPRNet from '../../../../../common/ai/cubie/\_model-zoo-LPRNet.mdx'; + + diff --git a/docs/cubie/a5e/app-dev/npu-dev/model-zoo/_ppocr.md b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/_ppocr.md new file mode 100644 index 000000000..8b4a3a85e --- /dev/null +++ b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/_ppocr.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 16 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-ppocr.mdx +--- + +# PPOCR + +import PPOCR from '../../../../../common/ai/cubie/\_model-zoo-ppocr.mdx'; + + diff --git a/docs/cubie/a5e/app-dev/npu-dev/model-zoo/_yolo26-pose.md b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/_yolo26-pose.md new file mode 100644 index 000000000..e7d477529 --- /dev/null +++ b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/_yolo26-pose.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 13 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo26-pose.mdx +--- + +# YOLO26 Pose + +import YOLO26Pose from '../../../../../common/ai/cubie/\_model-zoo-yolo26-pose.mdx'; + + diff --git a/docs/cubie/a5e/app-dev/npu-dev/model-zoo/_yolov8-hybrid.md b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/_yolov8-hybrid.md new file mode 100644 index 000000000..d06d5de33 --- /dev/null +++ b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/_yolov8-hybrid.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 2 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8-hybrid.mdx +--- + +# YOLOv8 Hybrid + +import YOLOv8Hybrid from '../../../../../common/ai/cubie/\_model-zoo-yolov8-hybrid.mdx'; + + diff --git a/docs/cubie/a5e/app-dev/npu-dev/model-zoo/densenet121-keras.md b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/densenet121-keras.md new file mode 100644 index 000000000..ce02f3a5c --- /dev/null +++ b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/densenet121-keras.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 22 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-densenet121-keras.mdx +--- + +# DenseNet121 + +import DenseNet121 from '../../../../../common/ai/cubie/\_model-zoo-densenet121-keras.mdx'; + + diff --git a/docs/cubie/a5e/app-dev/npu-dev/model-zoo/lenet-caffe.md b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/lenet-caffe.md new file mode 100644 index 000000000..da2ea0ff6 --- /dev/null +++ b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/lenet-caffe.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 24 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-lenet-caffe.mdx +--- + +# LeNet + +import LeNet from '../../../../../common/ai/cubie/\_model-zoo-lenet-caffe.mdx'; + + diff --git a/docs/cubie/a5e/app-dev/npu-dev/model-zoo/mobilenetv1-tensorflow.md b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/mobilenetv1-tensorflow.md new file mode 100644 index 000000000..a552c39b2 --- /dev/null +++ b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/mobilenetv1-tensorflow.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 18 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-mobilenetv1-tensorflow.mdx +--- + +# MobileNetV1 + +import MobileNetV1 from '../../../../../common/ai/cubie/\_model-zoo-mobilenetv1-tensorflow.mdx'; + + diff --git a/docs/cubie/a5e/app-dev/npu-dev/model-zoo/mobilenetv2.md b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/mobilenetv2.md new file mode 100644 index 000000000..51ef4f03f --- /dev/null +++ b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/mobilenetv2.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 19 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-mobilenetv2.mdx +--- + +# MobileNetV2 + +import MobileNetV2 from '../../../../../common/ai/cubie/\_model-zoo-mobilenetv2.mdx'; + + diff --git a/docs/cubie/a5e/app-dev/npu-dev/model-zoo/model-zoo-download.md b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/model-zoo-download.md new file mode 100644 index 000000000..96a870dec --- /dev/null +++ b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/model-zoo-download.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 1 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-download.mdx +--- + +# Model Zoo 下载 + +import ModelZooDownload from '../../../../../common/ai/cubie/\_model-zoo-download.mdx'; + + diff --git a/docs/cubie/a5e/app-dev/npu-dev/model-zoo/ppseg.md b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/ppseg.md new file mode 100644 index 000000000..fe48c4427 --- /dev/null +++ b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/ppseg.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 17 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-ppseg.mdx +--- + +# PPSeg + +import PPSeg from '../../../../../common/ai/cubie/\_model-zoo-ppseg.mdx'; + + diff --git a/docs/cubie/a5e/app-dev/npu-dev/model-zoo/resnet50-tflite.md b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/resnet50-tflite.md new file mode 100644 index 000000000..475aa0858 --- /dev/null +++ b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/resnet50-tflite.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 20 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-resnet50-tflite.mdx +--- + +# ResNet50 TFLite + +import ResNet50TFLite from '../../../../../common/ai/cubie/\_model-zoo-resnet50-tflite.mdx'; + + diff --git a/docs/cubie/a5e/app-dev/npu-dev/model-zoo/resnet50v2.md b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/resnet50v2.md new file mode 100644 index 000000000..274417b93 --- /dev/null +++ b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/resnet50v2.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 21 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-resnet50v2.mdx +--- + +# ResNet50 V2 + +import ResNet50V2 from '../../../../../common/ai/cubie/\_model-zoo-resnet50v2.mdx'; + + diff --git a/docs/cubie/a5e/app-dev/npu-dev/model-zoo/retinaface.md b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/retinaface.md new file mode 100644 index 000000000..ee3da187a --- /dev/null +++ b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/retinaface.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 15 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-retinaface.mdx +--- + +# RetinaFace + +import RetinaFace from '../../../../../common/ai/cubie/\_model-zoo-retinaface.mdx'; + + diff --git a/docs/cubie/a5e/app-dev/npu-dev/model-zoo/squeezenet-pytorch.md b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/squeezenet-pytorch.md new file mode 100644 index 000000000..690bafe69 --- /dev/null +++ b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/squeezenet-pytorch.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 23 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-squeezenet-pytorch.mdx +--- + +# SqueezeNet + +import SqueezeNet from '../../../../../common/ai/cubie/\_model-zoo-squeezenet-pytorch.mdx'; + + diff --git a/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolo11-pose.md b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolo11-pose.md new file mode 100644 index 000000000..f1255f584 --- /dev/null +++ b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolo11-pose.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 5 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo11-pose.mdx +--- + +# YOLO11 Pose + +import YOLO11Pose from '../../../../../common/ai/cubie/\_model-zoo-yolo11-pose.mdx'; + + diff --git a/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolo11-seg.md b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolo11-seg.md new file mode 100644 index 000000000..672afce80 --- /dev/null +++ b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolo11-seg.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 4 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo11-seg.mdx +--- + +# YOLO11 Seg + +import YOLO11Seg from '../../../../../common/ai/cubie/\_model-zoo-yolo11-seg.mdx'; + + diff --git a/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolo11.md b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolo11.md new file mode 100644 index 000000000..300d57295 --- /dev/null +++ b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolo11.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 3 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo11.mdx +--- + +# YOLO11 + +import YOLO11 from '../../../../../common/ai/cubie/\_model-zoo-yolo11.mdx'; + + diff --git a/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolo26.md b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolo26.md new file mode 100644 index 000000000..f9987edc1 --- /dev/null +++ b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolo26.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 12 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo26.mdx +--- + +# YOLO26 + +import YOLO26 from '../../../../../common/ai/cubie/\_model-zoo-yolo26.mdx'; + + diff --git a/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolov3-darknet.md b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolov3-darknet.md new file mode 100644 index 000000000..7d191f006 --- /dev/null +++ b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolov3-darknet.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 9 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov3-darknet.mdx +--- + +# YOLOv3 + +import YOLOv3 from '../../../../../common/ai/cubie/\_model-zoo-yolov3-darknet.mdx'; + + diff --git a/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolov5.md b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolov5.md new file mode 100644 index 000000000..cd33bd150 --- /dev/null +++ b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolov5.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 10 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov5.mdx +--- + +# YOLOv5 + +import YOLOv5 from '../../../../../common/ai/cubie/\_model-zoo-yolov5.mdx'; + + diff --git a/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolov8-pose.md b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolov8-pose.md new file mode 100644 index 000000000..d29446157 --- /dev/null +++ b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolov8-pose.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 8 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8-pose.mdx +--- + +# YOLOv8 Pose + +import YOLOv8Pose from '../../../../../common/ai/cubie/\_model-zoo-yolov8-pose.mdx'; + + diff --git a/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolov8-seg.md b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolov8-seg.md new file mode 100644 index 000000000..451da2b12 --- /dev/null +++ b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolov8-seg.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 7 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8-seg.mdx +--- + +# YOLOv8 Seg + +import YOLOv8Seg from '../../../../../common/ai/cubie/\_model-zoo-yolov8-seg.mdx'; + + diff --git a/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolov8.md b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolov8.md new file mode 100644 index 000000000..1eb3ebbd7 --- /dev/null +++ b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolov8.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 6 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8.mdx +--- + +# YOLOv8 + +import YOLOv8 from '../../../../../common/ai/cubie/\_model-zoo-yolov8.mdx'; + + diff --git a/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolox.md b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolox.md new file mode 100644 index 000000000..ed26d8739 --- /dev/null +++ b/docs/cubie/a5e/app-dev/npu-dev/model-zoo/yolox.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 14 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolox.mdx +--- + +# YOLOX + +import YOLOX from '../../../../../common/ai/cubie/\_model-zoo-yolox.mdx'; + + diff --git a/docs/cubie/a7a/app-dev/npu-dev/cubie-lenet.md b/docs/cubie/a7a/app-dev/npu-dev/_cubie-lenet.md similarity index 91% rename from docs/cubie/a7a/app-dev/npu-dev/cubie-lenet.md rename to docs/cubie/a7a/app-dev/npu-dev/_cubie-lenet.md index 8ef2fbf53..99c031d8f 100644 --- a/docs/cubie/a7a/app-dev/npu-dev/cubie-lenet.md +++ b/docs/cubie/a7a/app-dev/npu-dev/_cubie-lenet.md @@ -1,5 +1,5 @@ --- -sidebar_position: 7 +sidebar_position: 8 doc_kind: wrapper source_of_truth: common diff --git a/docs/cubie/a7s/app-dev/npu-dev/cubie-resnet50.md b/docs/cubie/a7a/app-dev/npu-dev/_cubie-resnet50.md similarity index 92% rename from docs/cubie/a7s/app-dev/npu-dev/cubie-resnet50.md rename to docs/cubie/a7a/app-dev/npu-dev/_cubie-resnet50.md index b3592bbe8..877e66357 100644 --- a/docs/cubie/a7s/app-dev/npu-dev/cubie-resnet50.md +++ b/docs/cubie/a7a/app-dev/npu-dev/_cubie-resnet50.md @@ -1,5 +1,5 @@ --- -sidebar_position: 5 +sidebar_position: 6 doc_kind: wrapper source_of_truth: common diff --git a/docs/cubie/a7z/app-dev/npu-dev/cubie-yolact.md b/docs/cubie/a7a/app-dev/npu-dev/_cubie-yolact.md similarity index 91% rename from docs/cubie/a7z/app-dev/npu-dev/cubie-yolact.md rename to docs/cubie/a7a/app-dev/npu-dev/_cubie-yolact.md index 4523dc768..f35e0d6f4 100644 --- a/docs/cubie/a7z/app-dev/npu-dev/cubie-yolact.md +++ b/docs/cubie/a7a/app-dev/npu-dev/_cubie-yolact.md @@ -1,5 +1,5 @@ --- -sidebar_position: 6 +sidebar_position: 7 doc_kind: wrapper source_of_truth: common diff --git a/docs/cubie/a7z/app-dev/npu-dev/cubie-yolov5.md b/docs/cubie/a7a/app-dev/npu-dev/_cubie-yolov5.md similarity index 91% rename from docs/cubie/a7z/app-dev/npu-dev/cubie-yolov5.md rename to docs/cubie/a7a/app-dev/npu-dev/_cubie-yolov5.md index 512a90864..3c362fec1 100644 --- a/docs/cubie/a7z/app-dev/npu-dev/cubie-yolov5.md +++ b/docs/cubie/a7a/app-dev/npu-dev/_cubie-yolov5.md @@ -1,5 +1,5 @@ --- -sidebar_position: 4 +sidebar_position: 5 doc_kind: wrapper source_of_truth: common diff --git a/docs/cubie/a7a/app-dev/npu-dev/cubie-acuity-usage.md b/docs/cubie/a7a/app-dev/npu-dev/cubie-acuity-usage.md index 2ad143d96..78270bd79 100644 --- a/docs/cubie/a7a/app-dev/npu-dev/cubie-acuity-usage.md +++ b/docs/cubie/a7a/app-dev/npu-dev/cubie-acuity-usage.md @@ -1,5 +1,5 @@ --- -sidebar_position: 2 +sidebar_position: 3 doc_kind: wrapper source_of_truth: common diff --git a/docs/cubie/a7a/app-dev/npu-dev/cubie-nbinfo.md b/docs/cubie/a7a/app-dev/npu-dev/cubie-nbinfo.md index 3ba693d35..4fffe9258 100644 --- a/docs/cubie/a7a/app-dev/npu-dev/cubie-nbinfo.md +++ b/docs/cubie/a7a/app-dev/npu-dev/cubie-nbinfo.md @@ -1,5 +1,5 @@ --- -sidebar_position: 9 +sidebar_position: 10 doc_kind: wrapper source_of_truth: common diff --git a/docs/cubie/a7a/app-dev/npu-dev/cubie-quant-acc-improve.md b/docs/cubie/a7a/app-dev/npu-dev/cubie-quant-acc-improve.md index 8429d5d6d..498b0a628 100644 --- a/docs/cubie/a7a/app-dev/npu-dev/cubie-quant-acc-improve.md +++ b/docs/cubie/a7a/app-dev/npu-dev/cubie-quant-acc-improve.md @@ -1,5 +1,5 @@ --- -sidebar_position: 3 +sidebar_position: 4 doc_kind: wrapper source_of_truth: common diff --git a/docs/cubie/a7a/app-dev/npu-dev/cubie-vpm-run.md b/docs/cubie/a7a/app-dev/npu-dev/cubie-vpm-run.md index 270ed0ff9..79c8d1970 100644 --- a/docs/cubie/a7a/app-dev/npu-dev/cubie-vpm-run.md +++ b/docs/cubie/a7a/app-dev/npu-dev/cubie-vpm-run.md @@ -1,5 +1,5 @@ --- -sidebar_position: 8 +sidebar_position: 9 doc_kind: wrapper source_of_truth: common diff --git a/docs/cubie/a7a/app-dev/npu-dev/model-zoo/README.md b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/README.md new file mode 100644 index 000000000..231911c91 --- /dev/null +++ b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/README.md @@ -0,0 +1,9 @@ +--- +sidebar_position: 11 +--- + +# Model Zoo + +这里有一些预先准备好的模型部署案例可供参考。 + + diff --git a/docs/cubie/a7a/app-dev/npu-dev/model-zoo/_LPRNet.md b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/_LPRNet.md new file mode 100644 index 000000000..1a2a23cdd --- /dev/null +++ b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/_LPRNet.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 25 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-LPRNet.mdx +--- + +# LPRNet + +import LPRNet from '../../../../../common/ai/cubie/\_model-zoo-LPRNet.mdx'; + + diff --git a/docs/cubie/a7a/app-dev/npu-dev/model-zoo/_ppocr.md b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/_ppocr.md new file mode 100644 index 000000000..8b4a3a85e --- /dev/null +++ b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/_ppocr.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 16 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-ppocr.mdx +--- + +# PPOCR + +import PPOCR from '../../../../../common/ai/cubie/\_model-zoo-ppocr.mdx'; + + diff --git a/docs/cubie/a7a/app-dev/npu-dev/model-zoo/_yolo26-pose.md b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/_yolo26-pose.md new file mode 100644 index 000000000..e7d477529 --- /dev/null +++ b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/_yolo26-pose.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 13 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo26-pose.mdx +--- + +# YOLO26 Pose + +import YOLO26Pose from '../../../../../common/ai/cubie/\_model-zoo-yolo26-pose.mdx'; + + diff --git a/docs/cubie/a7a/app-dev/npu-dev/model-zoo/_yolov8-hybrid.md b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/_yolov8-hybrid.md new file mode 100644 index 000000000..d06d5de33 --- /dev/null +++ b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/_yolov8-hybrid.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 2 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8-hybrid.mdx +--- + +# YOLOv8 Hybrid + +import YOLOv8Hybrid from '../../../../../common/ai/cubie/\_model-zoo-yolov8-hybrid.mdx'; + + diff --git a/docs/cubie/a7a/app-dev/npu-dev/model-zoo/densenet121-keras.md b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/densenet121-keras.md new file mode 100644 index 000000000..ce02f3a5c --- /dev/null +++ b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/densenet121-keras.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 22 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-densenet121-keras.mdx +--- + +# DenseNet121 + +import DenseNet121 from '../../../../../common/ai/cubie/\_model-zoo-densenet121-keras.mdx'; + + diff --git a/docs/cubie/a7a/app-dev/npu-dev/model-zoo/lenet-caffe.md b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/lenet-caffe.md new file mode 100644 index 000000000..da2ea0ff6 --- /dev/null +++ b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/lenet-caffe.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 24 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-lenet-caffe.mdx +--- + +# LeNet + +import LeNet from '../../../../../common/ai/cubie/\_model-zoo-lenet-caffe.mdx'; + + diff --git a/docs/cubie/a7a/app-dev/npu-dev/model-zoo/mobilenetv1-tensorflow.md b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/mobilenetv1-tensorflow.md new file mode 100644 index 000000000..a552c39b2 --- /dev/null +++ b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/mobilenetv1-tensorflow.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 18 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-mobilenetv1-tensorflow.mdx +--- + +# MobileNetV1 + +import MobileNetV1 from '../../../../../common/ai/cubie/\_model-zoo-mobilenetv1-tensorflow.mdx'; + + diff --git a/docs/cubie/a7a/app-dev/npu-dev/model-zoo/mobilenetv2.md b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/mobilenetv2.md new file mode 100644 index 000000000..51ef4f03f --- /dev/null +++ b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/mobilenetv2.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 19 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-mobilenetv2.mdx +--- + +# MobileNetV2 + +import MobileNetV2 from '../../../../../common/ai/cubie/\_model-zoo-mobilenetv2.mdx'; + + diff --git a/docs/cubie/a7a/app-dev/npu-dev/model-zoo/model-zoo-download.md b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/model-zoo-download.md new file mode 100644 index 000000000..96a870dec --- /dev/null +++ b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/model-zoo-download.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 1 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-download.mdx +--- + +# Model Zoo 下载 + +import ModelZooDownload from '../../../../../common/ai/cubie/\_model-zoo-download.mdx'; + + diff --git a/docs/cubie/a7a/app-dev/npu-dev/model-zoo/ppseg.md b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/ppseg.md new file mode 100644 index 000000000..fe48c4427 --- /dev/null +++ b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/ppseg.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 17 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-ppseg.mdx +--- + +# PPSeg + +import PPSeg from '../../../../../common/ai/cubie/\_model-zoo-ppseg.mdx'; + + diff --git a/docs/cubie/a7a/app-dev/npu-dev/model-zoo/resnet50-tflite.md b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/resnet50-tflite.md new file mode 100644 index 000000000..475aa0858 --- /dev/null +++ b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/resnet50-tflite.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 20 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-resnet50-tflite.mdx +--- + +# ResNet50 TFLite + +import ResNet50TFLite from '../../../../../common/ai/cubie/\_model-zoo-resnet50-tflite.mdx'; + + diff --git a/docs/cubie/a7a/app-dev/npu-dev/model-zoo/resnet50v2.md b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/resnet50v2.md new file mode 100644 index 000000000..274417b93 --- /dev/null +++ b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/resnet50v2.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 21 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-resnet50v2.mdx +--- + +# ResNet50 V2 + +import ResNet50V2 from '../../../../../common/ai/cubie/\_model-zoo-resnet50v2.mdx'; + + diff --git a/docs/cubie/a7a/app-dev/npu-dev/model-zoo/retinaface.md b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/retinaface.md new file mode 100644 index 000000000..ee3da187a --- /dev/null +++ b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/retinaface.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 15 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-retinaface.mdx +--- + +# RetinaFace + +import RetinaFace from '../../../../../common/ai/cubie/\_model-zoo-retinaface.mdx'; + + diff --git a/docs/cubie/a7a/app-dev/npu-dev/model-zoo/squeezenet-pytorch.md b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/squeezenet-pytorch.md new file mode 100644 index 000000000..690bafe69 --- /dev/null +++ b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/squeezenet-pytorch.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 23 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-squeezenet-pytorch.mdx +--- + +# SqueezeNet + +import SqueezeNet from '../../../../../common/ai/cubie/\_model-zoo-squeezenet-pytorch.mdx'; + + diff --git a/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolo11-pose.md b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolo11-pose.md new file mode 100644 index 000000000..f1255f584 --- /dev/null +++ b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolo11-pose.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 5 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo11-pose.mdx +--- + +# YOLO11 Pose + +import YOLO11Pose from '../../../../../common/ai/cubie/\_model-zoo-yolo11-pose.mdx'; + + diff --git a/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolo11-seg.md b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolo11-seg.md new file mode 100644 index 000000000..672afce80 --- /dev/null +++ b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolo11-seg.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 4 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo11-seg.mdx +--- + +# YOLO11 Seg + +import YOLO11Seg from '../../../../../common/ai/cubie/\_model-zoo-yolo11-seg.mdx'; + + diff --git a/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolo11.md b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolo11.md new file mode 100644 index 000000000..300d57295 --- /dev/null +++ b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolo11.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 3 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo11.mdx +--- + +# YOLO11 + +import YOLO11 from '../../../../../common/ai/cubie/\_model-zoo-yolo11.mdx'; + + diff --git a/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolo26.md b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolo26.md new file mode 100644 index 000000000..f9987edc1 --- /dev/null +++ b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolo26.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 12 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo26.mdx +--- + +# YOLO26 + +import YOLO26 from '../../../../../common/ai/cubie/\_model-zoo-yolo26.mdx'; + + diff --git a/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolov3-darknet.md b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolov3-darknet.md new file mode 100644 index 000000000..7d191f006 --- /dev/null +++ b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolov3-darknet.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 9 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov3-darknet.mdx +--- + +# YOLOv3 + +import YOLOv3 from '../../../../../common/ai/cubie/\_model-zoo-yolov3-darknet.mdx'; + + diff --git a/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolov5.md b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolov5.md new file mode 100644 index 000000000..cd33bd150 --- /dev/null +++ b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolov5.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 10 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov5.mdx +--- + +# YOLOv5 + +import YOLOv5 from '../../../../../common/ai/cubie/\_model-zoo-yolov5.mdx'; + + diff --git a/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolov8-pose.md b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolov8-pose.md new file mode 100644 index 000000000..d29446157 --- /dev/null +++ b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolov8-pose.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 8 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8-pose.mdx +--- + +# YOLOv8 Pose + +import YOLOv8Pose from '../../../../../common/ai/cubie/\_model-zoo-yolov8-pose.mdx'; + + diff --git a/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolov8-seg.md b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolov8-seg.md new file mode 100644 index 000000000..451da2b12 --- /dev/null +++ b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolov8-seg.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 7 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8-seg.mdx +--- + +# YOLOv8 Seg + +import YOLOv8Seg from '../../../../../common/ai/cubie/\_model-zoo-yolov8-seg.mdx'; + + diff --git a/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolov8.md b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolov8.md new file mode 100644 index 000000000..1eb3ebbd7 --- /dev/null +++ b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolov8.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 6 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8.mdx +--- + +# YOLOv8 + +import YOLOv8 from '../../../../../common/ai/cubie/\_model-zoo-yolov8.mdx'; + + diff --git a/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolox.md b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolox.md new file mode 100644 index 000000000..ed26d8739 --- /dev/null +++ b/docs/cubie/a7a/app-dev/npu-dev/model-zoo/yolox.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 14 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolox.mdx +--- + +# YOLOX + +import YOLOX from '../../../../../common/ai/cubie/\_model-zoo-yolox.mdx'; + + diff --git a/docs/cubie/a7s/app-dev/npu-dev/cuibie-lenet.md b/docs/cubie/a7s/app-dev/npu-dev/_cubie-lenet.md similarity index 91% rename from docs/cubie/a7s/app-dev/npu-dev/cuibie-lenet.md rename to docs/cubie/a7s/app-dev/npu-dev/_cubie-lenet.md index 8ef2fbf53..99c031d8f 100644 --- a/docs/cubie/a7s/app-dev/npu-dev/cuibie-lenet.md +++ b/docs/cubie/a7s/app-dev/npu-dev/_cubie-lenet.md @@ -1,5 +1,5 @@ --- -sidebar_position: 7 +sidebar_position: 8 doc_kind: wrapper source_of_truth: common diff --git a/docs/cubie/a7a/app-dev/npu-dev/cubie-resnet50.md b/docs/cubie/a7s/app-dev/npu-dev/_cubie-resnet50.md similarity index 92% rename from docs/cubie/a7a/app-dev/npu-dev/cubie-resnet50.md rename to docs/cubie/a7s/app-dev/npu-dev/_cubie-resnet50.md index b3592bbe8..877e66357 100644 --- a/docs/cubie/a7a/app-dev/npu-dev/cubie-resnet50.md +++ b/docs/cubie/a7s/app-dev/npu-dev/_cubie-resnet50.md @@ -1,5 +1,5 @@ --- -sidebar_position: 5 +sidebar_position: 6 doc_kind: wrapper source_of_truth: common diff --git a/docs/cubie/a7s/app-dev/npu-dev/cubie-yolact.md b/docs/cubie/a7s/app-dev/npu-dev/_cubie-yolact.md similarity index 91% rename from docs/cubie/a7s/app-dev/npu-dev/cubie-yolact.md rename to docs/cubie/a7s/app-dev/npu-dev/_cubie-yolact.md index 4523dc768..f35e0d6f4 100644 --- a/docs/cubie/a7s/app-dev/npu-dev/cubie-yolact.md +++ b/docs/cubie/a7s/app-dev/npu-dev/_cubie-yolact.md @@ -1,5 +1,5 @@ --- -sidebar_position: 6 +sidebar_position: 7 doc_kind: wrapper source_of_truth: common diff --git a/docs/cubie/a7s/app-dev/npu-dev/cubie-yolov5.md b/docs/cubie/a7s/app-dev/npu-dev/_cubie-yolov5.md similarity index 91% rename from docs/cubie/a7s/app-dev/npu-dev/cubie-yolov5.md rename to docs/cubie/a7s/app-dev/npu-dev/_cubie-yolov5.md index 512a90864..3c362fec1 100644 --- a/docs/cubie/a7s/app-dev/npu-dev/cubie-yolov5.md +++ b/docs/cubie/a7s/app-dev/npu-dev/_cubie-yolov5.md @@ -1,5 +1,5 @@ --- -sidebar_position: 4 +sidebar_position: 5 doc_kind: wrapper source_of_truth: common diff --git a/docs/cubie/a7s/app-dev/npu-dev/cubie-acuity-env.md b/docs/cubie/a7s/app-dev/npu-dev/cubie-acuity-env.md index 371b4b9c5..f0c737182 100644 --- a/docs/cubie/a7s/app-dev/npu-dev/cubie-acuity-env.md +++ b/docs/cubie/a7s/app-dev/npu-dev/cubie-acuity-env.md @@ -1,5 +1,5 @@ --- -sidebar_position: 1 +sidebar_position: 2 doc_kind: wrapper source_of_truth: common diff --git a/docs/cubie/a7s/app-dev/npu-dev/cubie-acuity-sdk.md b/docs/cubie/a7s/app-dev/npu-dev/cubie-acuity-sdk.md index 728b72067..b06622de7 100644 --- a/docs/cubie/a7s/app-dev/npu-dev/cubie-acuity-sdk.md +++ b/docs/cubie/a7s/app-dev/npu-dev/cubie-acuity-sdk.md @@ -1,5 +1,5 @@ --- -sidebar_position: 0 +sidebar_position: 1 doc_kind: wrapper source_of_truth: common diff --git a/docs/cubie/a7s/app-dev/npu-dev/cubie-acuity-usage.md b/docs/cubie/a7s/app-dev/npu-dev/cubie-acuity-usage.md index 2ad143d96..78270bd79 100644 --- a/docs/cubie/a7s/app-dev/npu-dev/cubie-acuity-usage.md +++ b/docs/cubie/a7s/app-dev/npu-dev/cubie-acuity-usage.md @@ -1,5 +1,5 @@ --- -sidebar_position: 2 +sidebar_position: 3 doc_kind: wrapper source_of_truth: common diff --git a/docs/cubie/a7s/app-dev/npu-dev/cubie-nbinfo.md b/docs/cubie/a7s/app-dev/npu-dev/cubie-nbinfo.md index 3ba693d35..4fffe9258 100644 --- a/docs/cubie/a7s/app-dev/npu-dev/cubie-nbinfo.md +++ b/docs/cubie/a7s/app-dev/npu-dev/cubie-nbinfo.md @@ -1,5 +1,5 @@ --- -sidebar_position: 9 +sidebar_position: 10 doc_kind: wrapper source_of_truth: common diff --git a/docs/cubie/a7s/app-dev/npu-dev/cubie-quant-acc-improve.md b/docs/cubie/a7s/app-dev/npu-dev/cubie-quant-acc-improve.md index 8429d5d6d..498b0a628 100644 --- a/docs/cubie/a7s/app-dev/npu-dev/cubie-quant-acc-improve.md +++ b/docs/cubie/a7s/app-dev/npu-dev/cubie-quant-acc-improve.md @@ -1,5 +1,5 @@ --- -sidebar_position: 3 +sidebar_position: 4 doc_kind: wrapper source_of_truth: common diff --git a/docs/cubie/a7s/app-dev/npu-dev/cubie-vpm-run.md b/docs/cubie/a7s/app-dev/npu-dev/cubie-vpm-run.md index 270ed0ff9..79c8d1970 100644 --- a/docs/cubie/a7s/app-dev/npu-dev/cubie-vpm-run.md +++ b/docs/cubie/a7s/app-dev/npu-dev/cubie-vpm-run.md @@ -1,5 +1,5 @@ --- -sidebar_position: 8 +sidebar_position: 9 doc_kind: wrapper source_of_truth: common diff --git a/docs/cubie/a7s/app-dev/npu-dev/model-zoo/README.md b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/README.md new file mode 100644 index 000000000..231911c91 --- /dev/null +++ b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/README.md @@ -0,0 +1,9 @@ +--- +sidebar_position: 11 +--- + +# Model Zoo + +这里有一些预先准备好的模型部署案例可供参考。 + + diff --git a/docs/cubie/a7s/app-dev/npu-dev/model-zoo/_LPRNet.md b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/_LPRNet.md new file mode 100644 index 000000000..1a2a23cdd --- /dev/null +++ b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/_LPRNet.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 25 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-LPRNet.mdx +--- + +# LPRNet + +import LPRNet from '../../../../../common/ai/cubie/\_model-zoo-LPRNet.mdx'; + + diff --git a/docs/cubie/a7s/app-dev/npu-dev/model-zoo/_ppocr.md b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/_ppocr.md new file mode 100644 index 000000000..8b4a3a85e --- /dev/null +++ b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/_ppocr.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 16 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-ppocr.mdx +--- + +# PPOCR + +import PPOCR from '../../../../../common/ai/cubie/\_model-zoo-ppocr.mdx'; + + diff --git a/docs/cubie/a7s/app-dev/npu-dev/model-zoo/_yolo26-pose.md b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/_yolo26-pose.md new file mode 100644 index 000000000..e7d477529 --- /dev/null +++ b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/_yolo26-pose.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 13 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo26-pose.mdx +--- + +# YOLO26 Pose + +import YOLO26Pose from '../../../../../common/ai/cubie/\_model-zoo-yolo26-pose.mdx'; + + diff --git a/docs/cubie/a7s/app-dev/npu-dev/model-zoo/_yolov8-hybrid.md b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/_yolov8-hybrid.md new file mode 100644 index 000000000..d06d5de33 --- /dev/null +++ b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/_yolov8-hybrid.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 2 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8-hybrid.mdx +--- + +# YOLOv8 Hybrid + +import YOLOv8Hybrid from '../../../../../common/ai/cubie/\_model-zoo-yolov8-hybrid.mdx'; + + diff --git a/docs/cubie/a7s/app-dev/npu-dev/model-zoo/densenet121-keras.md b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/densenet121-keras.md new file mode 100644 index 000000000..ce02f3a5c --- /dev/null +++ b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/densenet121-keras.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 22 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-densenet121-keras.mdx +--- + +# DenseNet121 + +import DenseNet121 from '../../../../../common/ai/cubie/\_model-zoo-densenet121-keras.mdx'; + + diff --git a/docs/cubie/a7s/app-dev/npu-dev/model-zoo/lenet-caffe.md b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/lenet-caffe.md new file mode 100644 index 000000000..da2ea0ff6 --- /dev/null +++ b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/lenet-caffe.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 24 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-lenet-caffe.mdx +--- + +# LeNet + +import LeNet from '../../../../../common/ai/cubie/\_model-zoo-lenet-caffe.mdx'; + + diff --git a/docs/cubie/a7s/app-dev/npu-dev/model-zoo/mobilenetv1-tensorflow.md b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/mobilenetv1-tensorflow.md new file mode 100644 index 000000000..a552c39b2 --- /dev/null +++ b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/mobilenetv1-tensorflow.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 18 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-mobilenetv1-tensorflow.mdx +--- + +# MobileNetV1 + +import MobileNetV1 from '../../../../../common/ai/cubie/\_model-zoo-mobilenetv1-tensorflow.mdx'; + + diff --git a/docs/cubie/a7s/app-dev/npu-dev/model-zoo/mobilenetv2.md b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/mobilenetv2.md new file mode 100644 index 000000000..51ef4f03f --- /dev/null +++ b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/mobilenetv2.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 19 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-mobilenetv2.mdx +--- + +# MobileNetV2 + +import MobileNetV2 from '../../../../../common/ai/cubie/\_model-zoo-mobilenetv2.mdx'; + + diff --git a/docs/cubie/a7s/app-dev/npu-dev/model-zoo/model-zoo-download.md b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/model-zoo-download.md new file mode 100644 index 000000000..96a870dec --- /dev/null +++ b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/model-zoo-download.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 1 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-download.mdx +--- + +# Model Zoo 下载 + +import ModelZooDownload from '../../../../../common/ai/cubie/\_model-zoo-download.mdx'; + + diff --git a/docs/cubie/a7s/app-dev/npu-dev/model-zoo/ppseg.md b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/ppseg.md new file mode 100644 index 000000000..fe48c4427 --- /dev/null +++ b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/ppseg.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 17 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-ppseg.mdx +--- + +# PPSeg + +import PPSeg from '../../../../../common/ai/cubie/\_model-zoo-ppseg.mdx'; + + diff --git a/docs/cubie/a7s/app-dev/npu-dev/model-zoo/resnet50-tflite.md b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/resnet50-tflite.md new file mode 100644 index 000000000..475aa0858 --- /dev/null +++ b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/resnet50-tflite.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 20 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-resnet50-tflite.mdx +--- + +# ResNet50 TFLite + +import ResNet50TFLite from '../../../../../common/ai/cubie/\_model-zoo-resnet50-tflite.mdx'; + + diff --git a/docs/cubie/a7s/app-dev/npu-dev/model-zoo/resnet50v2.md b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/resnet50v2.md new file mode 100644 index 000000000..274417b93 --- /dev/null +++ b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/resnet50v2.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 21 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-resnet50v2.mdx +--- + +# ResNet50 V2 + +import ResNet50V2 from '../../../../../common/ai/cubie/\_model-zoo-resnet50v2.mdx'; + + diff --git a/docs/cubie/a7s/app-dev/npu-dev/model-zoo/retinaface.md b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/retinaface.md new file mode 100644 index 000000000..ee3da187a --- /dev/null +++ b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/retinaface.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 15 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-retinaface.mdx +--- + +# RetinaFace + +import RetinaFace from '../../../../../common/ai/cubie/\_model-zoo-retinaface.mdx'; + + diff --git a/docs/cubie/a7s/app-dev/npu-dev/model-zoo/squeezenet-pytorch.md b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/squeezenet-pytorch.md new file mode 100644 index 000000000..690bafe69 --- /dev/null +++ b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/squeezenet-pytorch.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 23 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-squeezenet-pytorch.mdx +--- + +# SqueezeNet + +import SqueezeNet from '../../../../../common/ai/cubie/\_model-zoo-squeezenet-pytorch.mdx'; + + diff --git a/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolo11-pose.md b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolo11-pose.md new file mode 100644 index 000000000..f1255f584 --- /dev/null +++ b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolo11-pose.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 5 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo11-pose.mdx +--- + +# YOLO11 Pose + +import YOLO11Pose from '../../../../../common/ai/cubie/\_model-zoo-yolo11-pose.mdx'; + + diff --git a/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolo11-seg.md b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolo11-seg.md new file mode 100644 index 000000000..672afce80 --- /dev/null +++ b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolo11-seg.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 4 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo11-seg.mdx +--- + +# YOLO11 Seg + +import YOLO11Seg from '../../../../../common/ai/cubie/\_model-zoo-yolo11-seg.mdx'; + + diff --git a/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolo11.md b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolo11.md new file mode 100644 index 000000000..300d57295 --- /dev/null +++ b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolo11.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 3 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo11.mdx +--- + +# YOLO11 + +import YOLO11 from '../../../../../common/ai/cubie/\_model-zoo-yolo11.mdx'; + + diff --git a/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolo26.md b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolo26.md new file mode 100644 index 000000000..f9987edc1 --- /dev/null +++ b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolo26.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 12 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo26.mdx +--- + +# YOLO26 + +import YOLO26 from '../../../../../common/ai/cubie/\_model-zoo-yolo26.mdx'; + + diff --git a/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolov3-darknet.md b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolov3-darknet.md new file mode 100644 index 000000000..7d191f006 --- /dev/null +++ b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolov3-darknet.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 9 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov3-darknet.mdx +--- + +# YOLOv3 + +import YOLOv3 from '../../../../../common/ai/cubie/\_model-zoo-yolov3-darknet.mdx'; + + diff --git a/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolov5.md b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolov5.md new file mode 100644 index 000000000..cd33bd150 --- /dev/null +++ b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolov5.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 10 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov5.mdx +--- + +# YOLOv5 + +import YOLOv5 from '../../../../../common/ai/cubie/\_model-zoo-yolov5.mdx'; + + diff --git a/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolov8-pose.md b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolov8-pose.md new file mode 100644 index 000000000..d29446157 --- /dev/null +++ b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolov8-pose.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 8 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8-pose.mdx +--- + +# YOLOv8 Pose + +import YOLOv8Pose from '../../../../../common/ai/cubie/\_model-zoo-yolov8-pose.mdx'; + + diff --git a/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolov8-seg.md b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolov8-seg.md new file mode 100644 index 000000000..451da2b12 --- /dev/null +++ b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolov8-seg.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 7 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8-seg.mdx +--- + +# YOLOv8 Seg + +import YOLOv8Seg from '../../../../../common/ai/cubie/\_model-zoo-yolov8-seg.mdx'; + + diff --git a/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolov8.md b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolov8.md new file mode 100644 index 000000000..1eb3ebbd7 --- /dev/null +++ b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolov8.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 6 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8.mdx +--- + +# YOLOv8 + +import YOLOv8 from '../../../../../common/ai/cubie/\_model-zoo-yolov8.mdx'; + + diff --git a/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolox.md b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolox.md new file mode 100644 index 000000000..ed26d8739 --- /dev/null +++ b/docs/cubie/a7s/app-dev/npu-dev/model-zoo/yolox.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 14 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolox.mdx +--- + +# YOLOX + +import YOLOX from '../../../../../common/ai/cubie/\_model-zoo-yolox.mdx'; + + diff --git a/docs/cubie/a7z/app-dev/npu-dev/cubie-lenet.md b/docs/cubie/a7z/app-dev/npu-dev/_cubie-lenet.md similarity index 91% rename from docs/cubie/a7z/app-dev/npu-dev/cubie-lenet.md rename to docs/cubie/a7z/app-dev/npu-dev/_cubie-lenet.md index 8ef2fbf53..99c031d8f 100644 --- a/docs/cubie/a7z/app-dev/npu-dev/cubie-lenet.md +++ b/docs/cubie/a7z/app-dev/npu-dev/_cubie-lenet.md @@ -1,5 +1,5 @@ --- -sidebar_position: 7 +sidebar_position: 8 doc_kind: wrapper source_of_truth: common diff --git a/docs/cubie/a7z/app-dev/npu-dev/cubie-resnet50.md b/docs/cubie/a7z/app-dev/npu-dev/_cubie-resnet50.md similarity index 92% rename from docs/cubie/a7z/app-dev/npu-dev/cubie-resnet50.md rename to docs/cubie/a7z/app-dev/npu-dev/_cubie-resnet50.md index b3592bbe8..877e66357 100644 --- a/docs/cubie/a7z/app-dev/npu-dev/cubie-resnet50.md +++ b/docs/cubie/a7z/app-dev/npu-dev/_cubie-resnet50.md @@ -1,5 +1,5 @@ --- -sidebar_position: 5 +sidebar_position: 6 doc_kind: wrapper source_of_truth: common diff --git a/docs/cubie/a7a/app-dev/npu-dev/cubie-yolact.md b/docs/cubie/a7z/app-dev/npu-dev/_cubie-yolact.md similarity index 91% rename from docs/cubie/a7a/app-dev/npu-dev/cubie-yolact.md rename to docs/cubie/a7z/app-dev/npu-dev/_cubie-yolact.md index 4523dc768..f35e0d6f4 100644 --- a/docs/cubie/a7a/app-dev/npu-dev/cubie-yolact.md +++ b/docs/cubie/a7z/app-dev/npu-dev/_cubie-yolact.md @@ -1,5 +1,5 @@ --- -sidebar_position: 6 +sidebar_position: 7 doc_kind: wrapper source_of_truth: common diff --git a/docs/cubie/a7a/app-dev/npu-dev/cubie-yolov5.md b/docs/cubie/a7z/app-dev/npu-dev/_cubie-yolov5.md similarity index 91% rename from docs/cubie/a7a/app-dev/npu-dev/cubie-yolov5.md rename to docs/cubie/a7z/app-dev/npu-dev/_cubie-yolov5.md index 512a90864..3c362fec1 100644 --- a/docs/cubie/a7a/app-dev/npu-dev/cubie-yolov5.md +++ b/docs/cubie/a7z/app-dev/npu-dev/_cubie-yolov5.md @@ -1,5 +1,5 @@ --- -sidebar_position: 4 +sidebar_position: 5 doc_kind: wrapper source_of_truth: common diff --git a/docs/cubie/a7z/app-dev/npu-dev/cubie-acuity-usage.md b/docs/cubie/a7z/app-dev/npu-dev/cubie-acuity-usage.md index 2ad143d96..78270bd79 100644 --- a/docs/cubie/a7z/app-dev/npu-dev/cubie-acuity-usage.md +++ b/docs/cubie/a7z/app-dev/npu-dev/cubie-acuity-usage.md @@ -1,5 +1,5 @@ --- -sidebar_position: 2 +sidebar_position: 3 doc_kind: wrapper source_of_truth: common diff --git a/docs/cubie/a7z/app-dev/npu-dev/cubie-nbinfo.md b/docs/cubie/a7z/app-dev/npu-dev/cubie-nbinfo.md index 3ba693d35..4fffe9258 100644 --- a/docs/cubie/a7z/app-dev/npu-dev/cubie-nbinfo.md +++ b/docs/cubie/a7z/app-dev/npu-dev/cubie-nbinfo.md @@ -1,5 +1,5 @@ --- -sidebar_position: 9 +sidebar_position: 10 doc_kind: wrapper source_of_truth: common diff --git a/docs/cubie/a7z/app-dev/npu-dev/cubie-quant-acc-improve.md b/docs/cubie/a7z/app-dev/npu-dev/cubie-quant-acc-improve.md index 8429d5d6d..498b0a628 100644 --- a/docs/cubie/a7z/app-dev/npu-dev/cubie-quant-acc-improve.md +++ b/docs/cubie/a7z/app-dev/npu-dev/cubie-quant-acc-improve.md @@ -1,5 +1,5 @@ --- -sidebar_position: 3 +sidebar_position: 4 doc_kind: wrapper source_of_truth: common diff --git a/docs/cubie/a7z/app-dev/npu-dev/cubie-vpm-run.md b/docs/cubie/a7z/app-dev/npu-dev/cubie-vpm-run.md index 270ed0ff9..79c8d1970 100644 --- a/docs/cubie/a7z/app-dev/npu-dev/cubie-vpm-run.md +++ b/docs/cubie/a7z/app-dev/npu-dev/cubie-vpm-run.md @@ -1,5 +1,5 @@ --- -sidebar_position: 8 +sidebar_position: 9 doc_kind: wrapper source_of_truth: common diff --git a/docs/cubie/a7z/app-dev/npu-dev/model-zoo/README.md b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/README.md new file mode 100644 index 000000000..231911c91 --- /dev/null +++ b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/README.md @@ -0,0 +1,9 @@ +--- +sidebar_position: 11 +--- + +# Model Zoo + +这里有一些预先准备好的模型部署案例可供参考。 + + diff --git a/docs/cubie/a7z/app-dev/npu-dev/model-zoo/_LPRNet.md b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/_LPRNet.md new file mode 100644 index 000000000..1a2a23cdd --- /dev/null +++ b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/_LPRNet.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 25 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-LPRNet.mdx +--- + +# LPRNet + +import LPRNet from '../../../../../common/ai/cubie/\_model-zoo-LPRNet.mdx'; + + diff --git a/docs/cubie/a7z/app-dev/npu-dev/model-zoo/_ppocr.md b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/_ppocr.md new file mode 100644 index 000000000..8b4a3a85e --- /dev/null +++ b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/_ppocr.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 16 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-ppocr.mdx +--- + +# PPOCR + +import PPOCR from '../../../../../common/ai/cubie/\_model-zoo-ppocr.mdx'; + + diff --git a/docs/cubie/a7z/app-dev/npu-dev/model-zoo/_yolo26-pose.md b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/_yolo26-pose.md new file mode 100644 index 000000000..e7d477529 --- /dev/null +++ b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/_yolo26-pose.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 13 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo26-pose.mdx +--- + +# YOLO26 Pose + +import YOLO26Pose from '../../../../../common/ai/cubie/\_model-zoo-yolo26-pose.mdx'; + + diff --git a/docs/cubie/a7z/app-dev/npu-dev/model-zoo/_yolov8-hybrid.md b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/_yolov8-hybrid.md new file mode 100644 index 000000000..d06d5de33 --- /dev/null +++ b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/_yolov8-hybrid.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 2 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8-hybrid.mdx +--- + +# YOLOv8 Hybrid + +import YOLOv8Hybrid from '../../../../../common/ai/cubie/\_model-zoo-yolov8-hybrid.mdx'; + + diff --git a/docs/cubie/a7z/app-dev/npu-dev/model-zoo/densenet121-keras.md b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/densenet121-keras.md new file mode 100644 index 000000000..ce02f3a5c --- /dev/null +++ b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/densenet121-keras.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 22 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-densenet121-keras.mdx +--- + +# DenseNet121 + +import DenseNet121 from '../../../../../common/ai/cubie/\_model-zoo-densenet121-keras.mdx'; + + diff --git a/docs/cubie/a7z/app-dev/npu-dev/model-zoo/lenet-caffe.md b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/lenet-caffe.md new file mode 100644 index 000000000..da2ea0ff6 --- /dev/null +++ b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/lenet-caffe.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 24 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-lenet-caffe.mdx +--- + +# LeNet + +import LeNet from '../../../../../common/ai/cubie/\_model-zoo-lenet-caffe.mdx'; + + diff --git a/docs/cubie/a7z/app-dev/npu-dev/model-zoo/mobilenetv1-tensorflow.md b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/mobilenetv1-tensorflow.md new file mode 100644 index 000000000..a552c39b2 --- /dev/null +++ b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/mobilenetv1-tensorflow.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 18 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-mobilenetv1-tensorflow.mdx +--- + +# MobileNetV1 + +import MobileNetV1 from '../../../../../common/ai/cubie/\_model-zoo-mobilenetv1-tensorflow.mdx'; + + diff --git a/docs/cubie/a7z/app-dev/npu-dev/model-zoo/mobilenetv2.md b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/mobilenetv2.md new file mode 100644 index 000000000..51ef4f03f --- /dev/null +++ b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/mobilenetv2.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 19 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-mobilenetv2.mdx +--- + +# MobileNetV2 + +import MobileNetV2 from '../../../../../common/ai/cubie/\_model-zoo-mobilenetv2.mdx'; + + diff --git a/docs/cubie/a7z/app-dev/npu-dev/model-zoo/model-zoo-download.md b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/model-zoo-download.md new file mode 100644 index 000000000..96a870dec --- /dev/null +++ b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/model-zoo-download.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 1 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-download.mdx +--- + +# Model Zoo 下载 + +import ModelZooDownload from '../../../../../common/ai/cubie/\_model-zoo-download.mdx'; + + diff --git a/docs/cubie/a7z/app-dev/npu-dev/model-zoo/ppseg.md b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/ppseg.md new file mode 100644 index 000000000..fe48c4427 --- /dev/null +++ b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/ppseg.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 17 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-ppseg.mdx +--- + +# PPSeg + +import PPSeg from '../../../../../common/ai/cubie/\_model-zoo-ppseg.mdx'; + + diff --git a/docs/cubie/a7z/app-dev/npu-dev/model-zoo/resnet50-tflite.md b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/resnet50-tflite.md new file mode 100644 index 000000000..475aa0858 --- /dev/null +++ b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/resnet50-tflite.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 20 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-resnet50-tflite.mdx +--- + +# ResNet50 TFLite + +import ResNet50TFLite from '../../../../../common/ai/cubie/\_model-zoo-resnet50-tflite.mdx'; + + diff --git a/docs/cubie/a7z/app-dev/npu-dev/model-zoo/resnet50v2.md b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/resnet50v2.md new file mode 100644 index 000000000..274417b93 --- /dev/null +++ b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/resnet50v2.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 21 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-resnet50v2.mdx +--- + +# ResNet50 V2 + +import ResNet50V2 from '../../../../../common/ai/cubie/\_model-zoo-resnet50v2.mdx'; + + diff --git a/docs/cubie/a7z/app-dev/npu-dev/model-zoo/retinaface.md b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/retinaface.md new file mode 100644 index 000000000..ee3da187a --- /dev/null +++ b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/retinaface.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 15 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-retinaface.mdx +--- + +# RetinaFace + +import RetinaFace from '../../../../../common/ai/cubie/\_model-zoo-retinaface.mdx'; + + diff --git a/docs/cubie/a7z/app-dev/npu-dev/model-zoo/squeezenet-pytorch.md b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/squeezenet-pytorch.md new file mode 100644 index 000000000..690bafe69 --- /dev/null +++ b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/squeezenet-pytorch.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 23 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-squeezenet-pytorch.mdx +--- + +# SqueezeNet + +import SqueezeNet from '../../../../../common/ai/cubie/\_model-zoo-squeezenet-pytorch.mdx'; + + diff --git a/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolo11-pose.md b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolo11-pose.md new file mode 100644 index 000000000..f1255f584 --- /dev/null +++ b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolo11-pose.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 5 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo11-pose.mdx +--- + +# YOLO11 Pose + +import YOLO11Pose from '../../../../../common/ai/cubie/\_model-zoo-yolo11-pose.mdx'; + + diff --git a/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolo11-seg.md b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolo11-seg.md new file mode 100644 index 000000000..672afce80 --- /dev/null +++ b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolo11-seg.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 4 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo11-seg.mdx +--- + +# YOLO11 Seg + +import YOLO11Seg from '../../../../../common/ai/cubie/\_model-zoo-yolo11-seg.mdx'; + + diff --git a/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolo11.md b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolo11.md new file mode 100644 index 000000000..300d57295 --- /dev/null +++ b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolo11.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 3 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo11.mdx +--- + +# YOLO11 + +import YOLO11 from '../../../../../common/ai/cubie/\_model-zoo-yolo11.mdx'; + + diff --git a/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolo26.md b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolo26.md new file mode 100644 index 000000000..f9987edc1 --- /dev/null +++ b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolo26.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 12 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo26.mdx +--- + +# YOLO26 + +import YOLO26 from '../../../../../common/ai/cubie/\_model-zoo-yolo26.mdx'; + + diff --git a/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolov3-darknet.md b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolov3-darknet.md new file mode 100644 index 000000000..7d191f006 --- /dev/null +++ b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolov3-darknet.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 9 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov3-darknet.mdx +--- + +# YOLOv3 + +import YOLOv3 from '../../../../../common/ai/cubie/\_model-zoo-yolov3-darknet.mdx'; + + diff --git a/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolov5.md b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolov5.md new file mode 100644 index 000000000..cd33bd150 --- /dev/null +++ b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolov5.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 10 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov5.mdx +--- + +# YOLOv5 + +import YOLOv5 from '../../../../../common/ai/cubie/\_model-zoo-yolov5.mdx'; + + diff --git a/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolov8-pose.md b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolov8-pose.md new file mode 100644 index 000000000..d29446157 --- /dev/null +++ b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolov8-pose.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 8 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8-pose.mdx +--- + +# YOLOv8 Pose + +import YOLOv8Pose from '../../../../../common/ai/cubie/\_model-zoo-yolov8-pose.mdx'; + + diff --git a/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolov8-seg.md b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolov8-seg.md new file mode 100644 index 000000000..451da2b12 --- /dev/null +++ b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolov8-seg.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 7 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8-seg.mdx +--- + +# YOLOv8 Seg + +import YOLOv8Seg from '../../../../../common/ai/cubie/\_model-zoo-yolov8-seg.mdx'; + + diff --git a/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolov8.md b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolov8.md new file mode 100644 index 000000000..1eb3ebbd7 --- /dev/null +++ b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolov8.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 6 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8.mdx +--- + +# YOLOv8 + +import YOLOv8 from '../../../../../common/ai/cubie/\_model-zoo-yolov8.mdx'; + + diff --git a/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolox.md b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolox.md new file mode 100644 index 000000000..ed26d8739 --- /dev/null +++ b/docs/cubie/a7z/app-dev/npu-dev/model-zoo/yolox.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 14 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolox.mdx +--- + +# YOLOX + +import YOLOX from '../../../../../common/ai/cubie/\_model-zoo-yolox.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/ai/_cubie_acuity_env.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/_cubie_acuity_env.mdx index 0264eb661..0a6bfd113 100644 --- a/i18n/en/docusaurus-plugin-content-docs/current/common/ai/_cubie_acuity_env.mdx +++ b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/_cubie_acuity_env.mdx @@ -1,8 +1,8 @@ -ACUITY Toolkit is installed as a Docker image. Before installing ACUITY, install Docker on your x86 PC. +ACUITY Toolkit is installed as a Docker image. Before installing ACUITY, install Docker on your X86 PC. ## Install Docker -Install Docker on your x86 PC according to your platform. For more information, refer to [docker docs](https://docs.docker.com/engine/install/). +Install Docker on your X86 PC according to your platform. For more Docker installation instructions, refer to [docker docs](https://docs.docker.com/engine/install/) Ubuntu is used as an example below. @@ -29,10 +29,11 @@ sudo curl -fsSL https://download.docker.com/linux/ubuntu/gpg -o /etc/apt/keyring sudo chmod a+r /etc/apt/keyrings/docker.asc # Add the repository to Apt sources: + echo \ - "deb [arch=$(dpkg --print-architecture) signed-by=/etc/apt/keyrings/docker.asc] https://download.docker.com/linux/ubuntu \ - $(. /etc/os-release && echo "${UBUNTU_CODENAME:-$VERSION_CODENAME}") stable" | \ - sudo tee /etc/apt/sources.list.d/docker.list > /dev/null + "deb [arch=$(dpkg --print-architecture) signed-by=/etc/apt/keyrings/docker.asc] https://download.docker.com/linux/ubuntu \ + $(. /etc/os-release && echo "${UBUNTU_CODENAME:-$VERSION_CODENAME}") stable" | \ + sudo tee /etc/apt/sources.list.d/docker.list > /dev/null sudo apt-get update ``` @@ -51,111 +52,92 @@ sudo apt-get install docker-ce docker-ce-cli containerd.io docker-buildx-plugin ## Install ACUITY :::tip -Due to SDK version compatibility, you need to choose the appropriate package based on the NPU. +Due to SDK version compatibility, you need to choose the appropriate package based on your NPU. ::: - - - - - ### Get the ACUITY package - - Download the ACUITY Docker archive from [Allwinner Netdisk](https://netstorage.allwinnertech.com:5001/sharing/Mh23BhPHq) and extract it. - - ```bash - unzip docker_images_v2.0.x.zip - ``` - - ### Load the image - - - - ```bash - cd docker_images_v2.0.x - unzip ubuntu-npu_v2.0.10.tar.zip - sudo docker load -i ubuntu-npu_v2.0.10.tar - ``` - - - - After the Docker image is loaded, you can see it in `docker images` with the name `ubuntu‑npu:v2.0.10`. - - ### Create a Docker container - - - - ```bash - mkdir docker_data && cd docker_data - sudo docker run --ipc=host -itd -v ${PWD}:/workspace --name allwinner_v2.0.10 ubuntu-npu:v2.0.10 /bin/bash - ``` - - +### Get the ACUITY Package - After the container is created, you can see it in `docker ps -a` with the name `allwinner_v2.0.10`. +Download the ACUITY Docker archive from [Allwinner Netdisk](https://netstorage.allwinnertech.com:5001/sharing/Mh23BhPHq) and extract it. - ### Enter the Docker container - - Use `docker ps -a` to get the container ID of `allwinner_v2.0.10`. + + + - +```bash +unzip docker_images_v2.0.x.zip +``` - ```bash - sudo docker exec -it /bin/bash - ``` + + + + - +```bash +unzip docker_images_v1.8.x.zip +``` - + + + - +### Load the Image - ### Get the ACUITY package + + + - Download the ACUITY Docker archive from [Allwinner Netdisk](https://netstorage.allwinnertech.com:5001/sharing/N6TVlZQVZ) and extract it. - - ```bash - unzip docker_images_v1.8.x.zip - ``` - - ### Load the image +```bash +cd docker_images_v2.0.x unzip ubuntu-npu_v2.0.10.1.tar.zip sudo docker load -i ubuntu-npu_v2.0.10.1.tar +``` - + + + + - ```bash - cd docker_images_v1.8.x - unzip ubuntu-npu_v1.8.11.tar.zip - sudo docker load -i ubuntu-npu_v1.8.11.tar - ``` +```bash +cd docker_images_v1.8.x unzip ubuntu-npu_v1.8.11.tar.zip sudo docker load -i ubuntu-npu_v1.8.11.tar +``` - + + + - After the Docker image is loaded, you can see it in `docker images` with the name `ubuntu-npu:v1.8.11`. +After the Docker image is loaded, you can see it with `docker images`. The image name is `ubuntu‑npu:v2.0.10.1` (A733) or `ubuntu-npu:v1.8.11` (T527). - ### Create a Docker container +### Create a Docker Container - + + + - ```bash - mkdir docker_data && cd docker_data - sudo docker run --ipc=host -itd -v ${PWD}:/workspace --name allwinner_v1.8.11 ubuntu-npu:v1.8.11 /bin/bash - ``` +```bash +mkdir docker_data && cd docker_data sudo docker run --ipc=host -itd -v ${PWD}:/workspace --name allwinner_v2.0.10.1 ubuntu-npu:v2.0.10.1 /bin/bash +``` - + + + + - After the container is created, you can see it in `docker ps -a` with the name `allwinner_v1.8.11`. +```bash +mkdir docker_data && cd docker_data sudo docker run --ipc=host -itd -v ${PWD}:/workspace --name allwinner_v1.8.11 ubuntu-npu:v1.8.11 /bin/bash +``` - ### Enter the Docker container + + + - Use `docker ps -a` to get the container ID of `allwinner_v1.8.11`. +After the Docker container is created, you can see it with `docker ps -a`. The container name is `allwinner_v2.0.10.1` (A733) or `allwinner_v1.8.11` (T527). - +### Enter the Docker Container - ```bash - sudo docker exec -it /bin/bash - ``` +Use `docker ps -a` to find the container ID. - + - +```bash +sudo docker exec -it /bin/bash +``` - + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-densenet121-keras.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-densenet121-keras.mdx new file mode 100644 index 000000000..58d54aaff --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-densenet121-keras.mdx @@ -0,0 +1,312 @@ +This document describes how to run DenseNet121 on NPU. + +:::info +Refer to [Model Zoo Download](./model-zoo-download) for the example. +::: + +DenseNet121 Example Directory Structure: + +```bash +$ tree ./ +./ +├── class_post.cpp +├── class_pre.cpp +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ └── convert_model_env.sh +├── label.h +├── main.cpp +├── model +│ └── space_shuttle_224x224.jpg +└── README.md +``` + +## Model Conversion + +### Download Model + +Click to download [densenet121_batch1_224x224.h5](http://netstorage.allwinnertech.com:5000/sharing/d7pP1omn5). + +Then move the model to the convert_model/ directory. + +### Create Symlink for Conversion Script + + + +```bash +./convert_model_env.sh +``` + + + +### Model Import/Quantization/Conversion + +You need to enter the container development environment first. Refer to the [Create Container](./model-zoo-download#创建并启动容器) section in Model Zoo Download. + +:::info +Different platforms use corresponding Docker images: + +- A733: ubuntu-npu:v2.0.10.1 +- T527: ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +After entering the container, navigate to the corresponding directory and run the script. + + + +```bash +cd /workspace/examples/densenet121_keras/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh densenet121_batch1_224x224 +./pegasus_quantize.sh densenet121_batch1_224x224 uint8 10 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh densenet121_batch1_224x224 uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh densenet121_batch1_224x224 uint8 t527 +``` + + + + + + +The exported model files are stored in the ../model directory. + +### Compile Example + +Now you can compile the example. **First exit the container**, then execute the following command to compile the example. + +First, you need to configure third-party libraries and cross-compilation toolchain. + +:::info +You can skip this step if you have already configured third-party libraries and cross-compilation toolchain in other examples. +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +You need to manually download via [this link](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4) first, then place it in 0-toolchains/ before executing the following command: + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/densenet121_keras/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## Model Deployment + +After compilation, the example will be installed in the install directory. You can use scp to transfer it to the board. + +### Configure NPU Driver + +:::info +You can skip this step if you have already configured NPU driver in other examples. +::: + +Transfer the driver library to the board's lib directory via scp. + +- A733 corresponds to the common/lib_linux_aarch64/A733 directory +- T527 corresponds to the common/lib_linux_aarch64/T527 directory + +Then execute the following command to export to environment variables. + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### Run Example + +After configuring the driver, you can run the example. + +:::tip +For T527 platform, you need to first enable NPU by referring to the A5E's "Enable NPU on Board" documentation, then use the following command to grant the current user permission to use /dev/vipcore. +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd densenet121_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./densenet121_demo_a733 +./densenet121_demo_a733 -nb model/densenet121_batch1_224x224_uint8_a733.nb -i model/space_shuttle_224x224.jpg +``` + + + +The running result is as follows: + +```bash +$ ./densenet121_demo_a733 -nb model/densenet121_batch1_224x224_uint8_a733.nb -i model/space_shuttle_224x224.jpg +model_file=model/densenet121_batch1_224x224_uint8_a733.nb, input=model/space_shuttle_224x224.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 224 224 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 1000 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/densenet121_batch1_224x224_uint8_a733.nb, size: 6397400. +create network 0: 10448 us. +prepare network: 3582 us. +network: 0, loop count: 1 +run time for this network 0: 8278 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 812, prob: 0.973633, label: space shuttle +class id: 569, prob: 0.005028, label: garbage truck, dustcart +class id: 403, prob: 0.003510, label: aircraft carrier, carrier, flattop, attack aircraft carrier +class id: 408, prob: 0.002764, label: amphibian, amphibious vehicle +class id: 895, prob: 0.002176, label: warplane, military plane +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :---------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner A733 | Vivante VIP9000 | densenet121 | 224×224 | 10.4 ms | 3.6 ms | 8.3 ms | | 22.3 ms | 44.8 FPS | + + + + + + + +```bash +cd densenet121_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./densenet121_demo_t527 +./densenet121_demo_t527 -nb model/densenet121_batch1_224x224_uint8_t527.nb -i model/space_shuttle_224x224.jpg +``` + + + +The running result is as follows: + +```bash +$ ./densenet121_demo_t527 -nb model/densenet121_batch1_224x224_uint8_t527.nb -i model/space_shuttle_224x224.jpg +model_file=model/densenet121_batch1_224x224_uint8_t527.nb, input=model/space_shuttle_224x224.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 224 224 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 1000 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/densenet121_batch1_224x224_uint8_t527.nb, size: 5857536. +create network 0: 12870 us. +prepare network: 2366 us. +network: 0, loop count: 1 +run time for this network 0: 10573 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 812, prob: 0.961914, label: space shuttle +class id: 569, prob: 0.008011, label: garbage truck, dustcart +class id: 408, prob: 0.004963, label: amphibian, amphibious vehicle +class id: 403, prob: 0.003466, label: aircraft carrier, carrier, flattop, attack aircraft carrier +class id: 895, prob: 0.003466, label: warplane, military plane +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :---------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner T527 | Vivante VIP9000 | densenet121 | 224×224 | 12.9 ms | 2.4 ms | 10.6 ms | | 25.9 ms | 38.6 FPS | + + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-download.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-download.mdx new file mode 100644 index 000000000..1c87b6987 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-download.mdx @@ -0,0 +1,112 @@ +You can obtain the Model Zoo from the [Allwinner Customer Service Platform](https://open.allwinnertech.com/). + +Click Register Now: + +![Allwinner Customer Service Platform Login](/img/cubie/allwinnertech-login.webp) + +After successfully registering and logging in, you will enter the dashboard interface. Click Resource Download: + +![Click Resource Download](/img/cubie/allwinnertech-model-zoo-download-guide1.webp) + +Click AI Development SDK under Tool Query on the left sidebar, then find the v0.9.0 version of Model Zoo and download it to your local machine. + +![Download Model Zoo](/img/cubie/allwinnertech-model-zoo-download-guide2.webp) + +Or you can download via dl.radxa: + + + +```bash +wget https://dl.radxa.com/cubie/allwinner-model-zoo.tar.gz +``` + + + +Before running examples in the Model Zoo, you need to download the container development package. You can find the corresponding version for download on the interface above, or refer to our [guidance document](../cubie-acuity-env). + +:::info +Note that T527 and A733 use different versions of the container development package. +::: + +After downloading both the container development package and Model Zoo, you can refer to the steps below: + +## Extract Model Zoo + + + +```bash +tar -xvf 1768567762439_awnpu_model_zoo-v0.9.0-20260116-83a67d4b.tar.gz +cd awnpu_model_zoo-v0.9.0-20260116-83a67d4b/ +``` + + + +## Create and Start Container + + + + + + +```bash +sudo docker run --ipc=host -d -v ${PWD}:/workspace --name model-zoo ubuntu-npu:v2.0.10.1 tail -f /dev/null +``` + + + + + + + + + +```bash +sudo docker run --ipc=host -d -v ${PWD}:/workspace --name model-zoo ubuntu-npu:v1.8.11 tail -f /dev/null +``` + + + + + + +When the container is created, you can see this container in `docker ps` with the name `model-zoo`. + +## Enter Container + + + +```bash +sudo docker exec -it model-zoo /bin/bash +``` + + + +After completion, the Model Zoo directory has been mounted to the `/workspace` directory in the container. Execute inside the container: + + + +```bash +cd /workspace +ls -al +``` + + + +You can see the Model Zoo directory structure: + +```bash +/workspace# ls -al +total 52 +drwxr-xr-x 10 1000 1000 4096 Apr 2 19:09 . +drwxr-xr-x 9 root root 4096 Apr 2 18:48 .. +drwxr-xr-x 10 1000 1000 4096 Jan 16 20:44 0-toolchains # Cross-compilation toolchain +drwxr-xr-x 3 1000 1000 4096 Jan 16 20:44 3rdparty # Third-party toolchain +-rwxr-xr-x 1 1000 1000 4976 Jan 16 20:44 README.md # README document +drwxr-xr-x 2 1000 1000 4096 Apr 3 12:06 cmake_toolchain # Compilation toolchain configuration +drwxr-xr-x 3 1000 1000 4096 Jan 16 20:44 common # Common library +drwxr-xr-x 3 1000 1000 4096 Jan 16 20:44 docs # Documentation +drwxr-xr-x 26 1000 1000 4096 Jan 16 20:44 examples # Example directory +drwxr-xr-x 6 1000 1000 4096 Jan 16 20:44 functions # Special features +drwxr-xr-x 2 1000 1000 4096 Jan 16 20:44 scripts_model_convert # Conversion scripts +-rwxr-xr-x 1 1000 1000 6 Jan 16 20:44 version # Version number +``` diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-lenet-caffe.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-lenet-caffe.mdx new file mode 100644 index 000000000..4c1b467a1 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-lenet-caffe.mdx @@ -0,0 +1,305 @@ +This document describes how to run LeNet on NPU. + +:::info +Refer to [Model Zoo Download](./model-zoo-download) for the example. +::: + +LeNet Example Directory Structure: + +```bash +$ tree ./ +./ +├── class_pre.cpp +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ └── convert_model_env.sh +├── main.cpp +├── model +│ ├── 3.jpg +│ ├── 4.jpg +│ └── 5.jpg +└── README.md +``` + +## Model Conversion + +### Download Model + +Click to download [lenet.caffemodel](http://netstorage.allwinnertech.com:5000/sharing/JG3iqvFvH). + +Click to download [lenet.prototxt](http://netstorage.allwinnertech.com:5000/sharing/Wf4J4DsTR). + +Then move the model to the convert_model/ directory. + +### Create Symlink for Conversion Script + + + +```bash +cd convert_model/ +./convert_model_env.sh +``` + + + +### Model Import/Quantization/Conversion + +You need to enter the container development environment first. Refer to the [Create Container](./model-zoo-download#创建并启动容器) section in Model Zoo Download. + +:::info +Different platforms use corresponding Docker images: + +- A733: ubuntu-npu:v2.0.10.1 +- T527: ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +After entering the container, navigate to the corresponding directory and run the script. + + + +```bash +cd /workspace/examples/lenet_caffe/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh lenet +./pegasus_quantize.sh lenet uint8 12 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh lenet uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh lenet uint8 t527 +``` + + + + + + +The exported model files are stored in the ../model directory. + +### Compile Example + +Now you can compile the example. **First exit the container**, then execute the following command to compile the example. + +First, you need to configure third-party libraries and cross-compilation toolchain. + +:::info +You can skip this step if you have already configured third-party libraries and cross-compilation toolchain in other examples. +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +You need to manually download via [this link](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4) first, then place it in 0-toolchains/ before executing the following command: + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/lenet_caffe/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## Model Deployment + +After compilation, the example will be installed in the install directory. You can use scp to transfer it to the board. + +### Configure NPU Driver + +:::info +You can skip this step if you have already configured NPU driver in other examples. +::: + +Transfer the driver library to the board's lib directory via scp. + +- A733 corresponds to the common/lib_linux_aarch64/A733 directory +- T527 corresponds to the common/lib_linux_aarch64/T527 directory + +Then execute the following command to export to environment variables. + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### Run Example + +After configuring the driver, you can run the example. + +:::tip +For T527 platform, you need to first enable NPU by referring to the A5E's "Enable NPU on Board" documentation, then use the following command to grant the current user permission to use /dev/vipcore. +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd lenet_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./lenet_demo_a733 +./lenet_demo_a733 -nb model/lenet_uint8_a733.nb -i model/3.jpg +``` + + + +The running result is as follows: + +```bash +$ ./lenet_demo_a733 -nb model/lenet_uint8_a733.nb -i model/3.jpg +model_file=model/lenet_uint8_a733.nb, input=model/3.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 28 28 1 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 10 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/lenet_uint8_a733.nb, size: 407776. +create network 0: 1486 us. +prepare network: 176 us. +network: 0, loop count: 1 +run time for this network 0: 281 us. +Image: model/3.jpg, Predicted digit: 3, Probability: 1.000000 +Class probabilities: 0 : 0.0000 1 : 0.0000 2 : 0.0000 3 : 1.0000 4 : 0.0000 5 : 0.0000 6 : 0.0000 7 : 0.0000 8 : 0.0000 9 : 0.0000 +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :---- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner A733 | Vivante VIP9000 | lenet | 28×28 | 1.5 ms | 0.2 ms | 0.3 ms | | 2.0 ms | 500.0 FPS | + + + + + + + +```bash +cd lenet_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./lenet_demo_t527 +./lenet_demo_t527 -nb model/lenet_uint8_t527.nb -i model/3.jpg +``` + + + +The running result is as follows: + +```bash +$ ./lenet_demo_t527 -nb model/lenet_uint8_t527.nb -i model/3.jpg +model_file=model/lenet_uint8_t527.nb, input=model/3.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 28 28 1 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 10 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/lenet_uint8_t527.nb, size: 403520. +create network 0: 744 us. +prepare network: 123 us. +network: 0, loop count: 1 +run time for this network 0: 219 us. +Image: model/3.jpg, Predicted digit: 1, Probability: 0.607422 +Class probabilities: 0 : 0.0523 1 : 0.6074 2 : 0.0154 3 : 0.0132 4 : 0.0243 5 : 0.0610 6 : 0.0449 7 : 0.1530 8 : 0.0045 9 : 0.0243 +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :---- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner T527 | Vivante VIP9000 | lenet | 28×28 | 0.7 ms | 0.1 ms | 0.2 ms | | 1.0 ms | 1000.0 FPS | + + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-mobilenetv1-tensorflow.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-mobilenetv1-tensorflow.mdx new file mode 100644 index 000000000..1ff09da09 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-mobilenetv1-tensorflow.mdx @@ -0,0 +1,342 @@ +This document describes how to run MobileNetV1 on NPU. + +:::info +Refer to [Model Zoo Download](./model-zoo-download) for the example. +::: + +MobileNetV1 Example Directory Structure: + +```bash +$ tree ./ +./ +├── class_post.cpp +├── class_pre.cpp +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ └── inputs_outputs.txt +├── label.h +├── main.cpp +├── model +│ └── space_shuttle_224x224.jpg +└── README.md +``` + +## Model Conversion + +### Download Model + +Click to download [mobilenet_v1_1.0_224_frozen.pb](http://netstorage.allwinnertech.com:5000/sharing/YtgpJ76vp). + +Then move the model to the convert_model/ directory. + +### Create Symlink for Conversion Script + + + +```bash +cd convert_model/ +./convert_model_env.sh +``` + + + +### Model Import/Quantization/Conversion + +You need to enter the container development environment first. Refer to the [Create Container](./model-zoo-download#创建并启动容器) section in Model Zoo Download. + +:::info +Different platforms use corresponding Docker images: + +- A733: ubuntu-npu:v2.0.10.1 +- T527: ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +After entering the container, navigate to the corresponding directory and run the script. + + + +```bash +cd /workspace/examples/mobilenetv1_tensorflow/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh mobilenet_v1_1.0_224_frozen +./pegasus_quantize.sh mobilenet_v1_1.0_224_frozen uint8 10 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh mobilenet_v1_1.0_224_frozen uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh mobilenet_v1_1.0_224_frozen uint8 t527 +``` + + + + + + +The exported model files are stored in the ../model directory. + +### Compile Example + +Now you can compile the example. **First exit the container**, then execute the following command to compile the example. + +First, you need to configure third-party libraries and cross-compilation toolchain. + +:::info +You can skip this step if you have already configured third-party libraries and cross-compilation toolchain in other examples. +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +You need to manually download via [this link](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4) first, then place it in 0-toolchains/ before executing the following command: + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/mobilenetv1_tensorflow/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## Model Deployment + +After compilation, the example will be installed in the install directory. You can use scp to transfer it to the board. + +### Configure NPU Driver + +:::info +You can skip this step if you have already configured NPU driver in other examples. +::: + +Transfer the driver library to the board's lib directory via scp. + +- A733 corresponds to the common/lib_linux_aarch64/A733 directory +- T527 corresponds to the common/lib_linux_aarch64/T527 directory + +Then execute the following command to export to environment variables. + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### Run Example + +After configuring the driver, you can run the example. + +:::tip +For T527 platform, you need to first enable NPU by referring to the A5E's "Enable NPU on Board" documentation, then use the following command to grant the current user permission to use /dev/vipcore. +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd mobilenetv1_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./mobilenetv1_demo_a733 +./mobilenetv1_demo_a733 -nb model/mobilenet_v1_1.0_224_frozen_uint8_a733.nb -i model/space_shuttle_224x224.jpg +``` + + + +The running result is as follows: + +```bash +$ ./mobilenetv1_demo_a733 -nb model/mobilenet_v1_1.0_224_frozen_uint8_a733.nb -i model/space_shuttle_224x224.jpg +model_file=model/mobilenet_v1_1.0_224_frozen_uint8_a733.nb, input=model/space_shuttle_224x224.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 224 224 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 1001 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/mobilenet_v1_1.0_224_frozen_uint8_a733.nb, size: 3182272. +create network 0: 5661 us. +prepare network: 1085 us. +network: 0, loop count: 1 +run time for this network 0: 1845 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 813, prob: 0.993652, label: space shuttle +class id: 868, prob: 0.001479, label: trailer truck, tractor trailer, trucking rig, rig, articulated lorry, semi +class id: 405, prob: 0.000976, label: airliner +class id: 406, prob: 0.000561, label: airship, dirigible +class id: 628, prob: 0.000370, label: limousine, limo +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :---------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner A733 | Vivante VIP9000 | mobilenetv1 | 224×224 | 5.7 ms | 1.1 ms | 1.8 ms | | 8.6 ms | 116.3 FPS | + + + + + + + +```bash +cd mobilenetv1_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./mobilenetv1_demo_t527 +./mobilenetv1_demo_t527 -nb model/mobilenet_v1_1.0_224_frozen_uint8_t527.nb -i model/space_shuttle_224x224.jpg +``` + + + +The running result is as follows: + +```bash +$ ./mobilenetv1_demo_t527 -nb model/mobilenet_v1_1.0_224_frozen_uint8_t527.nb -i model/space_shuttle_224x224.jpg +model_file=model/mobilenet_v1_1.0_224_frozen_uint8_t527.nb, input=model/space_shuttle_224x224.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 224 224 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 1001 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/mobilenet_v1_1.0_224_frozen_uint8_t527.nb, size: 3097728. +create network 0: 5639 us. +prepare network: 980 us. +network: 0, loop count: 1 +run time for this network 0: 3809 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 813, prob: 0.993652, label: space shuttle +class id: 868, prob: 0.001479, label: trailer truck, tractor trailer, trucking rig, rig, articulated lorry, semi +class id: 405, prob: 0.000976, label: airliner +class id: 406, prob: 0.000561, label: airship, dirigible +class id: 628, prob: 0.000370, label: limousine, limo +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :---------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner T527 | Vivante VIP9000 | mobilenetv1 | 224×224 | 5.6 ms | 1.0 ms | 3.8 ms | | 10.4 ms | 96.2 FPS | + + + + +The running result is as follows: + +```bash +$ ./mobilenetv1_demo_a733 -nb model/mobilenet_v1_1.0_224_frozen_uint8_a733.nb -i model/space_shuttle_224x224.jpg +model_file=model/mobilenet_v1_1.0_224_frozen_uint8_a733.nb, input=model/space_shuttle_224x224.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 224 224 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 1001 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/mobilenet_v1_1.0_224_frozen_uint8_a733.nb, size: 3182272. +create network 0: 5661 us. +prepare network: 1085 us. +network: 0, loop count: 1 +run time for this network 0: 1845 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 813, prob: 0.995117, label: space shuttle +class id: 868, prob: 0.001483, label: trailer truck, tractor trailer, trucking rig, rig, articulated lorry, semi +class id: 405, prob: 0.000979, label: airliner +class id: 406, prob: 0.000563, label: airship, dirigible +class id: 628, prob: 0.000371, label: limousine, limo +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :---------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner A733 | Vivante VIP9000 | mobilenetv1 | 224×224 | 5.7 ms | 1.1 ms | 1.8 ms | | 8.6 ms | 116.3 FPS | diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-mobilenetv2.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-mobilenetv2.mdx new file mode 100644 index 000000000..baa31da92 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-mobilenetv2.mdx @@ -0,0 +1,341 @@ +This document describes how to run MobileNetV2 on NPU. + +:::info +Refer to [Model Zoo Download](./model-zoo-download) for the example. +::: + +MobileNetV2 Example Directory Structure: + +```bash +$ tree ./ +./ +├── class_post.cpp +├── class_pre.cpp +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ └── mobilenetv2-12.onnx +├── label.h +├── main.cpp +├── model +│ ├── 1.jpg +│ └── mobilenetv2-12_pcq_t527.nb +└── README.md +``` + +## Model Conversion + +### Download Model + +The model already exists in the directory, no download needed. + +### Create Symlink for Conversion Script + + + +```bash +cd convert_model/ +./convert_model_env.sh +``` + + + +### Model Import/Quantization/Conversion + +You need to enter the container development environment first. Refer to the [Create Container](./model-zoo-download#创建并启动容器) section in Model Zoo Download. + +:::info +Different platforms use corresponding Docker images: + +- A733: ubuntu-npu:v2.0.10.1 +- T527: ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +After entering the container, navigate to the corresponding directory and run the script. + + + +```bash +cd /workspace/examples/mobilenetv2/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh mobilenetv2-12 +./pegasus_quantize.sh mobilenetv2-12 pcq 10 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh mobilenetv2-12 pcq a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh mobilenetv2-12 pcq t527 +``` + + + + + + +The exported model files are stored in the ../model directory. + +### Compile Example + +Now you can compile the example. **First exit the container**, then execute the following command to compile the example. + +First, you need to configure third-party libraries and cross-compilation toolchain. + +:::info +You can skip this step if you have already configured third-party libraries and cross-compilation toolchain in other examples. +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +You need to manually download via [this link](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4) first, then place it in 0-toolchains/ before executing the following command: + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/mobilenetv2/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## Model Deployment + +After compilation, the example will be installed in the install directory. You can use scp to transfer it to the board. + +### Configure NPU Driver + +:::info +You can skip this step if you have already configured NPU driver in other examples. +::: + +Transfer the driver library to the board's lib directory via scp. + +- A733 corresponds to the common/lib_linux_aarch64/A733 directory +- T527 corresponds to the common/lib_linux_aarch64/T527 directory + +Then execute the following command to export to environment variables. + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### Run Example + +After configuring the driver, you can run the example. + +:::tip +For T527 platform, you need to first enable NPU by referring to the A5E's "Enable NPU on Board" documentation, then use the following command to grant the current user permission to use /dev/vipcore. +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd mobilenetv2_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./mobilenetv2_demo_a733 +./mobilenetv2_demo_a733 -nb model/mobilenetv2-12_pcq_a733.nb -i model/1.jpg +``` + + + +The running result is as follows: + +```bash +$ ./mobilenetv2_demo_a733 -nb model/mobilenetv2-12_pcq_a733.nb -i model/1.jpg +model_file=model/mobilenetv2-12_pcq_a733.nb, input=model/1.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 224 224 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 1000 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/mobilenetv2-12_pcq_a733.nb, size: 3617080. +create network 0: 6194 us. +prepare network: 1134 us. +network: 0, loop count: 1 +run time for this network 0: 1792 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 281, prob: 15.055555, label: tabby, tabby cat +class id: 285, prob: 13.879630, label: Egyptian cat +class id: 287, prob: 11.990741, label: lynx, catamount +class id: 282, prob: 8.842592, label: tiger cat +class id: 631, prob: 4.837963, label: grey fox +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :---------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner A733 | Vivante VIP9000 | mobilenetv2 | 224×224 | 6.2 ms | 1.1 ms | 1.8 ms | | 9.1 ms | 109.9 FPS | + + + + + + + +```bash +cd mobilenetv2_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./mobilenetv2_demo_t527 +./mobilenetv2_demo_t527 -nb model/mobilenetv2-12_pcq_t527.nb -i model/1.jpg +``` + + + +The running result is as follows: + +```bash +$ ./mobilenetv2_demo_t527 -nb model/mobilenetv2-12_pcq_t527.nb -i model/1.jpg +model_file=model/mobilenetv2-12_pcq_t527.nb, input=model/1.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 224 224 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 1000 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/mobilenetv2-12_pcq_t527.nb, size: 3515008. +create network 0: 6479 us. +prepare network: 1872 us. +network: 0, loop count: 1 +run time for this network 0: 3127 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 281, prob: 14.506104, label: tabby, tabby cat +class id: 282, prob: 13.106445, label: tiger cat +class id: 285, prob: 12.215820, label: Egyptian cat +class id: 287, prob: 8.143799, label: lynx, catamount +class id: 478, prob: 7.889282, label: carton +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :---------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner T527 | Vivante VIP9000 | mobilenetv2 | 224×224 | 6.5 ms | 1.9 ms | 3.1 ms | | 11.5 ms | 87.0 FPS | + + + + +The running result is as follows: + +```bash +$ ./mobilenetv2_demo_a733 -nb model/mobilenetv2-12_pcq_a733.nb -i model/1.jpg +model_file=model/mobilenetv2-12_pcq_a733.nb, input=model/1.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 224 224 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 1000 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/mobilenetv2-12_pcq_a733.nb, size: 3617080. +create network 0: 6194 us. +prepare network: 1502 us. +network: 0, loop count: 1 +run time for this network 0: 2028 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 281, prob: 14.801300, label: tabby, tabby cat +class id: 282, prob: 13.230242, label: tiger cat +class id: 285, prob: 12.491362, label: Egyptian cat +class id: 287, prob: 8.243347, label: lynx, catamount +class id: 478, prob: 8.116148, label: carton +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :---------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner A733 | Vivante VIP9000 | mobilenetv2 | 224×224 | 6.2 ms | 1.5 ms | 2.0 ms | | 9.7 ms | 103.1 FPS | diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-ppseg.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-ppseg.mdx new file mode 100644 index 000000000..5f9758d93 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-ppseg.mdx @@ -0,0 +1,311 @@ +This document describes how to run PPSeg on NPU. + +:::info +Refer to [Model Zoo Download](./model-zoo-download) for the example. +::: + +PPSeg Example Directory Structure: + +```bash +$ tree ./ +./ +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ ├── model.pdparams +│ └── pp_liteseg_cityscapes.txt +├── figures +│ └── out_ppseg.png +├── main.cpp +├── model +│ └── munster_000022_000019_leftImg8bit.png +├── model_config.h +├── ppseg_post.cpp +├── ppseg_pre.cpp +└── README.md +``` + +## Model Conversion + +### Download Model + +Click to download [pp_liteseg_cityscapes](http://netstorage.allwinnertech.com:5000/sharing/H2GNL3AuC). + +Then move the model to the convert_model/ directory. + +### Create Symlink for Conversion Script + + + +```bash +cd convert_model/ +./convert_model_env.sh +``` + + + +### Model Import/Quantization/Conversion + +You need to enter the container development environment first. Refer to the [Create Container](./model-zoo-download#创建并启动容器) section in Model Zoo Download. + +:::info +Different platforms use corresponding Docker images: + +- A733: ubuntu-npu:v2.0.10.1 +- T527: ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +After entering the container, navigate to the corresponding directory and run the script. + + + +```bash +cd /workspace/examples/ppseg/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh pp_liteseg_cityscapes +./pegasus_quantize.sh pp_liteseg_cityscapes pcq 10 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh pp_liteseg_cityscapes pcq a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh pp_liteseg_cityscapes pcq t527 +``` + + + + + + +The exported model files are stored in the ../model directory. + +### Compile Example + +Now you can compile the example. **First exit the container**, then execute the following command to compile the example. + +First, you need to configure third-party libraries and cross-compilation toolchain. + +:::info +You can skip this step if you have already configured third-party libraries and cross-compilation toolchain in other examples. +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +You need to manually download via [this link](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4) first, then place it in 0-toolchains/ before executing the following command: + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/ppseg/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## Model Deployment + +After compilation, the example will be installed in the install directory. You can use scp to transfer it to the board. + +### Configure NPU Driver + +:::info +You can skip this step if you have already configured NPU driver in other examples. +::: + +Transfer the driver library to the board's lib directory via scp. + +- A733 corresponds to the common/lib_linux_aarch64/A733 directory +- T527 corresponds to the common/lib_linux_aarch64/T527 directory + +Then execute the following command to export to environment variables. + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### Run Example + +After configuring the driver, you can run the example. + +:::tip +For T527 platform, you need to first enable NPU by referring to the A5E's "Enable NPU on Board" documentation, then use the following command to grant the current user permission to use /dev/vipcore. +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd ppseg_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./ppseg_demo_a733 +./ppseg_demo_a733 -nb model/pp_liteseg_cityscapes_pcq_a733.nb -i model/munster_000022_000019_leftImg8bit.png +``` + + + +The running result is as follows: + +```bash +$ ./ppseg_demo_a733 -nb model/pp_liteseg_cityscapes_pcq_a733.nb -i model/munster_000022_000019_leftImg8bit.png +model_file=model/pp_liteseg_cityscapes_pcq_a733.nb, input=model/munster_000022_000019_leftImg8bit.png, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 512 512 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 512 512 19 1, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/pp_liteseg_cityscapes_pcq_a733.nb, size: 10615496. +create network 0: 30474 us. +prepare network: 1578 us. +buffer ptr: 0x190c0300, buffer size: 786432 +network: 0, loop count: 1 +run time for this network 0: 81842 us. +output 0, ptr 0xffffa2a2b040, size 4980736. +post process time : 146 ms +ppseg_postprocess finished. +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :-------------------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner A733 | Vivante VIP9000 | pp_liteseg_cityscapes | 512×512 | 30.5 ms | 1.6 ms | 81.8 ms | 146.0 ms | 260 ms | 3.8 FPS | + + + + + + + +```bash +cd ppseg_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./ppseg_demo_t527 +./ppseg_demo_t527 -nb model/pp_liteseg_cityscapes_pcq_t527.nb -i model/munster_000022_000019_leftImg8bit.png +``` + + + +The running result is as follows: + +```bash +$ ./ppseg_demo_t527 -nb model/pp_liteseg_cityscapes_pcq_t527.nb -i model/munster_000022_000019_leftImg8bit.png +model_file=model/pp_liteseg_cityscapes_pcq_t527.nb, input=model/munster_000022_000019_leftImg8bit.png, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 512 512 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 512 512 19 1, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/pp_liteseg_cityscapes_pcq_t527.nb, size: 10894912. +create network 0: 39396 us. +prepare network: 6386 us. +buffer ptr: 0x2fa182c0, buffer size: 786432 +network: 0, loop count: 1 +run time for this network 0: 122524 us. +output 0, ptr 0xffff8f7d9040, size 4980736. +post process time : 346 ms +ppseg_postprocess finished. +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :-------------------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner T527 | Vivante VIP9000 | pp_liteseg_cityscapes | 512×512 | 39.4 ms | 6.4 ms | 122.5 ms | 346.0 ms | 514.3 ms | 1.9 FPS | + + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-resnet50-tflite.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-resnet50-tflite.mdx new file mode 100644 index 000000000..335958bbd --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-resnet50-tflite.mdx @@ -0,0 +1,345 @@ +This document describes how to run ResNet50 TFLite on NPU. + +:::info +Refer to [Model Zoo Download](./model-zoo-download) for the example. +::: + +ResNet50 TFLite Example Directory Structure: + +```bash +$ tree ./ +./ +├── class_post.cpp +├── class_pre.cpp +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ └── convert_model_env.sh +├── label.h +├── main.cpp +├── model +│ └── space_shuttle_224x224.jpg +└── README.md +``` + +## Model Conversion + +### Download Model + + + +```bash +cd convert_model/ +wget https://huggingface.co/qualcomm/ResNet50/resolve/18ab0a0ae3c14bc3ee7006c017f12802ab89cdf2/ResNet50.tflite +``` + + + +### Create Symlink for Conversion Script + + + +```bash +./convert_model_env.sh +``` + + + +### Model Import/Quantization/Conversion + +You need to enter the container development environment first. Refer to the [Create Container](./model-zoo-download#创建并启动容器) section in Model Zoo Download. + +:::info +Different platforms use corresponding Docker images: + +- A733: ubuntu-npu:v2.0.10.1 +- T527: ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +After entering the container, navigate to the corresponding directory and run the script. + + + +```bash +cd /workspace/examples/resnet50/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh ResNet50 +./pegasus_quantize.sh ResNet50 uint8 10 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh ResNet50 uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh ResNet50 uint8 t527 +``` + + + + + + +The exported model files are stored in the ../model directory. + +### Compile Example + +Now you can compile the example. **First exit the container**, then execute the following command to compile the example. + +First, you need to configure third-party libraries and cross-compilation toolchain. + +:::info +You can skip this step if you have already configured third-party libraries and cross-compilation toolchain in other examples. +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +You need to manually download via [this link](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4) first, then place it in 0-toolchains/ before executing the following command: + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/resnet50/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## Model Deployment + +After compilation, the example will be installed in the install directory. You can use scp to transfer it to the board. + +### Configure NPU Driver + +:::info +You can skip this step if you have already configured NPU driver in other examples. +::: + +Transfer the driver library to the board's lib directory via scp. + +- A733 corresponds to the common/lib_linux_aarch64/A733 directory +- T527 corresponds to the common/lib_linux_aarch64/T527 directory + +Then execute the following command to export to environment variables. + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### Run Example + +After configuring the driver, you can run the example. + +:::tip +For T527 platform, you need to first enable NPU by referring to the A5E's "Enable NPU on Board" documentation, then use the following command to grant the current user permission to use /dev/vipcore. +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd resnet50_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./resnet50_demo_a733 +./resnet50_demo_a733 -nb model/ResNet50_uint8_a733.nb -i model/space_shuttle_224x224.jpg +``` + + + +The running result is as follows: + +```bash +$ ./resnet50_demo_a733 -nb model/ResNet50_uint8_a733.nb -i model/space_shuttle_224x224.jpg +model_file=model/ResNet50_uint8_a733.nb, input=model/space_shuttle_224x224.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 224 224 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 1000 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/ResNet50_uint8_a733.nb, size: 16737832. +create network 0: 15664 us. +prepare network: 1734 us. +network: 0, loop count: 1 +run time for this network 0: 8913 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 812, prob: 26.991699, label: space shuttle +class id: 404, prob: 12.526367, label: airliner +class id: 867, prob: 11.482666, label: trailer truck, tractor trailer, trucking rig, rig, articulated lorry, semi +class id: 833, prob: 11.184326, label: submarine, pigboat, sub, U-boat +class id: 675, prob: 9.693115, label: moving van +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner A733 | Vivante VIP9000 | resnet50 | 224×224 | 15.7 ms | 1.7 ms | 8.9 ms | | 26.3 ms | 38.0 FPS | + + + + + + + +```bash +cd resnet50_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./resnet50_demo_t527 +./resnet50_demo_t527 -nb model/ResNet50_uint8_t527.nb -i model/space_shuttle_224x224.jpg +``` + + + +The running result is as follows: + +```bash +$ ./resnet50_demo_t527 -nb model/ResNet50_uint8_t527.nb -i model/space_shuttle_224x224.jpg +model_file=model/ResNet50_uint8_t527.nb, input=model/space_shuttle_224x224.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 224 224 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 1000 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/ResNet50_uint8_t527.nb, size: 16724480. +create network 0: 21618 us. +prepare network: 2776 us. +network: 0, loop count: 1 +run time for this network 0: 14320 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 812, prob: 26.991699, label: space shuttle +class id: 404, prob: 12.526367, label: airliner +class id: 867, prob: 11.482666, label: trailer truck, tractor trailer, trucking rig, rig, articulated lorry, semi +class id: 833, prob: 11.184326, label: submarine, pigboat, sub, U-boat +class id: 675, prob: 9.693115, label: moving van +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner T527 | Vivante VIP9000 | resnet50 | 224×224 | 21.6 ms | 2.8 ms | 14.3 ms | | 38.7 ms | 25.8 FPS | + + + + +The running result is as follows: + +```bash +$ ./resnet50_demo_a733 -nb model/ResNet50_uint8_a733.nb -i model/space_shuttle_224x224.jpg +model_file=model/ResNet50_uint8_a733.nb, input=model/space_shuttle_224x224.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 224 224 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 1000 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/ResNet50_uint8_a733.nb, size: 16737832. +create network 0: 19026 us. +prepare network: 646 us. +network: 0, loop count: 1 +run time for this network 0: 7255 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 812, prob: 26.843445, label: space shuttle +class id: 404, prob: 12.652412, label: airliner +class id: 867, prob: 11.439104, label: trailer truck, tractor trailer, trucking rig, rig, articulated lorry, semi +class id: 833, prob: 10.994865, label: submarine, pigboat, sub, U-boat +class id: 675, prob: 9.653847, label: moving van +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner A733 | Vivante VIP9000 | resnet50 | 224×224 | 19.0 ms | 0.6 ms | 7.3 ms | | 26.9 ms | 37.2 FPS | diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-resnet50v2.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-resnet50v2.mdx new file mode 100644 index 000000000..5f57c6634 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-resnet50v2.mdx @@ -0,0 +1,341 @@ +This document describes how to run ResNet50 V2 on NPU. + +:::info +Refer to [Model Zoo Download](./model-zoo-download) for the example. +::: + +ResNet50 V2 Example Directory Structure: + +```bash +$ tree ./ +./ +├── class_post.cpp +├── class_pre.cpp +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ └── convert_model_env.sh +├── label.h +├── main.cpp +├── model +│ └── 1.jpg +└── README.md +``` + +## Model Conversion + +### Download Model + +Click to download [resnet50v2.onnx](http://netstorage.allwinnertech.com:5000/sharing/WQ6FcVgs5). + +Then move the model to the convert_model/ directory. + +### Create Symlink for Conversion Script + + + +```bash +cd convert_model/ +./convert_model_env.sh +``` + + + +### Model Import/Quantization/Conversion + +You need to enter the container development environment first. Refer to the [Create Container](./model-zoo-download#创建并启动容器) section in Model Zoo Download. + +:::info +Different platforms use corresponding Docker images: + +- A733: ubuntu-npu:v2.0.10.1 +- T527: ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +After entering the container, navigate to the corresponding directory and run the script. + + + +```bash +cd /workspace/examples/resnet50v2/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh resnet50v2 +./pegasus_quantize.sh resnet50v2 uint8 10 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh resnet50v2 uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh resnet50v2 uint8 t527 +``` + + + + + + +The exported model files are stored in the ../model directory. + +### Compile Example + +Now you can compile the example. **First exit the container**, then execute the following command to compile the example. + +First, you need to configure third-party libraries and cross-compilation toolchain. + +:::info +You can skip this step if you have already configured third-party libraries and cross-compilation toolchain in other examples. +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +You need to manually download via [this link](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4) first, then place it in 0-toolchains/ before executing the following command: + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/resnet50v2/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## Model Deployment + +After compilation, the example will be installed in the install directory. You can use scp to transfer it to the board. + +### Configure NPU Driver + +:::info +You can skip this step if you have already configured NPU driver in other examples. +::: + +Transfer the driver library to the board's lib directory via scp. + +- A733 corresponds to the common/lib_linux_aarch64/A733 directory +- T527 corresponds to the common/lib_linux_aarch64/T527 directory + +Then execute the following command to export to environment variables. + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### Run Example + +After configuring the driver, you can run the example. + +:::tip +For T527 platform, you need to first enable NPU by referring to the A5E's "Enable NPU on Board" documentation, then use the following command to grant the current user permission to use /dev/vipcore. +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd resnet50v2_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./resnet50v2_demo_a733 +./resnet50v2_demo_a733 -nb model/resnet50v2_uint8_a733.nb -i model/1.jpg +``` + + + +The running result is as follows: + +```bash +$ ./resnet50v2_demo_a733 -nb model/resnet50v2_uint8_a733.nb -i model/1.jpg +model_file=model/resnet50v2_uint8_a733.nb, input=model/1.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 224 224 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 1000 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/resnet50v2_uint8_a733.nb, size: 17593328. +create network 0: 15664 us. +prepare network: 1734 us. +network: 0, loop count: 1 +run time for this network 0: 8900 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 281, prob: 11.682373, label: tabby, tabby cat +class id: 285, prob: 11.270020, label: Egyptian cat +class id: 282, prob: 10.033203, label: tiger cat +class id: 287, prob: 6.047363, label: lynx, catamount +class id: 292, prob: 5.085327, label: tiger, Panthera tigris +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :--------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner A733 | Vivante VIP9000 | resnet50v2 | 224×224 | 15.7 ms | 1.7 ms | 8.9 ms | | 26.3 ms | 38.0 FPS | + + + + + + + +```bash +cd resnet50v2_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./resnet50v2_demo_t527 +./resnet50v2_demo_t527 -nb model/resnet50v2_uint8_t527.nb -i model/1.jpg +``` + + + +The running result is as follows: + +```bash +$ ./resnet50v2_demo_t527 -nb model/resnet50v2_uint8_t527.nb -i model/1.jpg +model_file=model/resnet50v2_uint8_t527.nb, input=model/1.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 224 224 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 1000 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/resnet50v2_uint8_t527.nb, size: 17309120. +create network 0: 27251 us. +prepare network: 3129 us. +network: 0, loop count: 1 +run time for this network 0: 16298 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 281, prob: 11.682373, label: tabby, tabby cat +class id: 285, prob: 11.270020, label: Egyptian cat +class id: 282, prob: 10.033203, label: tiger cat +class id: 287, prob: 6.047363, label: lynx, catamount +class id: 292, prob: 5.085327, label: tiger, Panthera tigris +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :--------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner T527 | Vivante VIP9000 | resnet50v2 | 224×224 | 27.3 ms | 3.1 ms | 16.3 ms | | 46.7 ms | 21.4 FPS | + + + + +The running result is as follows: + +```bash +$ ./resnet50v2_demo_a733 -nb model/resnet50v2_uint8_a733.nb -i model/1.jpg +model_file=model/resnet50v2_uint8_a733.nb, input=model/1.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 224 224 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 1000 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/resnet50v2_uint8_a733.nb, size: 17593328. +create network 0: 15664 us. +prepare network: 1734 us. +network: 0, loop count: 1 +run time for this network 0: 8900 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 281, prob: 11.564558, label: tabby, tabby cat +class id: 285, prob: 10.988928, label: Egyptian cat +class id: 282, prob: 9.769331, label: tiger cat +class id: 287, prob: 5.913674, label: lynx, catamount +class id: 292, prob: 4.939133, label: tiger, Panthera tigris +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :--------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner A733 | Vivante VIP9000 | resnet50v2 | 224×224 | 15.7 ms | 1.7 ms | 8.9 ms | | 26.3 ms | 38.0 FPS | diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-retinaface.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-retinaface.mdx new file mode 100644 index 000000000..5b4e9a572 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-retinaface.mdx @@ -0,0 +1,372 @@ +This document describes how to run RetinaFace on NPU. + +:::info +Refer to [Model Zoo Download](./model-zoo-download) for the example. +::: + +RetinaFace Example Directory Structure: + +```bash +$ tree ./ +./ +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ └── Retinaface_resnet50_320.txt +├── figures +│ └── out_retinaface.png +├── main.cpp +├── model +│ └── test.jpg +├── model_config.h +├── README.md +├── retinaface_post.cpp +└── retinaface_pre.cpp +``` + +## Model Conversion + +### Export ONNX Model + +Click to download [Resnet50_Final.pth](https://drive.google.com/drive/folders/1oZRSG0ZegbVkVwUd8wUIQx8W7yfZ_ki1). + +### Download ONNX Model + +You can download the modified model. + +Click to download [Retinaface_resnet50_320.onnx](http://netstorage.allwinnertech.com:5000/sharing/YObunQV8S). + +Click to download [Retinaface_mobilenet0.25_320.onnx](http://netstorage.allwinnertech.com:5000/sharing/xt9rDVXzI). + +Then move to the convert_model/ directory. + +### Create Symlink for Conversion Script + + + +```bash +./convert_model_env.sh +``` + + + +### Model Import/Quantization/Conversion + +You need to enter the container development environment first. Refer to the [Create Container](./model-zoo-download#创建并启动容器) section in Model Zoo Download. + +:::info +Different platforms use corresponding Docker images: + +- A733: ubuntu-npu:v2.0.10.1 +- T527: ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +After entering the container, navigate to the corresponding directory and run the script. + + + +```bash +cd /workspace/examples/retinaface/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh Retinaface_resnet50_320 +./pegasus_quantize.sh Retinaface_resnet50_320 uint8 10 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh Retinaface_resnet50_320 uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh Retinaface_resnet50_320 uint8 t527 +``` + + + + + + +The exported model files are stored in the ../model directory. + +### Compile Example + +Now you can compile the example. **First exit the container**, then execute the following command to compile the example. + +First, you need to configure third-party libraries and cross-compilation toolchain. + +:::info +You can skip this step if you have already configured third-party libraries and cross-compilation toolchain in other examples. +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +You need to manually download via [this link](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4) first, then place it in 0-toolchains/ before executing the following command: + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/retinaface/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## Model Deployment + +After compilation, the example will be installed in the install directory. You can use scp to transfer it to the board. + +### Configure NPU Driver + +:::info +You can skip this step if you have already configured NPU driver in other examples. +::: + +Transfer the driver library to the board's lib directory via scp. + +- A733 corresponds to the common/lib_linux_aarch64/A733 directory +- T527 corresponds to the common/lib_linux_aarch64/T527 directory + +Then execute the following command to export to environment variables. + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### Run Example + +After configuring the driver, you can run the example. + +:::tip +For T527 platform, you need to first enable NPU by referring to the A5E's "Enable NPU on Board" documentation, then use the following command to grant the current user permission to use /dev/vipcore. +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd retinaface_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./retinaface_demo_a733 +./retinaface_demo_a733 -nb model/Retinaface_resnet50_320_uint8_a733.nb -i model/test.jpg +``` + + + +The running result is as follows: + +```bash +$ ./retinaface_demo_a733 -nb model/Retinaface_resnet50_320_uint8_a733.nb -i model/test.jpg +model_file=model/Retinaface_resnet50_320_uint8_a733.nb, input=model/test.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 320 320 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 4 4200 1 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 2 4200 1 0, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 10 4200 1 0, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +nbg name=model/Retinaface_resnet50_320_uint8_a733.nb, size: 19056048. +create network 0: 20781 us. +prepare network: 2285 us. +buffer ptr: 0x25971380, buffer size: 307200 +network: 0, loop count: 1 +run time for this network 0: 18564 us. +output 0, ptr 0x259d7240, size 16800. +output 1, ptr 0x259e5240, size 8400. +output 2, ptr 0x259f3240, size 42000. +post process time : 1 ms +detection num: 1 +100%, [ 244, 45, 363, 208], face +275.10 113.49 +328.03 112.26 +300.95 147.94 +277.57 166.40 +326.80 165.17 +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :------------------ | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner A733 | Vivante VIP9000 | Retinaface_resnet50 | 320×320 | 20.8 ms | 2.3 ms | 18.6 ms | 1.0 ms | 42.7 ms | 23.4 FPS | + + + + + + + +```bash +cd retinaface_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./retinaface_demo_t527 +./retinaface_demo_t527 -nb model/Retinaface_resnet50_320_uint8_t527.nb -i model/test.jpg +``` + + + +The running result is as follows: + +```bash +$ ./retinaface_demo_t527 -nb model/Retinaface_resnet50_320_uint8_t527.nb -i model/test.jpg +model_file=model/Retinaface_resnet50_320_uint8_t527.nb, input=model/test.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 320 320 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 4 4200 1 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 2 4200 1 0, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 10 4200 1 0, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +nbg name=model/Retinaface_resnet50_320_uint8_t527.nb, size: 18714688. +create network 0: 27602 us. +prepare network: 5276 us. +buffer ptr: 0x23c57380, buffer size: 307200 +network: 0, loop count: 1 +run time for this network 0: 30483 us. +output 0, ptr 0x23ca2440, size 16800. +output 1, ptr 0x23cb2b40, size 8400. +output 2, ptr 0x23cbaf40, size 42000. +post process time : 1 ms +detection num: 1 +100%, [ 244, 45, 363, 208], face +275.10 113.49 +328.03 112.26 +300.95 147.94 +277.57 166.40 +326.80 165.17 +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :------------------ | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner T527 | Vivante VIP9000 | Retinaface_resnet50 | 320×320 | 27.6 ms | 5.3 ms | 30.5 ms | 1.0 ms | 64.4 ms | 15.5 FPS | + + + + +The running result is as follows: + +```bash +$ ./retinaface_demo_a733 -nb model/Retinaface_resnet50_320_uint8_a733.nb -i model/test.jpg +model_file=model/Retinaface_resnet50_320_uint8_a733.nb, input=model/test.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 320 320 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 4 4200 1 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 2 4200 1 0, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 10 4200 1 0, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +nbg name=model/Retinaface_resnet50_320_uint8_a733.nb, size: 19056048. +create network 0: 20781 us. +prepare network: 2285 us. +buffer ptr: 0x25971380, buffer size: 307200 +network: 0, loop count: 1 +run time for this network 0: 15703 us. +output 0, ptr 0x259bc480, size 16800. +output 1, ptr 0x259ccb80, size 8400. +output 2, ptr 0x259d4f40, size 42000. +post process time : 0 ms +detection num: 1 +100%, [ 244, 46, 363, 209], face +275.10 113.49 +328.03 112.26 +300.95 147.94 +277.57 165.17 +326.80 165.17 +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :------------------ | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner A733 | Vivante VIP9000 | Retinaface_resnet50 | 320×320 | 20.8 ms | 2.3 ms | 15.7 ms | 0.0 ms | 38.8 ms | 25.8 FPS | diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-squeezenet-pytorch.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-squeezenet-pytorch.mdx new file mode 100644 index 000000000..e12e4d8d3 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-squeezenet-pytorch.mdx @@ -0,0 +1,342 @@ +This document describes how to run SqueezeNet on NPU. + +:::info +Refer to [Model Zoo Download](./model-zoo-download) for the example. +::: + +SqueezeNet Example Directory Structure: + +```bash +$ tree ./ +./ +├── class_post.cpp +├── class_pre.cpp +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ └── inputs_outputs.txt +├── label.h +├── main.cpp +├── model +│ └── space_shuttle_227x227.jpg +└── README.md +``` + +## Model Conversion + +### Download Model + +Click to download [squeezenet1_0.pt](http://netstorage.allwinnertech.com:5000/sharing/bSFCrqSqC). + +Then move the model to the convert_model/ directory. + +### Create Symlink for Conversion Script + + + +```bash +cd convert_model/ +./convert_model_env.sh +``` + + + +### Model Import/Quantization/Conversion + +You need to enter the container development environment first. Refer to the [Create Container](./model-zoo-download#创建并启动容器) section in Model Zoo Download. + +:::info +Different platforms use corresponding Docker images: + +- A733: ubuntu-npu:v2.0.10.1 +- T527: ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +After entering the container, navigate to the corresponding directory and run the script. + + + +```bash +cd /workspace/examples/squeezenet_pytorch/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh squeezenet1_0 +./pegasus_quantize.sh squeezenet1_0 uint8 10 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh squeezenet1_0 uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh squeezenet1_0 uint8 t527 +``` + + + + + + +The exported model files are stored in the ../model directory. + +### Compile Example + +Now you can compile the example. **First exit the container**, then execute the following command to compile the example. + +First, you need to configure third-party libraries and cross-compilation toolchain. + +:::info +You can skip this step if you have already configured third-party libraries and cross-compilation toolchain in other examples. +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +You need to manually download via [this link](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4) first, then place it in 0-toolchains/ before executing the following command: + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/squeezenet_pytorch/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## Model Deployment + +After compilation, the example will be installed in the install directory. You can use scp to transfer it to the board. + +### Configure NPU Driver + +:::info +You can skip this step if you have already configured NPU driver in other examples. +::: + +Transfer the driver library to the board's lib directory via scp. + +- A733 corresponds to the common/lib_linux_aarch64/A733 directory +- T527 corresponds to the common/lib_linux_aarch64/T527 directory + +Then execute the following command to export to environment variables. + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### Run Example + +After configuring the driver, you can run the example. + +:::tip +For T527 platform, you need to first enable NPU by referring to the A5E's "Enable NPU on Board" documentation, then use the following command to grant the current user permission to use /dev/vipcore. +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd squeezenet_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./squeezenet_demo_a733 +./squeezenet_demo_a733 -nb model/squeezenet1_0_uint8_a733.nb -i model/space_shuttle_227x227.jpg +``` + + + +The running result is as follows: + +```bash +$ ./squeezenet_demo_a733 -nb model/squeezenet1_0_uint8_a733.nb -i model/space_shuttle_227x227.jpg +model_file=model/squeezenet1_0_uint8_a733.nb, input=model/space_shuttle_227x227.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 227 227 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 1000 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/squeezenet1_0_uint8_a733.nb, size: 1066072. +create network 0: 2628 us. +prepare network: 844 us. +network: 0, loop count: 1 +run time for this network 0: 2455 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 812, prob: 16.893208, label: space shuttle +class id: 404, prob: 15.730225, label: airliner +class id: 833, prob: 14.984619, label: submarine, pigboat, sub, U-boat +class id: 554, prob: 14.686523, label: fireboat +class id: 895, prob: 14.611816, label: warplane, military plane +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :--------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner A733 | Vivante VIP9000 | squeezenet | 227×227 | 2.6 ms | 0.8 ms | 2.5 ms | | 5.9 ms | 169.5 FPS | + + + + + + + +```bash +cd squeezenet_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./squeezenet_demo_t527 +./squeezenet_demo_t527 -nb model/squeezenet1_0_uint8_t527.nb -i model/space_shuttle_227x227.jpg +``` + + + +The running result is as follows: + +```bash +$ ./squeezenet_demo_t527 -nb model/squeezenet1_0_uint8_t527.nb -i model/space_shuttle_227x227.jpg +model_file=model/squeezenet1_0_uint8_t527.nb, input=model/space_shuttle_227x227.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 227 227 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 1000 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/squeezenet1_0_uint8_t527.nb, size: 1078912. +create network 0: 2536 us. +prepare network: 1301 us. +network: 0, loop count: 1 +run time for this network 0: 3103 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 812, prob: 16.922852, label: space shuttle +class id: 404, prob: 15.730225, label: airliner +class id: 833, prob: 14.984619, label: submarine, pigboat, sub, U-boat +class id: 554, prob: 14.686523, label: fireboat +class id: 895, prob: 14.611816, label: warplane, military plane +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :--------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner T527 | Vivante VIP9000 | squeezenet | 227×227 | 2.5 ms | 1.3 ms | 3.1 ms | | 6.9 ms | 144.9 FPS | + + + + +The running result is as follows: + +```bash +$ ./squeezenet_demo_a733 -nb model/squeezenet1_0_uint8_a733.nb -i model/space_shuttle_227x227.jpg +model_file=model/squeezenet1_0_uint8_a733.nb, input=model/space_shuttle_227x227.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 227 227 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 1000 1 0 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +nbg name=model/squeezenet1_0_uint8_a733.nb, size: 1066072. +create network 0: 2628 us. +prepare network: 844 us. +network: 0, loop count: 1 +run time for this network 0: 2455 us. +class_postprocess.cpp run. +========== top5 ========== +class id: 812, prob: 16.893208, label: space shuttle +class id: 404, prob: 15.714762, label: airliner +class id: 833, prob: 14.979437, label: submarine, pigboat, sub, U-boat +class id: 554, prob: 14.664755, label: fireboat +class id: 895, prob: 14.632645, label: warplane, military plane +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :--------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner A733 | Vivante VIP9000 | squeezenet | 227×227 | 2.6 ms | 0.8 ms | 2.5 ms | | 5.9 ms | 169.5 FPS | diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolo11-pose.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolo11-pose.mdx new file mode 100644 index 000000000..0db392b91 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolo11-pose.mdx @@ -0,0 +1,481 @@ +This document describes how to run YOLO11 Pose on NPU. + +:::info +Refer to [Model Zoo Download](./model-zoo-download) for the example. +::: + +YOLO11 Pose Example Directory Structure: + +```bash +$ tree ./ +./ +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ ├── python +│ │ ├── onnx_extract.py +│ │ └── yolo11s-pose_640.txt +│ └── yolo11s-pose_9.txt +├── figures +│ ├── diff_img.png +│ └── out_yolo11_pose_pcq.png +├── main.cpp +├── model +│ └── COCO_train2014_000000500390.jpg +├── model_config.h +├── README.md +├── yolo11_pose_9_post.cpp +└── yolo11_pose_9_pre.cpp +``` + +## Model Conversion + +### Configure Virtual Environment + + + +```bash +python -m venv .venv && source .venv/bin/activate +pip install ultralytics +``` + + + +### Export ONNX Model + + + +```bash +cd convert_model/python/ +yolo export model=yolo11s-pose.pt format=onnx simplify=True dynamic=False opset=11 nms=False batch=1 device=cpu +``` + + + +### Prune Model + + + +```bash +python onnx_extract.py +mv ./yolo11s-pose_9.onnx ../ +cd .. +``` + + + +### Create Symlink for Conversion Script + + + +```bash +./convert_model_env.sh +``` + + + +### Model Import/Quantization/Conversion + +You need to enter the container development environment first. Refer to the [Create Container](./model-zoo-download#创建并启动容器) section in Model Zoo Download. + +:::info +Different platforms use corresponding Docker images: + +- A733: ubuntu-npu:v2.0.10.1 +- T527: ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +After entering the container, navigate to the corresponding directory and run the script. + + + +```bash +cd /workspace/examples/yolo11_pose/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh yolo11s-pose_9 +./pegasus_quantize.sh yolo11s-pose_9 uint8 12 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolo11s-pose_9 uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolo11s-pose_9 uint8 t527 +``` + + + + + + +The exported model files are stored in the ../model directory. + +### Compile Example + +Now you can compile the example. **First exit the container**, then execute the following command to compile the example. + +First, you need to configure third-party libraries and cross-compilation toolchain. + +:::info +You can skip this step if you have already configured third-party libraries and cross-compilation toolchain in other examples. +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +You need to manually download via [this link](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4) first, then place it in 0-toolchains/ before executing the following command: + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/yolo11_pose/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## Model Deployment + +After compilation, the example will be installed in the install directory. You can use scp to transfer it to the board. + +### Configure NPU Driver + +:::info +You can skip this step if you have already configured NPU driver in other examples. +::: + +Transfer the driver library to the board's lib directory via scp. + +- A733 corresponds to the common/lib_linux_aarch64/A733 directory +- T527 corresponds to the common/lib_linux_aarch64/T527 directory + +Then execute the following command to export to environment variables. + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### Run Example + +After configuring the driver, you can run the example. + +:::tip +For T527 platform, you need to first enable NPU by referring to the A5E's "Enable NPU on Board" documentation, then use the following command to grant the current user permission to use /dev/vipcore. +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd yolo11_pose_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./yolo11_pose_demo_a733 +./yolo11_pose_demo_a733 -nb model/yolo11s-pose_9_uint8_a733.nb -i model/COCO_train2014_000000500390.jpg +``` + + + +The running result is as follows: + +```bash +$ ./yolo11_pose_demo_a733 -nb model/yolo11s-pose_9_uint8_a733.nb -i model/COCO_train2014_000000500390.jpg +model_file=model/yolo11s-pose_9_uint8_a733.nb, input=model/COCO_train2014_000000500390.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 80 80 64 1, data_format=0, name=uid_17_out_0b_uid_1_out_0, none-quant +output 1 dim 80 80 1 1, data_format=0, name=uid_16_out_0b_uid_1_out_0, none-quant +output 2 dim 80 80 51 1, data_format=0, name=uid_15_out_0b_uid_1_out_0, none-quant +output 3 dim 40 40 64 1, data_format=0, name=uid_14_out_0b_uid_1_out_0, none-quant +output 4 dim 40 40 1 1, data_format=0, name=uid_13_out_0b_uid_1_out_0, none-quant +output 5 dim 40 40 51 1, data_format=0, name=uid_12_out_0b_uid_1_out_0, none-quant +output 6 dim 20 20 64 1, data_format=0, name=uid_11_out_0b_uid_1_out_0, none-quant +output 7 dim 20 20 1 1, data_format=0, name=uid_10_out_0b_uid_1_out_0, none-quant +output 8 dim 20 20 51 1, data_format=0, name=uid_9_out_0ub_uid_1_out_0, none-quant +nbg name=model/yolo11s-pose_9_uint8_a733.nb, size: 7284048. +create network 0: 16110 us. +prepare network: 3977 us. +buffer ptr: 0x202f5380, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 32374 us. +output 0, ptr 0x20421480, size 409600. +output 1, ptr 0x205b1500, size 6400. +output 2, ptr 0x205b7980, size 326400. +output 3, ptr 0x206f6640, size 102400. +output 4, ptr 0x2075a6c0, size 1600. +output 5, ptr 0x2075c040, size 81600. +output 6, ptr 0x207abbc0, size 25600. +output 7, ptr 0x207c4c80, size 400. +output 8, ptr 0x207c5340, size 20400. +post process time : 4 ms +detection num: 3 + 0: 94%, [ 370, 0, 589, 346], person +405.75 26.20 = 0.96988 +419.11 23.03 = 0.96501 +405.65 21.63 = 0.29929 +441.04 31.18 = 0.99146 +421.11 22.33 = 0.04379 +455.76 67.51 = 0.99977 +430.35 62.14 = 0.99950 +466.39 121.18 = 0.99797 +405.08 109.99 = 0.98330 +447.50 96.32 = 0.98985 +382.14 70.42 = 0.94582 +466.06 166.69 = 0.99986 +455.44 165.19 = 0.99974 +411.43 242.60 = 0.99939 +497.02 230.87 = 0.99880 +408.66 307.99 = 0.98213 +562.98 301.11 = 0.97806 + 0: 88%, [ 86, 27, 292, 389], person +146.77 66.48 = 0.99659 +157.56 60.56 = 0.99517 +138.54 62.15 = 0.93738 +177.10 58.15 = 0.97191 +136.75 58.00 = 0.22714 +182.16 88.95 = 0.99876 +146.17 100.58 = 0.99755 +210.99 144.46 = 0.99757 +161.48 152.14 = 0.98186 +171.08 179.70 = 0.99453 +131.07 188.60 = 0.97326 +222.98 197.17 = 0.99975 +178.42 204.61 = 0.99950 +250.05 264.09 = 0.99831 +151.41 290.25 = 0.99650 +287.21 296.16 = 0.97016 +127.74 355.67 = 0.95755 + 0: 92%, [ 228, 39, 399, 407], person +275.86 94.61 = 0.99351 +286.44 88.28 = 0.98999 +267.42 87.73 = 0.88035 +308.03 73.30 = 0.97833 +265.48 74.83 = 0.23741 +339.54 98.91 = 0.99963 +280.47 109.74 = 0.99938 +372.16 125.90 = 0.99505 +272.82 170.12 = 0.98157 +380.93 163.22 = 0.98073 +243.21 204.51 = 0.94730 +339.07 225.60 = 0.99986 +302.82 223.45 = 0.99980 +294.02 310.02 = 0.99952 +314.44 286.69 = 0.99926 +270.90 355.43 = 0.99344 +374.00 318.30 = 0.99277 +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :----------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner A733 | Vivante VIP9000 | yolo11s-pose | 640×640 | 16.1 ms | 4.0 ms | 32.4 ms | 4.0 ms | 56.5 ms | 17.7 FPS | + + + + + + + +```bash +cd yolo11_pose_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./yolo11_pose_demo_t527 +./yolo11_pose_demo_t527 -nb model/yolo11s-pose_9_uint8_t527.nb -i model/COCO_train2014_000000500390.jpg +``` + + + +The running result is as follows: + +```bash +$ ./yolo11_pose_demo_t527 -nb model/yolo11s-pose_9_uint8_t527.nb -i model/COCO_train2014_000000500390.jpg +model_file=model/yolo11s-pose_9_uint8_t527.nb, input=model/COCO_train2014_000000500390.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 80 80 64 1, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 80 80 1 1, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 80 80 51 1, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +output 3 dim 40 40 64 1, data_format=0, name=uid_20003_sub_uid_1_out_0, none-quant +output 4 dim 40 40 1 1, data_format=0, name=uid_20004_sub_uid_1_out_0, none-quant +output 5 dim 40 40 51 1, data_format=0, name=uid_20005_sub_uid_1_out_0, none-quant +output 6 dim 20 20 64 1, data_format=0, name=uid_20006_sub_uid_1_out_0, none-quant +output 7 dim 20 20 1 1, data_format=0, name=uid_20007_sub_uid_1_out_0, none-quant +output 8 dim 20 20 51 1, data_format=0, name=uid_20008_sub_uid_1_out_0, none-quant +nbg name=model/yolo11s-pose_9_uint8_t527.nb, size: 8148288. +create network 0: 23417 us. +prepare network: 10280 us. +buffer ptr: 0x22f74380, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 75989 us. +output 0, ptr 0x230a0440, size 409600. +output 1, ptr 0x23230500, size 6400. +output 2, ptr 0x23236980, size 326400. +output 3, ptr 0x23375600, size 102400. +output 4, ptr 0x233d9680, size 1600. +output 5, ptr 0x233db040, size 81600. +output 6, ptr 0x2342abc0, size 25600. +output 7, ptr 0x23443c40, size 400. +output 8, ptr 0x23444300, size 20400. +post process time : 11 ms +detection num: 3 + 0: 94%, [ 371, 0, 587, 346], person +406.01 30.36 = 0.96783 +418.34 26.25 = 0.95618 +406.01 26.25 = 0.45198 +434.78 30.36 = 0.98485 +418.34 26.25 = 0.12014 +455.34 71.46 = 0.99981 +426.56 67.35 = 0.99938 +467.67 120.79 = 0.99790 +406.01 104.35 = 0.98050 +451.23 96.12 = 0.98748 +389.57 75.57 = 0.93298 +475.89 174.23 = 0.99984 +459.45 170.12 = 0.99967 +414.23 244.11 = 0.99955 +484.11 235.88 = 0.99903 +410.12 301.65 = 0.99145 +558.10 297.54 = 0.98825 + 0: 87%, [ 86, 27, 292, 389], person +147.67 66.46 = 0.99650 +160.00 58.25 = 0.99518 +139.45 58.25 = 0.94058 +180.55 58.25 = 0.96977 +135.34 54.13 = 0.22788 +180.55 87.02 = 0.99866 +147.67 99.35 = 0.99729 +213.44 144.57 = 0.99729 +164.11 148.68 = 0.98050 +172.33 177.45 = 0.99417 +131.23 189.78 = 0.97160 +221.66 198.00 = 0.99971 +176.44 202.12 = 0.99946 +250.43 263.77 = 0.99816 +151.78 288.44 = 0.99627 +283.32 292.55 = 0.96783 +123.00 354.21 = 0.95618 + 0: 92%, [ 228, 38, 399, 408], person +275.67 96.12 = 0.99247 +288.00 87.90 = 0.98897 +267.45 87.90 = 0.86560 +308.55 75.57 = 0.97646 +267.45 75.57 = 0.22788 +341.44 100.23 = 0.99960 +279.78 108.46 = 0.99930 +374.32 124.90 = 0.99487 +275.67 170.12 = 0.97924 +382.54 161.89 = 0.98050 +242.78 203.00 = 0.94407 +341.44 227.66 = 0.99985 +304.44 223.55 = 0.99978 +292.11 309.88 = 0.99949 +312.66 285.21 = 0.99920 +271.56 355.09 = 0.99294 +374.32 318.10 = 0.99198 +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :----------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner T527 | Vivante VIP9000 | yolo11s-pose | 640×640 | 23.4 ms | 10.3 ms | 76.0 ms | 11.0 ms | 120.7 ms | 8.3 FPS | + + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolo11-seg.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolo11-seg.mdx new file mode 100644 index 000000000..a83f1e79f --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolo11-seg.mdx @@ -0,0 +1,383 @@ +This document describes how to run YOLO11 Seg on NPU. + +:::info +Refer to [Model Zoo Download](./model-zoo-download) for the example. +::: + +YOLO11 Seg Example Directory Structure: + +```bash +$ tree ./ +./ +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ ├── python +│ │ ├── onnx_extract.py +│ │ └── yolo11s-seg_640.txt +│ └── yolo11s-seg_10.txt +├── figures +│ ├── diff_img.png +│ └── out_yolo11_seg_pcq.png +├── main.cpp +├── model +│ └── dog.jpg +├── model_config.h +├── README.md +├── yolo11_seg_10_post.cpp +└── yolo11_seg_10_pre.cpp +``` + +## Model Conversion + +### Configure Virtual Environment + + + +```bash +python -m venv .venv && source .venv/bin/activate +pip install ultralytics +``` + + + +### Export ONNX Model + + + +```bash +cd convert_model/python/ +yolo export model=yolo11s-seg.pt format=onnx imgsz=640 dynamic=False simplify=True opset=11 nms=False batch=1 device=cpu +``` + + + +### Prune Model + + + +```bash +python onnx_extract.py +mv yolo11s-seg_10.onnx ../ +cd .. +``` + + + +### Create Symlink for Conversion Script + + + +```bash +./convert_model_env.sh +``` + + + +### Model Import/Quantization/Conversion + +You need to enter the container development environment first. Refer to the [Create Container](./model-zoo-download#创建并启动容器) section in Model Zoo Download. + +:::info +Different platforms use corresponding Docker images: + +- A733: ubuntu-npu:v2.0.10.1 +- T527: ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +After entering the container, navigate to the corresponding directory and run the script. + + + +```bash +cd /workspace/examples/yolo11_seg/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh yolo11s-seg_10 +./pegasus_quantize.sh yolo11s-seg_10 uint8 12 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolo11s-seg_10 uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolo11s-seg_10 uint8 t527 +``` + + + + + + +The exported model files are stored in the ../model directory. + +### Compile Example + +Now you can compile the example. **First exit the container**, then execute the following command to compile the example. + +First, you need to configure third-party libraries and cross-compilation toolchain. + +:::info +You can skip this step if you have already configured third-party libraries and cross-compilation toolchain in other examples. +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +You need to manually download via [this link](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4) first, then place it in 0-toolchains/ before executing the following command: + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/yolo11_seg/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## Model Deployment + +After compilation, the example will be installed in the install directory. You can use scp to transfer it to the board. + +### Configure NPU Driver + +:::info +You can skip this step if you have already configured NPU driver in other examples. +::: + +Transfer the driver library to the board's lib directory via scp. + +- A733 corresponds to the common/lib_linux_aarch64/A733 directory +- T527 corresponds to the common/lib_linux_aarch64/T527 directory + +Then execute the following command to export to environment variables. + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### Run Example + +After configuring the driver, you can run the example. + +:::tip +For T527 platform, you need to first enable NPU by referring to the A5E's "Enable NPU on Board" documentation, then use the following command to grant the current user permission to use /dev/vipcore. +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd yolo11_seg_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./yolo11_seg_demo_a733 +./yolo11_seg_demo_a733 -nb model/yolo11s-seg_10_uint8_a733.nb -i model/dog.jpg +``` + + + +The running result is as follows: + +```bash +$ ./yolo11_seg_demo_a733 -nb model/yolo11s-seg_10_uint8_a733.nb -i model/dog.jpg +model_file=model/yolo11s-seg_10_uint8_a733.nb, input=model/dog.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 80 80 64 1, data_format=0, name=uid_19_out_0b_uid_1_out_0, none-quant +output 1 dim 80 80 80 1, data_format=0, name=uid_18_out_0b_uid_1_out_0, none-quant +output 2 dim 80 80 32 1, data_format=0, name=uid_17_out_0b_uid_1_out_0, none-quant +output 3 dim 40 40 64 1, data_format=0, name=uid_16_out_0b_uid_1_out_0, none-quant +output 4 dim 40 40 80 1, data_format=0, name=uid_15_out_0b_uid_1_out_0, none-quant +output 5 dim 40 40 32 1, data_format=0, name=uid_14_out_0b_uid_1_out_0, none-quant +output 6 dim 20 20 64 1, data_format=0, name=uid_13_out_0b_uid_1_out_0, none-quant +output 7 dim 20 20 80 1, data_format=0, name=uid_12_out_0b_uid_1_out_0, none-quant +output 8 dim 20 20 32 1, data_format=0, name=uid_11_out_0b_uid_1_out_0, none-quant +output 9 dim 160 160 32 1, data_format=0, name=uid_20009_sub_uid_1_out_0, none-quant +nbg name=model/yolo11s-seg_10_uint8_a733.nb, size: 7326672. +create network 0: 24693 us. +prepare network: 2986 us. +buffer ptr: 0x226f1600, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 37744 us. +output 0, ptr 0x2281d780, size 409600. +output 1, ptr 0x229ad800, size 512000. +output 2, ptr 0x22ba1880, size 204800. +output 3, ptr 0x22c69900, size 102400. +output 4, ptr 0x22ccd9c0, size 128000. +output 5, ptr 0x22d4aa40, size 51200. +output 6, ptr 0x22d7cac0, size 25600. +output 7, ptr 0x22d95b40, size 32000. +output 8, ptr 0x22db5000, size 12800. +output 9, ptr 0x22dc1880, size 819200. +post process time : 11 ms +detection num: 3 + 1: 95%, [ 126, 126, 568, 420], bicycle +16: 95%, [ 131, 221, 311, 541], dog + 2: 86%, [ 467, 75, 691, 172], car +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :---------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner A733 | Vivante VIP9000 | yolo11s-seg | 640×640 | 24.7 ms | 3.0 ms | 37.7 ms | 11.0 ms | 76.4 ms | 13.1 FPS | + + + + + + + +```bash +cd yolo11_seg_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./yolo11_seg_demo_t527 +./yolo11_seg_demo_t527 -nb model/yolo11s-seg_10_uint8_t527.nb -i model/dog.jpg +``` + + + +The running result is as follows: + +```bash +$ ./yolo11_seg_demo_t527 -nb model/yolo11s-seg_10_uint8_t527.nb -i model/dog.jpg +model_file=model/yolo11s-seg_10_uint8_t527.nb, input=model/dog.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 80 80 64 1, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 80 80 80 1, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 80 80 32 1, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +output 3 dim 40 40 64 1, data_format=0, name=uid_20003_sub_uid_1_out_0, none-quant +output 4 dim 40 40 80 1, data_format=0, name=uid_20004_sub_uid_1_out_0, none-quant +output 5 dim 40 40 32 1, data_format=0, name=uid_20005_sub_uid_1_out_0, none-quant +output 6 dim 20 20 64 1, data_format=0, name=uid_20006_sub_uid_1_out_0, none-quant +output 7 dim 20 20 80 1, data_format=0, name=uid_20007_sub_uid_1_out_0, none-quant +output 8 dim 20 20 32 1, data_format=0, name=uid_20008_sub_uid_1_out_0, none-quant +output 9 dim 160 160 32 1, data_format=0, name=uid_20009_sub_uid_1_out_0, none-quant +nbg name=model/yolo11s-seg_10_uint8_t527.nb, size: 8522240. +create network 0: 26153 us. +prepare network: 11813 us. +buffer ptr: 0x38e48600, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 94147 us. +output 0, ptr 0x38f74740, size 409600. +output 1, ptr 0x391047c0, size 512000. +output 2, ptr 0x392f8880, size 204800. +output 3, ptr 0x393c0900, size 102400. +output 4, ptr 0x39424980, size 128000. +output 5, ptr 0x394a1a00, size 51200. +output 6, ptr 0x394d3ac0, size 25600. +output 7, ptr 0x394ecb40, size 32000. +output 8, ptr 0x3950bfc0, size 12800. +output 9, ptr 0x39518840, size 819200. +post process time : 51 ms +detection num: 3 + 1: 94%, [ 126, 124, 568, 420], bicycle +16: 95%, [ 132, 222, 311, 541], dog + 2: 82%, [ 467, 76, 692, 172], car +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :---------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner T527 | Vivante VIP9000 | yolo11s-seg | 640×640 | 26.2 ms | 11.8 ms | 94.1 ms | 51.0 ms | 183.1 ms | 5.5 FPS | + + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolo11.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolo11.mdx new file mode 100644 index 000000000..8afd4951c --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolo11.mdx @@ -0,0 +1,365 @@ +This document describes how to run YOLO11 on NPU. + +:::info +Refer to [Model Zoo Download](./model-zoo-download) for the example. +::: + +YOLO11 Example Directory Structure: + +```bash +$ tree ./ +./ +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ ├── python +│ │ ├── onnx_extract.py +│ │ └── yolo11s_640.txt +│ └── yolo11s_6.txt +├── figures +│ ├── diff_img.png +│ └── out_yolo11.png +├── main.cpp +├── model +│ └── dog.jpg +├── model_config.h +├── README.md +├── yolo11_6_post.cpp +└── yolo11_6_pre.cpp +``` + +## Model Conversion + +### Configure Virtual Environment + + + +```bash +python -m venv .venv && source .venv/bin/activate +pip install ultralytics +``` + + + +### Export ONNX Model + + + +```bash +cd convert_model/python +yolo export model=yolo11s.pt format=onnx imgsz=640 simplify=True dynamic=False opset=11 nms=False batch=1 device=cpu +``` + + + +### Prune Model + + + +```bash +python onnx_extract.py +mv yolo11s_6.onnx ../ +cd ../ +``` + + + +### Create Symlink for Conversion Script + + + +```bash +./convert_model_env.sh +``` + + + +### Model Import/Quantization/Conversion + +You need to enter the container development environment first. Refer to the [Create Container](./model-zoo-download#创建并启动容器) section in Model Zoo Download. + +:::info +Different platforms use corresponding Docker images: + +- A733: ubuntu-npu:v2.0.10.1 +- T527: ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +After entering the container, navigate to the corresponding directory and run the script. + + + +```bash +cd /workspace/examples/yolo11/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh yolo11s_6 +./pegasus_quantize.sh yolo11s_6 uint8 12 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolo11s_6 uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolo11s_6 uint8 t527 +``` + + + + + + +The exported model files are stored in the ../model directory. + +### Compile Example + +Now you can compile the example. **First exit the container**, then execute the following command to compile the example. + +First, you need to configure third-party libraries and cross-compilation toolchain. + +:::info +You can skip this step if you have already configured third-party libraries and cross-compilation toolchain in other examples. +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +You need to manually download via [this link](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4) first, then place it in 0-toolchains/ before executing the following command: + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/yolo11/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## Model Deployment + +After compilation, the example will be installed in the install directory. You can use scp to transfer it to the board. + +### Configure NPU Driver + +:::info +You can skip this step if you have already configured NPU driver in other examples. +::: + +Transfer the driver library to the board's lib directory via scp. + +- A733 corresponds to the common/lib_linux_aarch64/A733 directory +- T527 corresponds to the common/lib_linux_aarch64/T527 directory + +Then execute the following command to export to environment variables. + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### Run Example + +After configuring the driver, you can run the example. + +:::tip +For T527 platform, you need to first enable NPU by referring to the A5E's "Enable NPU on Board" documentation, then use the following command to grant the current user permission to use /dev/vipcore. +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd yolo11_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./yolo11_demo_a733 +./yolo11_demo_a733 -nb model/yolo11s_6_uint8_a733.nb -i model/dog.jpg +``` + + + +The running result is as follows: + +```bash +$ ./yolo11_demo_a733 -nb model/yolo11s_6_uint8_a733.nb -i model/dog.jpg +model_file=model/yolo11s_6_uint8_a733.nb, input=model/dog.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 80 80 64 1, data_format=0, name=uid_11_out_0b_uid_1_out_0, none-quant +output 1 dim 80 80 80 1, data_format=0, name=uid_10_out_0b_uid_1_out_0, none-quant +output 2 dim 40 40 64 1, data_format=0, name=uid_9_out_0ub_uid_1_out_0, none-quant +output 3 dim 40 40 80 1, data_format=0, name=uid_8_out_0ub_uid_1_out_0, none-quant +output 4 dim 20 20 64 1, data_format=0, name=uid_7_out_0ub_uid_1_out_0, none-quant +output 5 dim 20 20 80 1, data_format=0, name=uid_6_out_0ub_uid_1_out_0, none-quant +nbg name=model/yolo11s_6_uint8_a733.nb, size: 6850432. +create network 0: 18253 us. +prepare network: 4469 us. +buffer ptr: 0x2baf600, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 31930 us. +output 0, ptr 0x2cdb740, size 409600. +output 1, ptr 0x2e6b7c0, size 512000. +output 2, ptr 0x305f840, size 102400. +output 3, ptr 0x30c38c0, size 128000. +output 4, ptr 0x3140980, size 25600. +output 5, ptr 0x3159a00, size 32000. +detection num: 3 + 1: 95%, [ 126, 130, 568, 419], bicycle +16: 93%, [ 132, 220, 311, 541], dog + 7: 50%, [ 465, 74, 692, 170], truck +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :------ | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner A733 | Vivante VIP9000 | yolo11s | 640×640 | 18.3 ms | 4.5 ms | 31.9 ms | 5 ms | 59.7 ms | 16.8 FPS | + + + + + + + +```bash +cd yolo11_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./yolo11_demo_t527 +./yolo11_demo_t527 -nb model/yolo11s_6_uint8_t527.nb -i model/dog.jpg +``` + + + +The running result is as follows: + +```bash +$ ./yolo11_demo_t527 -nb model/yolo11s_6_uint8_t527.nb -i model/dog.jpg +model_file=model/yolo11s_6_uint8_t527.nb, input=model/dog.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 80 80 64 1, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 80 80 80 1, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 40 40 64 1, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +output 3 dim 40 40 80 1, data_format=0, name=uid_20003_sub_uid_1_out_0, none-quant +output 4 dim 20 20 64 1, data_format=0, name=uid_20004_sub_uid_1_out_0, none-quant +output 5 dim 20 20 80 1, data_format=0, name=uid_20005_sub_uid_1_out_0, none-quant +nbg name=model/yolo11s_6_uint8_t527.nb, size: 7783808. +create network 0: 21201 us. +prepare network: 10246 us. +buffer ptr: 0xbf9e600, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 75191 us. +output 0, ptr 0xc0ca700, size 409600. +output 1, ptr 0xc25a780, size 512000. +output 2, ptr 0xc44e840, size 102400. +output 3, ptr 0xc4b28c0, size 128000. +output 4, ptr 0xc52f940, size 25600. +output 5, ptr 0xc5489c0, size 32000. +detection num: 3 + 1: 94%, [ 127, 129, 567, 419], bicycle +16: 93%, [ 132, 220, 312, 541], dog + 2: 46%, [ 465, 75, 693, 171], car +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :------ | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner T527 | Vivante VIP9000 | yolo11s | 640×640 | 21.2 ms | 10.2 ms | 75.2 ms | | 106.6 ms | 9.4 FPS | + + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolo26.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolo26.mdx new file mode 100644 index 000000000..3d6b8a49e --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolo26.mdx @@ -0,0 +1,357 @@ +This document describes how to run YOLO26 on NPU. + +:::info +Refer to [Model Zoo Download](./model-zoo-download) for the example. +::: + +YOLO26 Example Directory Structure: + +```bash +$ tree ./ +./ +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ ├── python +│ │ ├── onnx_extract.py +│ │ └── yolo26s_640.txt +│ └── yolo26s_6.txt +├── figures +│ ├── banner-yolo26.png +│ ├── bus.jpg +│ ├── out_yolo26_6_pcq.png +│ └── performance-comparison.png +├── main.cpp +├── model +│ └── dog.jpg +├── model_config.h +├── README.md +├── yolov26_6_post.cpp +└── yolov26_6_pre.cpp +``` + +## Model Conversion + +### Configure Virtual Environment + + + +```bash +python -m venv .venv && source .venv/bin/activate +pip install ultralytics +``` + + + +### Export ONNX Model + +ultralytics will automatically download the model and any missing dependencies. Please be patient. + + + +```bash +cd convert_model/python/ +yolo export model=yolo26s.pt format=onnx simplify=True dynamic=False opset=16 +``` + + + +### Prune Model + + + +```bash +python onnx_extract.py +cd .. +``` + + + +### Create Symlink for Conversion Script + + + +```bash +./convert_model_env.sh +``` + + + +### Model Import/Quantization/Conversion + +You need to enter the container development environment first. Refer to the [Create Container](./model-zoo-download#创建并启动容器) section in Model Zoo Download. + +:::info +Different platforms use corresponding Docker images: + +- A733: ubuntu-npu:v2.0.10.1 +- T527: ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +After entering the container, navigate to the corresponding directory and run the script. + + + +```bash +# Navigate to the corresponding directory in the container +cd /workspace/examples/yolo26/convert_model/ +./pegasus_import.sh yolo26s_6 # Model name without suffix +./pegasus_quantize.sh yolo26s_6 pcq 12 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolo26s_6 pcq a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolo26s_6 pcq t527 +``` + + + + + + +The exported model files are stored in the ../model directory. + +### Compile Example + +Now you can compile the example. **First exit the container**, then execute the following command to compile the example. + +First, you need to configure third-party libraries and cross-compilation toolchain. + +:::info +You can skip this step if you have already configured third-party libraries and cross-compilation toolchain in other examples. +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +You need to manually download via [this link](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4) first, then place it in 0-toolchains/ before executing the following command: + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + + + + +```bash +cd ../examples/yolo26/ +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +cd ../examples/yolo26/ +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## Model Deployment + +After compilation, the example will be installed in the install directory. You can use scp to transfer it to the board. + +### Configure NPU Driver + +:::info +You can skip this step if you have already configured NPU driver in other examples. +::: + +Transfer the driver library to the board's lib directory via scp. + +- A733 corresponds to the common/lib_linux_aarch64/A733 directory +- T527 corresponds to the common/lib_linux_aarch64/T527 directory + +Then execute the following command to export to environment variables. + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### Run Example + +After configuring the driver, you can run the example. + +:::tip +For T527 platform, you need to first enable NPU by referring to the A5E's "Enable NPU on Board" documentation, then use the following command to grant the current user permission to use /dev/vipcore. +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd yolo26_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./yolo26_demo_a733 +./yolo26_demo_a733 -nb model/yolo26s_6_pcq_a733.nb -i model/dog.jpg +``` + + + +The running result is as follows: + +```bash +$ ./yolo26_demo_a733 -nb model/yolo26s_6_pcq_a733.nb -i model/dog.jpg +model_file=model/yolo26s_6_pcq_a733.nb, input=model/dog.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 6400 4 1 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 1600 4 1 0, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 400 4 1 0, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +output 3 dim 6400 80 1 0, data_format=0, name=uid_20003_sub_uid_1_out_0, none-quant +output 4 dim 1600 80 1 0, data_format=0, name=uid_20004_sub_uid_1_out_0, none-quant +output 5 dim 400 80 1 0, data_format=0, name=uid_20005_sub_uid_1_out_0, none-quant +nbg name=model/yolo26s_6_pcq_a733.nb, size: 9362920. +create network 0: 18319 us. +prepare network: 8871 us. +buffer ptr: 0x5cb2600, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 34765 us. +output 0, ptr 0x5dde740, size 25600. +output 1, ptr 0x5df77c0, size 6400. +output 2, ptr 0x5dfdc40, size 1600. +output 3, ptr 0x5dff5c0, size 512000. +output 4, ptr 0x5ff3680, size 128000. +output 5, ptr 0x6070700, size 32000. +postprocess time : 6 ms +detection num: 3 + 7: 68%, [ 466, 74, 690, 171], truck + 1: 89%, [ 130, 136, 566, 420], bicycle +16: 90%, [ 133, 221, 310, 540], dog +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :------ | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner A733 | Vivante VIP9000 | yolo26s | 640×640 | 18.3 ms | 8.9 ms | 34.8 ms | 6 ms | 68.0 ms | 14.7 FPS | + + + + + + + +```bash +cd yolo26_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./yolo26_demo_t527 +./yolo26_demo_t527 -nb model/yolo26s_6_pcq_t527.nb -i model/dog.jpg +``` + + + +The running result is as follows: + +```bash +$ ./yolo26_demo_t527 -nb model/yolo26s_6_pcq_t527.nb -i model/dog.jpg +model_file=model/yolo26s_6_pcq_t527.nb, input=model/dog.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 6400 4 1 0, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 1600 4 1 0, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 400 4 1 0, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +output 3 dim 6400 80 1 0, data_format=0, name=uid_20003_sub_uid_1_out_0, none-quant +output 4 dim 1600 80 1 0, data_format=0, name=uid_20004_sub_uid_1_out_0, none-quant +output 5 dim 400 80 1 0, data_format=0, name=uid_20005_sub_uid_1_out_0, none-quant +nbg name=model/yolo26s_6_pcq_t527.nb, size: 9920576. +create network 0: 20397 us. +prepare network: 11311 us. +buffer ptr: 0x6fe6600, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 87270 us. +output 0, ptr 0x7112700, size 25600. +output 1, ptr 0x712b780, size 6400. +output 2, ptr 0x7131c40, size 1600. +output 3, ptr 0x71335c0, size 512000. +output 4, ptr 0x7327640, size 128000. +output 5, ptr 0x73a46c0, size 32000. +postprocess time : 20 ms +detection num: 2 + 1: 94%, [ 128, 135, 566, 418], bicycle +16: 92%, [ 133, 219, 311, 541], dog +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :------ | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner T527 | Vivante VIP9000 | yolo26s | 640×640 | 20.4 ms | 11.3 ms | 87.3 ms | 20.0 ms | 139.0 ms | 7.2 FPS | + + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolov3-darknet.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolov3-darknet.mdx new file mode 100644 index 000000000..40aad4f3b --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolov3-darknet.mdx @@ -0,0 +1,320 @@ +This document describes how to run YOLOv3 on NPU. + +:::info +Refer to [Model Zoo Download](./model-zoo-download) for the example. +::: + +YOLOv3 Example Directory Structure: + +```bash +$ tree ./ +./ +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ └── yolov3.cfg +├── main.cpp +├── model +│ └── horses_416x416.jpg +├── model_config.h +├── README.md +├── yolov3_post.cpp +└── yolov3_pre.cpp +``` + +## Model Conversion + +### Download Model File + + + +```bash +cd convert_model/ +wget https://pjreddie.com/media/files/yolov3.weights +``` + + + +### Create Symlink for Conversion Script + + + +```bash +./convert_model_env.sh +``` + + + +### Model Import/Quantization/Conversion + +You need to enter the container development environment first. Refer to the [Create Container](./model-zoo-download#创建并启动容器) section in Model Zoo Download. + +:::info +Different platforms use corresponding Docker images: + +- A733: ubuntu-npu:v2.0.10.1 +- T527: ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +After entering the container, navigate to the corresponding directory and run the script. + + + +```bash +cd /workspace/examples/yolov3_darknet/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh yolov3 +./pegasus_quantize.sh yolov3 uint8 12 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolov3 uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolov3 uint8 t527 +``` + + + + + + +The exported model files are stored in the ../model directory. + +### Compile Example + +Now you can compile the example. **First exit the container**, then execute the following command to compile the example. + +First, you need to configure third-party libraries and cross-compilation toolchain. + +:::info +You can skip this step if you have already configured third-party libraries and cross-compilation toolchain in other examples. +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +You need to manually download via [this link](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4) first, then place it in 0-toolchains/ before executing the following command: + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/yolov3/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## Model Deployment + +After compilation, the example will be installed in the install directory. You can use scp to transfer it to the board. + +### Configure NPU Driver + +:::info +You can skip this step if you have already configured NPU driver in other examples. +::: + +Transfer the driver library to the board's lib directory via scp. + +- A733 corresponds to the common/lib_linux_aarch64/A733 directory +- T527 corresponds to the common/lib_linux_aarch64/T527 directory + +Then execute the following command to export to environment variables. + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### Run Example + +After configuring the driver, you can run the example. + +:::tip +For T527 platform, you need to first enable NPU by referring to the A5E's "Enable NPU on Board" documentation, then use the following command to grant the current user permission to use /dev/vipcore. +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd yolov3_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./yolov3_demo_a733 +./yolov3_demo_a733 -nb model/yolov3_uint8_a733.nb -i model/horses_416x416.jpg +``` + + + +The running result is as follows: + +```bash +$ ./yolov3_demo_a733 -nb model/yolov3_uint8_a733.nb -i model/horses_416x416.jpg +model_file=model/yolov3_uint8_a733.nb, input=model/horses_416x416.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 416 416 3 1, data_format=2, quant_format=2, name=input/output[0], scale=0.003906, zero_point=0 +output 0 dim 13 13 255 1, data_format=2, name=uid_198_out_0, scale=0.191153, zero_point=189 +output 1 dim 26 26 255 1, data_format=2, name=uid_224_out_0, scale=0.213471, zero_point=198 +output 2 dim 52 52 255 1, data_format=2, name=uid_250_out_0, scale=0.323550, zero_point=184 +nbg name=model/yolov3_uint8_a733.nb, size: 38927048. +create network 0: 22932 us. +prepare network: 3310 us. +buffer ptr: 0xffffab170040, buffer size: 519168 +feed input cost: 19560 us. +network: 0, loop count: 1 +run time for this network 0: 35442 us. +detection num: 3 +17: 98%, [ 234, 169, 324, 288], horse +17: 88%, [ 137, 149, 233, 281], horse +17: 88%, [ 8, 165, 197, 329], horse +draw objects time : 13 ms +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :----- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner A733 | Vivante VIP9000 | yolov3 | 416×416 | 22.9 ms | 3.3 ms | 35.4 ms | 13 ms | 74.6 ms | 13.4 FPS | + + + + + + + +```bash +cd yolov3_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./yolov3_demo_t527 +./yolov3_demo_t527 -nb model/yolov3_uint8_t527.nb -i model/horses_416x416.jpg +``` + + + +The running result is as follows: + +```bash +$ ./yolov3_demo_t527 -nb model/yolov3_uint8_t527.nb -i model/horses_416x416.jpg +model_file=model/yolov3_uint8_t527.nb, input=model/horses_416x416.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 416 416 3 1, data_format=2, quant_format=2, name=input[0], scale=0.003906, zero_point=0 +output 0 dim 13 13 255 1, data_format=2, name=uid_20000_out_0, scale=0.178953, zero_point=178 +output 1 dim 26 26 255 1, data_format=2, name=uid_20001_out_0, scale=0.199956, zero_point=191 +output 2 dim 52 52 255 1, data_format=2, name=uid_20002_out_0, scale=0.319580, zero_point=170 +nbg name=model/yolov3_uint8_t527.nb, size: 38927048. +create network 0: 54287 us. +prepare network: 10433 us. +buffer ptr: 0xffff9d7e7040, buffer size: 519168 +network: 0, loop count: 1 +run time for this network 0: 63582 us. +detection num: 3 +17: 91%, [ 8, 165, 197, 329], horse +17: 88%, [ 127, 148, 243, 280], horse +draw objects time : 44 ms +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :----- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner T527 | Vivante VIP9000 | yolov3 | 416×416 | 54.3 ms | 10.4 ms | 63.6 ms | 44.0 ms | 172.3 ms | 5.8 FPS | + + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolov5.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolov5.mdx new file mode 100644 index 000000000..841ebe91d --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolov5.mdx @@ -0,0 +1,329 @@ +This document describes how to run YOLOv5 on NPU. + +:::info +Refer to [Model Zoo Download](./model-zoo-download) for the example. +::: + +YOLOv5 Example Directory Structure: + +```bash +$ tree ./ +./ +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ ├── python +│ │ ├── onnx_extract.py +│ │ └── yolov5s-sim.onnx +│ └── yolov5s_rt.onnx +├── figures +│ ├── dog.jpg +│ ├── onnx.jpg +│ ├── onnx-sim.jpg +│ ├── output.jpg +│ └── output_yolov5_uint8.png +├── main.cpp +├── model +│ ├── dog.jpg +│ ├── yolov5s_rt_uint8_a733.nb +│ └── yolov5s_rt_uint8_t527.nb +├── model_config.h +├── README.md +├── yolov5_post.cpp +└── yolov5_pre.cpp +``` + +## Model Conversion + +No need to export to ONNX model. Just use the yolov5s_rt.onnx provided in the repository. + +### Create Symlink for Conversion Script + + + +```bash +cd convert_model/ +./convert_model_env.sh +``` + + + +### Model Import/Quantization/Conversion + +You need to enter the container development environment first. Refer to the [Create Container](./model-zoo-download#创建并启动容器) section in Model Zoo Download. + +:::info +Different platforms use corresponding Docker images: + +- A733: ubuntu-npu:v2.0.10.1 +- T527: ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +After entering the container, navigate to the corresponding directory and run the script. + + + +```bash +cd /workspace/examples/yolov5/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh yolov5s_rt +./pegasus_quantize.sh yolov5s_rt uint8 12 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolov5s_rt uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolov5s_rt uint8 t527 +``` + + + + + + +The exported model files are stored in the ../model directory. + +### Compile Example + +Now you can compile the example. **First exit the container**, then execute the following command to compile the example. + +First, you need to configure third-party libraries and cross-compilation toolchain. + +:::info +You can skip this step if you have already configured third-party libraries and cross-compilation toolchain in other examples. +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +You need to manually download via [this link](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4) first, then place it in 0-toolchains/ before executing the following command: + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/yolov5/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## Model Deployment + +After compilation, the example will be installed in the install directory. You can use scp to transfer it to the board. + +### Configure NPU Driver + +:::info +You can skip this step if you have already configured NPU driver in other examples. +::: + +Transfer the driver library to the board's lib directory via scp. + +- A733 corresponds to the common/lib_linux_aarch64/A733 directory +- T527 corresponds to the common/lib_linux_aarch64/T527 directory + +Then execute the following command to export to environment variables. + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### Run Example + +After configuring the driver, you can run the example. + +:::tip +For T527 platform, you need to first enable NPU by referring to the A5E's "Enable NPU on Board" documentation, then use the following command to grant the current user permission to use /dev/vipcore. +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd yolov5_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./yolov5_demo_a733 +./yolov5_demo_a733 -nb model/yolov5s_rt_uint8_a733.nb -i model/dog.jpg +``` + + + +The running result is as follows: + +```bash +$ ./yolov5_demo_a733 -nb model/yolov5s_rt_uint8_a733.nb -i model/dog.jpg +model_file=model/yolov5s_rt_uint8_a733.nb, input=model/dog.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 80 80 3 1, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 40 40 3 1, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 20 20 3 1, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +output 3 dim 80 80 80 1, data_format=0, name=uid_20003_sub_uid_1_out_0, none-quant +output 4 dim 40 40 80 1, data_format=0, name=uid_20005_sub_uid_1_out_0, none-quant +output 5 dim 20 20 80 1, data_format=0, name=uid_20006_sub_uid_1_out_0, none-quant +nbg name=model/yolov5s_rt_uint8_a733.nb, size: 7246236. +create network 0: 10429 us. +prepare network: 1587 us. +buffer ptr: 0x190c0300, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 31862 us. +detection num: 3 +16: 93%, [ 137, 229, 303, 539], dog + 7: 70%, [ 471, 79, 687, 171], truck + 1: 47%, [ 151, 120, 560, 429], bicycle +draw objects time : 44 ms +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :------ | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner A733 | Vivante VIP9000 | yolov5s | 640×640 | 10.4 ms | 1.6 ms | 31.9 ms | 44.0 ms | 87.9 ms | 11.4 FPS | + + + + + + + +```bash +cd yolov5_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./yolov5_demo_t527 +./yolov5_demo_t527 -nb model/yolov5s_rt_uint8_t527.nb -i model/dog.jpg +``` + + + +The running result is as follows: + +```bash +$ ./yolov5_demo_t527 -nb model/yolov5s_rt_uint8_t527.nb -i model/dog.jpg +model_file=model/yolov5s_rt_uint8_t527.nb, input=model/dog.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 80 80 3 1, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 40 40 3 1, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 20 20 3 1, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +output 3 dim 80 80 80 1, data_format=0, name=uid_20003_sub_uid_1_out_0, none-quant +output 4 dim 40 40 80 1, data_format=0, name=uid_20004_sub_uid_1_out_0, none-quant +output 5 dim 20 20 80 1, data_format=0, name=uid_20005_sub_uid_1_out_0, none-quant +nbg name=model/yolov5s_rt_uint8_t527.nb, size: 7246236. +create network 0: 24280 us. +prepare network: 10836 us. +buffer ptr: 0xbe8ff600, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 49985 us. +detection num: 3 +16: 92%, [ 137, 229, 303, 539], dog + 7: 69%, [ 471, 79, 687, 171], truck + 1: 44%, [ 151, 120, 560, 429], bicycle +draw objects time : 81 ms +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :------ | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner T527 | Vivante VIP9000 | yolov5s | 640×640 | 24.3 ms | 10.8 ms | 50.0 ms | 81.0 ms | 166.1 ms | 6.0 FPS | + + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolov8-pose.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolov8-pose.mdx new file mode 100644 index 000000000..8bf56c508 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolov8-pose.mdx @@ -0,0 +1,490 @@ +This document describes how to run YOLOv8 Pose on NPU. + +:::info +Refer to [Model Zoo Download](./model-zoo-download) for the example. +::: + +YOLOv8 Pose Example Directory Structure: + +```bash +$ tree ./ +./ +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ ├── python +│ │ ├── onnx_extract.py +│ │ └── yolov8s-pose_640.txt +│ └── yolov8s-pose_9.txt +├── figures +│ ├── diff_img.png +│ └── out_yolov8_pose_pcq.png +├── main.cpp +├── model +│ └── COCO_train2014_000000500390.jpg +├── model_config.h +├── README.md +├── yolov8_pose_9_post.cpp +└── yolov8_pose_9_pre.cpp +``` + +## Model Conversion + +### Configure Virtual Environment + + + +```bash +python -m venv .venv && source .venv/bin/activate +pip install ultralytics==8.1.0 onnxsim +``` + + + +### Export ONNX Model + + + +```bash +cd convert_model/python/ +yolo export model=yolov8s-pose.pt format=onnx dynamic=True opset=11 +``` + + + +### Fixed Shape + + + +```bash +python3 -m onnxsim yolov8s-pose.onnx yolov8s-pose_640.onnx --input-shape=1,3,640,640 +``` + + + +### Prune Model + + + +```bash +python3 onnx_extract.py +cd .. +``` + + + +### Create Symlink for Conversion Script + + + +```bash +./convert_model_env.sh +``` + + + +### Model Import/Quantization/Conversion + +You need to enter the container development environment first. Refer to the [Create Container](./model-zoo-download#创建并启动容器) section in Model Zoo Download. + +:::info +Different platforms use corresponding Docker images: + +- A733: ubuntu-npu:v2.0.10.1 +- T527: ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +After entering the container, navigate to the corresponding directory and run the script. + + + +```bash +cd /workspace/examples/yolov8_pose/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh yolov8s-pose_9 +./pegasus_quantize.sh yolov8s-pose_9 uint8 12 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolov8s-pose_9 uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolov8s-pose_9 uint8 t527 +``` + + + + + + +The exported model files are stored in the ../model directory. + +### Compile Example + +Now you can compile the example. **First exit the container**, then execute the following command to compile the example. + +First, you need to configure third-party libraries and cross-compilation toolchain. + +:::info +You can skip this step if you have already configured third-party libraries and cross-compilation toolchain in other examples. +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +You need to manually download via [this link](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4) first, then place it in 0-toolchains/ before executing the following command: + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/yolov8_pose/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## Model Deployment + +After compilation, the example will be installed in the install directory. You can use scp to transfer it to the board. + +### Configure NPU Driver + +:::info +You can skip this step if you have already configured NPU driver in other examples. +::: + +Transfer the driver library to the board's lib directory via scp. + +- A733 corresponds to the common/lib_linux_aarch64/A733 directory +- T527 corresponds to the common/lib_linux_aarch64/T527 directory + +Then execute the following command to export to environment variables. + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### Run Example + +After configuring the driver, you can run the example. + +:::tip +For T527 platform, you need to first enable NPU by referring to the A5E's "Enable NPU on Board" documentation, then use the following command to grant the current user permission to use /dev/vipcore. +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd yolov8_pose_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./yolov8_pose_demo_a733 +./yolov8_pose_demo_a733 -nb model/yolov8s-pose_9_uint8_a733.nb -i model/COCO_train2014_000000500390.jpg +``` + + + +The running result is as follows: + +```bash +$ ./yolov8_pose_demo_a733 -nb model/yolov8s-pose_9_uint8_a733.nb -i model/COCO_train2014_000000500390.jpg +model_file=model/yolov8s-pose_9_uint8_a733.nb, input=model/COCO_train2014_000000500390.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 80 80 64 1, data_format=0, name=uid_17_out_0b_uid_1_out_0, none-quant +output 1 dim 80 80 1 1, data_format=0, name=uid_16_out_0b_uid_1_out_0, none-quant +output 2 dim 80 80 51 1, data_format=0, name=uid_15_out_0b_uid_1_out_0, none-quant +output 3 dim 40 40 64 1, data_format=0, name=uid_14_out_0b_uid_1_out_0, none-quant +output 4 dim 40 40 1 1, data_format=0, name=uid_13_out_0b_uid_1_out_0, none-quant +output 5 dim 40 40 51 1, data_format=0, name=uid_12_out_0b_uid_1_out_0, none-quant +output 6 dim 20 20 64 1, data_format=0, name=uid_11_out_0b_uid_1_out_0, none-quant +output 7 dim 20 20 1 1, data_format=0, name=uid_10_out_0b_uid_1_out_0, none-quant +output 8 dim 20 20 51 1, data_format=0, name=uid_9_out_0ub_uid_1_out_0, none-quant +nbg name=model/yolov8s-pose_9_uint8_a733.nb, size: 7768344. +create network 0: 18985 us. +prepare network: 5711 us. +buffer ptr: 0x28344380, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 32958 us. +output 0, ptr 0x28470480, size 409600. +output 1, ptr 0x28600500, size 6400. +output 2, ptr 0x28606980, size 326400. +output 3, ptr 0x28745640, size 102400. +output 4, ptr 0x287a96c0, size 1600. +output 5, ptr 0x287ab040, size 81600. +output 6, ptr 0x287fabc0, size 25600. +output 7, ptr 0x28813c80, size 400. +output 8, ptr 0x28814340, size 20400. +post process time : 4 ms +detection num: 3 + 0: 93%, [ 373, 1, 587, 346], person +411.58 41.32 = 0.98922 +419.64 35.78 = 0.98396 +416.36 37.19 = 0.76423 +440.57 37.33 = 0.97060 +422.52 38.08 = 0.12822 +450.85 69.58 = 0.99924 +422.50 75.59 = 0.99804 +473.26 121.09 = 0.99354 +405.71 108.13 = 0.95213 +449.46 97.35 = 0.98640 +389.93 81.26 = 0.92587 +461.08 161.27 = 0.99969 +461.53 162.40 = 0.99945 +405.04 226.47 = 0.99954 +489.77 240.76 = 0.99892 +415.75 320.01 = 0.99481 +555.85 276.33 = 0.99307 + 0: 93%, [ 86, 28, 288, 390], person +155.68 76.87 = 0.99271 +162.21 68.34 = 0.98739 +145.10 65.45 = 0.95864 +175.03 64.92 = 0.91619 +141.19 64.45 = 0.68796 +199.98 93.83 = 0.99730 +160.28 98.94 = 0.99395 +214.27 138.31 = 0.99138 +164.10 156.53 = 0.98026 +175.57 174.47 = 0.98414 +136.65 193.82 = 0.97464 +216.19 199.03 = 0.99952 +180.07 198.95 = 0.99935 +240.79 270.84 = 0.99790 +150.18 279.74 = 0.99727 +293.96 281.26 = 0.98766 +128.72 359.77 = 0.98534 + 0: 91%, [ 227, 36, 398, 405], person +281.36 106.56 = 0.99230 +287.73 97.59 = 0.98680 +279.19 104.41 = 0.88281 +308.31 83.30 = 0.95046 +275.74 96.24 = 0.19450 +328.64 102.08 = 0.99900 +275.67 126.89 = 0.99741 +373.27 126.41 = 0.99145 +278.76 161.70 = 0.95360 +382.76 163.57 = 0.98179 +249.65 205.81 = 0.91721 +332.66 214.56 = 0.99969 +309.45 218.78 = 0.99948 +293.53 304.20 = 0.99923 +310.05 306.35 = 0.99817 +279.65 380.53 = 0.99397 +363.50 304.91 = 0.99248 +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :----------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner A733 | Vivante VIP9000 | yolov8s-pose | 640×640 | 19.0 ms | 5.7 ms | 33.0 ms | 4.0 ms | 61.7 ms | 16.2 FPS | + + + + + + + +```bash +cd yolov8_pose_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./yolov8_pose_demo_t527 +./yolov8_pose_demo_t527 -nb model/yolov8s-pose_9_uint8_t527.nb -i model/COCO_train2014_000000500390.jpg +``` + + + +The running result is as follows: + +```bash +$ ./yolov8_pose_demo_t527 -nb model/yolov8s-pose_9_uint8_t527.nb -i model/COCO_train2014_000000500390.jpg +model_file=model/yolov8s-pose_9_uint8_t527.nb, input=model/COCO_train2014_000000500390.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 80 80 64 1, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 80 80 1 1, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 80 80 51 1, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +output 3 dim 40 40 64 1, data_format=0, name=uid_20003_sub_uid_1_out_0, none-quant +output 4 dim 40 40 1 1, data_format=0, name=uid_20004_sub_uid_1_out_0, none-quant +output 5 dim 40 40 51 1, data_format=0, name=uid_20005_sub_uid_1_out_0, none-quant +output 6 dim 20 20 64 1, data_format=0, name=uid_20006_sub_uid_1_out_0, none-quant +output 7 dim 20 20 1 1, data_format=0, name=uid_20007_sub_uid_1_out_0, none-quant +output 8 dim 20 20 51 1, data_format=0, name=uid_20008_sub_uid_1_out_0, none-quant +nbg name=model/yolov8s-pose_9_uint8_t527.nb, size: 10369024. +create network 0: 25460 us. +prepare network: 20276 us. +buffer ptr: 0x2a796380, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 71053 us. +output 0, ptr 0x2a8c2440, size 409600. +output 1, ptr 0x2aa52500, size 6400. +output 2, ptr 0x2aa58980, size 326400. +output 3, ptr 0x2ab97600, size 102400. +output 4, ptr 0x2abfb680, size 1600. +output 5, ptr 0x2abfd040, size 81600. +output 6, ptr 0x2ac4cbc0, size 25600. +output 7, ptr 0x2ac65c40, size 400. +output 8, ptr 0x2ac66300, size 20400. +post process time : 11 ms +detection num: 3 + 0: 93%, [ 373, 1, 587, 347], person +411.64 37.38 = 0.98651 +419.72 33.34 = 0.98042 +415.68 33.34 = 0.72047 +435.88 33.34 = 0.97493 +423.76 37.38 = 0.13819 +452.04 65.66 = 0.99938 +423.76 73.73 = 0.99807 +472.24 118.17 = 0.99438 +407.60 102.01 = 0.94482 +452.04 93.93 = 0.98564 +391.45 81.81 = 0.90655 +460.12 158.56 = 0.99969 +460.12 158.56 = 0.99945 +403.56 227.24 = 0.99963 +492.44 243.40 = 0.99904 +411.64 320.15 = 0.99614 +557.07 283.79 = 0.99438 + 0: 93%, [ 86, 28, 288, 389], person +155.96 77.46 = 0.99278 +164.04 69.38 = 0.98809 +143.84 65.34 = 0.95914 +176.16 65.34 = 0.92141 +139.80 65.34 = 0.68079 +200.40 93.62 = 0.99736 +160.00 97.66 = 0.99363 +212.51 138.05 = 0.99128 +164.04 158.25 = 0.97784 +176.16 174.41 = 0.98374 +135.76 194.60 = 0.97166 +216.55 198.64 = 0.99949 +180.20 198.64 = 0.99930 +240.79 271.36 = 0.99781 +151.92 279.44 = 0.99700 +293.30 283.47 = 0.98732 +127.68 356.19 = 0.98472 + 0: 92%, [ 227, 36, 398, 406], person +279.60 105.73 = 0.99278 +287.68 97.66 = 0.98809 +279.60 105.73 = 0.88286 +307.88 81.50 = 0.95390 +275.56 97.66 = 0.18974 +328.08 101.70 = 0.99904 +275.56 125.93 = 0.99736 +372.51 125.93 = 0.99181 +279.60 162.29 = 0.94802 +384.63 162.29 = 0.98270 +251.33 206.72 = 0.91177 +332.12 214.80 = 0.99969 +307.88 218.84 = 0.99945 +291.72 303.67 = 0.99925 +307.88 307.71 = 0.99807 +279.60 380.42 = 0.99438 +364.44 303.67 = 0.99231 +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :----------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner T527 | Vivante VIP9000 | yolov8s-pose | 640×640 | 25.5 ms | 20.3 ms | 71.1 ms | 11.0 ms | 127.9 ms | 7.8 FPS | + + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolov8-seg.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolov8-seg.mdx new file mode 100644 index 000000000..327291305 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolov8-seg.mdx @@ -0,0 +1,393 @@ +This document describes how to run YOLOv8 Seg on NPU. + +:::info +Refer to [Model Zoo Download](./model-zoo-download) for the example. +::: + +YOLOv8 Seg Example Directory Structure: + +```bash +$ tree ./ +./ +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ ├── python +│ │ ├── onnx_extract.py +│ │ └── yolov8s-seg_640.txt +│ └── yolov8s-seg_10.txt +├── figures +│ ├── diff_img.png +│ └── out_yolov8_seg.png +├── main.cpp +├── model +│ ├── bus.jpg +│ └── dog.jpg +├── model_config.h +├── README.md +├── yolov8_seg_10_post.cpp +└── yolov8_seg_10_pre.cpp +``` + +## Model Conversion + +### Configure Virtual Environment + + + +```bash +python -m venv .venv && source .venv/bin/activate +pip install ultralytics onnxsim +``` + + + +### Export ONNX Model + + + +```bash +cd convert_model/python/ +yolo export model=yolov8s-seg.pt format=onnx dynamic=True opset=11 +``` + + + +### Fixed Shape + + + +```bash +python3 -m onnxsim yolov8s-seg.onnx yolov8s-seg_640.onnx --input-shape=1,3,640,640 +``` + + + +### Prune Model + + + +```bash +python3 onnx_extract.py +cd .. +``` + + + +### Create Symlink for Conversion Script + + + +```bash +./convert_model_env.sh +``` + + + +### Model Import/Quantization/Conversion + +You need to enter the container development environment first. Refer to the [Create Container](./model-zoo-download#创建并启动容器) section in Model Zoo Download. + +:::info +Different platforms use corresponding Docker images: + +- A733: ubuntu-npu:v2.0.10.1 +- T527: ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +After entering the container, navigate to the corresponding directory and run the script. + + + +```bash +cd /workspace/examples/yolov8_seg/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh yolov8s-seg_10 +./pegasus_quantize.sh yolov8s-seg_10 uint8 12 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolov8s-seg_10 uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolov8s-seg_10 uint8 t527 +``` + + + + + + +The exported model files are stored in the ../model directory. + +### Compile Example + +Now you can compile the example. **First exit the container**, then execute the following command to compile the example. + +First, you need to configure third-party libraries and cross-compilation toolchain. + +:::info +You can skip this step if you have already configured third-party libraries and cross-compilation toolchain in other examples. +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +You need to manually download via [this link](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4) first, then place it in 0-toolchains/ before executing the following command: + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/yolov8_seg/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## Model Deployment + +After compilation, the example will be installed in the install directory. You can use scp to transfer it to the board. + +### Configure NPU Driver + +:::info +You can skip this step if you have already configured NPU driver in other examples. +::: + +Transfer the driver library to the board's lib directory via scp. + +- A733 corresponds to the common/lib_linux_aarch64/A733 directory +- T527 corresponds to the common/lib_linux_aarch64/T527 directory + +Then execute the following command to export to environment variables. + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### Run Example + +After configuring the driver, you can run the example. + +:::tip +For T527 platform, you need to first enable NPU by referring to the A5E's "Enable NPU on Board" documentation, then use the following command to grant the current user permission to use /dev/vipcore. +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd yolov8_seg_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./yolov8_seg_demo_a733 +./yolov8_seg_demo_a733 -nb model/yolov8s-seg_10_uint8_a733.nb -i model/dog.jpg +``` + + + +The running result is as follows: + +```bash +$ ./yolov8_seg_demo_a733 -nb model/yolov8s-seg_10_uint8_a733.nb -i model/dog.jpg +model_file=model/yolov8s-seg_10_uint8_a733.nb, input=model/dog.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 80 80 64 1, data_format=0, name=uid_19_out_0b_uid_1_out_0, none-quant +output 1 dim 80 80 80 1, data_format=0, name=uid_18_out_0b_uid_1_out_0, none-quant +output 2 dim 80 80 32 1, data_format=0, name=uid_17_out_0b_uid_1_out_0, none-quant +output 3 dim 40 40 64 1, data_format=0, name=uid_16_out_0b_uid_1_out_0, none-quant +output 4 dim 40 40 80 1, data_format=0, name=uid_15_out_0b_uid_1_out_0, none-quant +output 5 dim 40 40 32 1, data_format=0, name=uid_14_out_0b_uid_1_out_0, none-quant +output 6 dim 20 20 64 1, data_format=0, name=uid_13_out_0b_uid_1_out_0, none-quant +output 7 dim 20 20 80 1, data_format=0, name=uid_12_out_0b_uid_1_out_0, none-quant +output 8 dim 20 20 32 1, data_format=0, name=uid_11_out_0b_uid_1_out_0, none-quant +output 9 dim 160 160 32 1, data_format=0, name=uid_20009_sub_uid_1_out_0, none-quant +nbg name=model/yolov8s-seg_10_uint8_a733.nb, size: 8089336. +create network 0: 21684 us. +prepare network: 4881 us. +buffer ptr: 0x127ba600, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 38144 us. +output 0, ptr 0x128e6780, size 409600. +output 1, ptr 0x12a76800, size 512000. +output 2, ptr 0x12c6a880, size 204800. +output 3, ptr 0x12d32900, size 102400. +output 4, ptr 0x12d969c0, size 128000. +output 5, ptr 0x12e13a40, size 51200. +output 6, ptr 0x12e45ac0, size 25600. +output 7, ptr 0x12e5eb40, size 32000. +output 8, ptr 0x12e7e000, size 12800. +output 9, ptr 0x12e8a880, size 819200. +post process time : 12 ms +detection num: 3 + 1: 89%, [ 126, 133, 568, 425], bicycle +16: 96%, [ 131, 220, 310, 541], dog + 7: 65%, [ 470, 73, 688, 171], truck +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :---------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner A733 | Vivante VIP9000 | yolov8s-seg | 640×640 | 21.7 ms | 4.9 ms | 38.1 ms | 12 ms | 76.7 ms | 13.0 FPS | + + + + + + + +```bash +cd yolov8_seg_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./yolov8_seg_demo_t527 +./yolov8_seg_demo_t527 -nb model/yolov8s-seg_10_uint8_t527.nb -i model/dog.jpg +``` + + + +The running result is as follows: + +```bash +$ ./yolov8_seg_demo_t527 -nb model/yolov8s-seg_10_uint8_t527.nb -i model/dog.jpg +model_file=model/yolov8s-seg_10_uint8_t527.nb, input=model/dog.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 80 80 64 1, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 80 80 80 1, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 80 80 32 1, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +output 3 dim 40 40 64 1, data_format=0, name=uid_20003_sub_uid_1_out_0, none-quant +output 4 dim 40 40 80 1, data_format=0, name=uid_20004_sub_uid_1_out_0, none-quant +output 5 dim 40 40 32 1, data_format=0, name=uid_20005_sub_uid_1_out_0, none-quant +output 6 dim 20 20 64 1, data_format=0, name=uid_20006_sub_uid_1_out_0, none-quant +output 7 dim 20 20 80 1, data_format=0, name=uid_20007_sub_uid_1_out_0, none-quant +output 8 dim 20 20 32 1, data_format=0, name=uid_20008_sub_uid_1_out_0, none-quant +output 9 dim 160 160 32 1, data_format=0, name=uid_20009_sub_uid_1_out_0, none-quant +nbg name=model/yolov8s-seg_10_uint8_t527.nb, size: 11076992. +create network 0: 34337 us. +prepare network: 22250 us. +buffer ptr: 0x10312600, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 88575 us. +output 0, ptr 0x1043e740, size 409600. +output 1, ptr 0x105ce7c0, size 512000. +output 2, ptr 0x107c2880, size 204800. +output 3, ptr 0x1088a900, size 102400. +output 4, ptr 0x108ee980, size 128000. +output 5, ptr 0x1096ba00, size 51200. +output 6, ptr 0x1099dac0, size 25600. +output 7, ptr 0x109b6b40, size 32000. +output 8, ptr 0x109d5fc0, size 12800. +output 9, ptr 0x109e2840, size 819200. +post process time : 54 ms +detection num: 3 + 1: 89%, [ 126, 133, 568, 426], bicycle +16: 95%, [ 131, 220, 310, 541], dog + 7: 66%, [ 470, 73, 688, 170], truck +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :---------- | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner T527 | Vivante VIP9000 | yolov8s-seg | 640×640 | 34.3 ms | 22.3 ms | 88.6 ms | 54.0 ms | 199.2 ms | 5.0 FPS | + + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolov8.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolov8.mdx new file mode 100644 index 000000000..5d67406ad --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolov8.mdx @@ -0,0 +1,375 @@ +This document describes how to run YOLOv8 on NPU. + +:::info +Refer to [Model Zoo Download](./model-zoo-download) for the example. +::: + +YOLOv8 Example Directory Structure: + +```bash +$ tree ./ +./ +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ ├── python +│ │ └── onnx_extract.py +│ └── yolov8n_6.onnx +├── figures +│ ├── banner-yolo-vision-2023.png +│ ├── bus.jpg +│ ├── out_yolov8_640.png +│ └── yolo-comparison-plots.png +├── main.cpp +├── model +│ └── dog.jpg +├── model_config.h +├── README.md +├── yolov8_6_post.cpp +└── yolov8_6_pre.cpp +``` + +## Model Conversion + +### Configure Virtual Environment + + + +```bash +python -m venv .venv && source .venv/bin/activate +pip install ultralytics==8.1.0 onnxsim +``` + + + +### Export ONNX Model + + + +```bash +cd convert_model/python/ +yolo export model=yolov8n.pt format=onnx dynamic=True opset=11 +``` + + + +### Fixed Shape + + + +```bash +python3 -m onnxsim yolov8n.onnx yolov8n_640_sim.onnx --input-shape=1,3,640,640 +``` + + + +### Prune Model + + + +```bash +python3 onnx_extract.py +cd .. +``` + + + +### Create Symlink for Conversion Script + + + +```bash +./convert_model_env.sh +``` + + + +### Model Import/Quantization/Conversion + +You need to enter the container development environment first. Refer to the [Create Container](./model-zoo-download#创建并启动容器) section in Model Zoo Download. + +:::info +Different platforms use corresponding Docker images: + +- A733: ubuntu-npu:v2.0.10.1 +- T527: ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +After entering the container, navigate to the corresponding directory and run the script. + + + +```bash +cd /workspace/examples/yolov8/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh yolov8n_6 +./pegasus_quantize.sh yolov8n_6 uint8 12 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolov8n_6 uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolov8n_6 uint8 t527 +``` + + + + + + +The exported model files are stored in the ../model directory. + +### Compile Example + +Now you can compile the example. **First exit the container**, then execute the following command to compile the example. + +First, you need to configure third-party libraries and cross-compilation toolchain. + +:::info +You can skip this step if you have already configured third-party libraries and cross-compilation toolchain in other examples. +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +You need to manually download via [this link](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4) first, then place it in 0-toolchains/ before executing the following command: + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/yolov8/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## Model Deployment + +After compilation, the example will be installed in the install directory. You can use scp to transfer it to the board. + +### Configure NPU Driver + +:::info +You can skip this step if you have already configured NPU driver in other examples. +::: + +Transfer the driver library to the board's lib directory via scp. + +- A733 corresponds to the common/lib_linux_aarch64/A733 directory +- T527 corresponds to the common/lib_linux_aarch64/T527 directory + +Then execute the following command to export to environment variables. + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### Run Example + +After configuring the driver, you can run the example. + +:::tip +For T527 platform, you need to first enable NPU by referring to the A5E's "Enable NPU on Board" documentation, then use the following command to grant the current user permission to use /dev/vipcore. +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd yolov8_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./yolov8_demo_a733 +./yolov8_demo_a733 -nb model/yolov8n_6_uint8_a733.nb -i model/dog.jpg +``` + + + +The running result is as follows: + +```bash +$ ./yolov8_demo_a733 -nb model/yolov8n_6_uint8_a733.nb -i model/dog.jpg +model_file=model/yolov8n_6_uint8_a733.nb, input=model/dog.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 80 80 64 1, data_format=0, name=uid_11_out_0b_uid_1_out_0, none-quant +output 1 dim 80 80 80 1, data_format=0, name=uid_10_out_0b_uid_1_out_0, none-quant +output 2 dim 40 40 64 1, data_format=0, name=uid_9_out_0ub_uid_1_out_0, none-quant +output 3 dim 40 40 80 1, data_format=0, name=uid_8_out_0ub_uid_1_out_0, none-quant +output 4 dim 20 20 64 1, data_format=0, name=uid_7_out_0ub_uid_1_out_0, none-quant +output 5 dim 20 20 80 1, data_format=0, name=uid_6_out_0ub_uid_1_out_0, none-quant +nbg name=model/yolov8n_6_uint8_a733.nb, size: 2452448. +create network 0: 11517 us. +prepare network: 1821 us. +buffer ptr: 0x10cdb600, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 12567 us. +output 0, ptr 0x10e07740, size 409600. +output 1, ptr 0x10f977c0, size 512000. +output 2, ptr 0x1118b840, size 102400. +output 3, ptr 0x111ef8c0, size 128000. +output 4, ptr 0x1126c980, size 25600. +output 5, ptr 0x11285a00, size 32000. +detection num: 3 + 1: 87%, [ 130, 136, 568, 419], bicycle +16: 95%, [ 131, 220, 308, 541], dog + 2: 68%, [ 467, 74, 695, 171], car +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :------ | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner A733 | Vivante VIP9000 | yolov8n | 640×640 | 11.5 ms | 1.8 ms | 12.6 ms | | 25.9 ms | 38.6 FPS | + + + + + + + +```bash +cd yolov8_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./yolov8_demo_t527 +./yolov8_demo_t527 -nb model/yolov8n_6_uint8_t527.nb -i model/dog.jpg +``` + + + +The running result is as follows: + +```bash +$ ./yolov8_demo_t527 -nb model/yolov8n_6_uint8_t527.nb -i model/dog.jpg +model_file=model/yolov8n_6_uint8_t527.nb, input=model/dog.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 80 80 64 1, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 80 80 80 1, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 40 40 64 1, data_format=0, name=uid_20003_sub_uid_1_out_0, none-quant +output 3 dim 40 40 80 1, data_format=0, name=uid_20004_sub_uid_1_out_0, none-quant +output 4 dim 20 20 64 1, data_format=0, name=uid_20005_sub_uid_1_out_0, none-quant +output 5 dim 20 20 80 1, data_format=0, name=uid_20006_sub_uid_1_out_0, none-quant +nbg name=model/yolov8n_6_uint8_t527.nb, size: 2915520. +create network 0: 16603 us. +prepare network: 5775 us. +buffer ptr: 0x3e36c600, buffer size: 1228800 +network: 0, loop count: 1 +run time for this network 0: 30967 us. +output 0, ptr 0x3e498700, size 409600. +output 1, ptr 0x3e628780, size 512000. +output 2, ptr 0x3e81c840, size 102400. +output 3, ptr 0x3e8808c0, size 128000. +output 4, ptr 0x3e8fd940, size 25600. +output 5, ptr 0x3e9169c0, size 32000. +detection num: 3 + 1: 88%, [ 130, 135, 568, 419], bicycle +16: 93%, [ 131, 219, 307, 540], dog + 2: 68%, [ 466, 74, 695, 171], car +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :------ | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner T527 | Vivante VIP9000 | yolov8n | 640×640 | 16.6 ms | 5.8 ms | 31.0 ms | | 53.4 ms | 18.7 FPS | + + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolox.mdx b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolox.mdx new file mode 100644 index 000000000..d26d99482 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/common/ai/cubie/_model-zoo-yolox.mdx @@ -0,0 +1,354 @@ +This document describes how to run YOLOX on NPU. + +:::info +Refer to [Model Zoo Download](./model-zoo-download) for the example. +::: + +YOLOX Example Directory Structure: + +```bash +$ tree ./ +./ +├── CMakeLists.txt +├── convert_model +│ ├── config_yml.py +│ ├── convert_model_env.sh +│ └── python +│ ├── coco_classes.py +│ ├── demo_utils.py +│ ├── sub_model.py +│ ├── visualize.py +│ └── yolox_sim.py +├── figures +│ ├── output_yolox.png +│ └── yolox_rt.png +├── main.cpp +├── model +│ └── bus.jpg +├── model_config.h +├── README.md +├── yolox_postprocess.cpp +└── yolox_preprocess.cpp +``` + +## Model Conversion + +### Download Model + + + +```bash +cd convert_model/ +wget https://github.com/Megvii-BaseDetection/YOLOX/releases/download/0.1.1rc0/yolox_s.onnx +``` + + + +Or download the modified model, click to download [yolox_s_sim.onnx](http://netstorage.allwinnertech.com:5000/sharing/G84PP2KvG). + +Then move to the convert_model/ directory. + +### Prune Model + +If you downloaded the already converted model, you can skip model pruning. + + + +```bash +cd python/ +python3 sub_model.py +cd ../ +``` + + + +### Create Symlink for Conversion Script + + + +```bash +./convert_model_env.sh +``` + + + +### Model Import/Quantization/Conversion + +You need to enter the container development environment first. Refer to the [Create Container](./model-zoo-download#创建并启动容器) section in Model Zoo Download. + +:::info +Different platforms use corresponding Docker images: + +- A733: ubuntu-npu:v2.0.10.1 +- T527: ubuntu-npu:v1.8.11 + ::: + + + +```bash +docker exec -it model-zoo /bin/bash +``` + + + +After entering the container, navigate to the corresponding directory and run the script. + + + +```bash +cd /workspace/examples/yolox/convert_model/ +``` + + + + + +```bash +./pegasus_import.sh yolox_s_sim +./pegasus_quantize.sh yolox_s_sim uint8 12 +``` + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolox_s_sim uint8 a733 +``` + + + + + + + + + +```bash +./pegasus_export_ovx_nbg.sh yolox_s_sim uint8 t527 +``` + + + + + + +The exported model files are stored in the ../model directory. + +### Compile Example + +Now you can compile the example. **First exit the container**, then execute the following command to compile the example. + +First, you need to configure third-party libraries and cross-compilation toolchain. + +:::info +You can skip this step if you have already configured third-party libraries and cross-compilation toolchain in other examples. +::: + + + +```bash +cd ../../../3rdparty/opencv/ +unzip opencv-4.9.0-aarch64-linux-sunxi-glibc.zip +cd ../../0-toolchains/ +``` + + + +You need to manually download via [this link](http://netstorage.allwinnertech.com:5000/sharing/e2nD8YwB4) first, then place it in 0-toolchains/ before executing the following command: + + + +```bash +tar -xvf gcc-arm-10.2-2020.11-x86_64-aarch64-none-linux-gnu.tar.xz +``` + + + + + +```bash +cd ../examples/yolox/ +``` + + + + + + + + +```bash +../build_linux.sh -t a733 -s debian11 +``` + + + + + + + + + +```bash +../build_linux.sh -t t527 -s debian11 +``` + + + + + + +## Model Deployment + +After compilation, the example will be installed in the install directory. You can use scp to transfer it to the board. + +### Configure NPU Driver + +:::info +You can skip this step if you have already configured NPU driver in other examples. +::: + +Transfer the driver library to the board's lib directory via scp. + +- A733 corresponds to the common/lib_linux_aarch64/A733 directory +- T527 corresponds to the common/lib_linux_aarch64/T527 directory + +Then execute the following command to export to environment variables. + + + +```bash +echo 'export LD_LIBRARY_PATH=$HOME/lib:$LD_LIBRARY_PATH' >> ~/.bashrc +``` + + + +### Run Example + +After configuring the driver, you can run the example. + +:::tip +For T527 platform, you need to first enable NPU by referring to the A5E's "Enable NPU on Board" documentation, then use the following command to grant the current user permission to use /dev/vipcore. +::: + + + +```bash +sudo chmod 777 /dev/vipcore +``` + + + + + + + + +```bash +cd yolox_demo_linux_a733/ +``` + + + + + +```bash +chmod +x ./yolox_demo_a733 +./yolox_demo_a733 -nb model/yolox_s_sim_uint8_a733.nb -i model/bus.jpg +``` + + + +The running result is as follows: + +```bash +$ ./yolox_demo_a733 -nb model/yolox_s_sim_uint8_a733.nb -i model/bus.jpg +model_file=model/yolox_s_sim_uint8_a733.nb, input=model/bus.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 2.0.3.2-AW-2024-08-30 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input/output[0], none-quant +output 0 dim 80 80 85 1, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 40 40 85 1, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 20 20 85 1, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +nbg name=model/yolox_s_sim_uint8_a733.nb, size: 7058512. +create network 0: 17292 us. +prepare network: 7783 us. +buffer ptr: 0x24831600, buffer size: 1228800 +Original image size: 640x640 +YOLOX preprocess completed: model/bus.jpg -> 640x640, buffer size: 1228800 +feed input cost: 11464 us. +network: 0, loop count: 1 +run time for this network 0: 30120 us. +detection num: 5 + 5: 93%, [ 85, 136, 555, 433], bus + 0: 89%, [ 113, 243, 199, 524], person + 0: 86%, [ 475, 239, 560, 520], person + 0: 89%, [ 213, 243, 283, 506], person + 0: 56%, [ 79, 328, 121, 515], person +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :------ | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner A733 | Vivante VIP9000 | yolox_s | 640×640 | 17.3 ms | 7.8 ms | 30.1 ms | | 55.2 ms | 18.1 FPS | + + + + + + + +```bash +cd yolox_demo_linux_t527/ +``` + + + + + +```bash +chmod +x ./yolox_demo_t527 +./yolox_demo_t527 -nb model/yolox_s_sim_uint8_t527.nb -i model/bus.jpg +``` + + + +The running result is as follows: + +```bash +$ ./yolox_demo_t527 -nb model/yolox_s_sim_uint8_t527.nb -i model/bus.jpg +model_file=model/yolox_s_sim_uint8_t527.nb, input=model/bus.jpg, loop_count=1, malloc_mbyte=10 +VIPLite driver software version 1.13.0.0-AW-2023-10-19 +input 0 dim 3 640 640 1, data_format=2, quant_format=0, name=input[0], none-quant +output 0 dim 80 80 85 1, data_format=0, name=uid_20000_sub_uid_1_out_0, none-quant +output 1 dim 40 40 85 1, data_format=0, name=uid_20001_sub_uid_1_out_0, none-quant +output 2 dim 20 20 85 1, data_format=0, name=uid_20002_sub_uid_1_out_0, none-quant +nbg name=model/yolox_s_sim_uint8_t527.nb, size: 9132672. +create network 0: 25385 us. +prepare network: 18164 us. +buffer ptr: 0x116c6600, buffer size: 1228800 +Original image size: 640x640 +YOLOX preprocess completed: model/bus.jpg -> 640x640, buffer size: 1228800 +feed input cost: 62033 us. +network: 0, loop count: 1 +run time for this network 0: 73697 us. +detection num: 5 + 5: 93%, [ 98, 137, 550, 435], bus + 0: 89%, [ 107, 239, 210, 533], person + 0: 87%, [ 477, 239, 560, 519], person + 0: 89%, [ 214, 243, 283, 506], person + 0: 58%, [ 79, 329, 120, 516], person +destroy npu finished. +~NpuUint. +``` + +| SoC | NPU | Model | Input Resolution | Network Creation Time | Network Preparation Time | Single Frame Inference Time | Post-processing Time | Total Time | Frame Rate | +| :------------- | :-------------- | :------ | :--------------- | :-------------------- | :----------------------- | :-------------------------- | :------------------- | :--------- | :--------- | +| Allwinner T527 | Vivante VIP9000 | yolox_s | 640×640 | 25.4 ms | 18.2 ms | 73.7 ms | | 117.3 ms | 8.5 FPS | + + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/cubie-lenet.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/_cubie-lenet.md similarity index 100% rename from i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/cubie-lenet.md rename to i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/_cubie-lenet.md diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/cubie-resnet50.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/_cubie-resnet50.md similarity index 100% rename from i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/cubie-resnet50.md rename to i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/_cubie-resnet50.md diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/cubie-yolact.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/_cubie-yolact.md similarity index 100% rename from i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/cubie-yolact.md rename to i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/_cubie-yolact.md diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/cubie-yolov5.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/_cubie-yolov5.md similarity index 100% rename from i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/cubie-yolov5.md rename to i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/_cubie-yolov5.md diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/README.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/README.md new file mode 100644 index 000000000..3c0300bea --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/README.md @@ -0,0 +1,9 @@ +--- +sidebar_position: 11 +--- + +# Model Zoo + +Here are some pre-prepared model deployment cases for your reference. + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/_LPRNet.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/_LPRNet.md new file mode 100644 index 000000000..1a2a23cdd --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/_LPRNet.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 25 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-LPRNet.mdx +--- + +# LPRNet + +import LPRNet from '../../../../../common/ai/cubie/\_model-zoo-LPRNet.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/_ppocr.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/_ppocr.md new file mode 100644 index 000000000..8b4a3a85e --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/_ppocr.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 16 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-ppocr.mdx +--- + +# PPOCR + +import PPOCR from '../../../../../common/ai/cubie/\_model-zoo-ppocr.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/_yolo26-pose.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/_yolo26-pose.md new file mode 100644 index 000000000..e7d477529 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/_yolo26-pose.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 13 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo26-pose.mdx +--- + +# YOLO26 Pose + +import YOLO26Pose from '../../../../../common/ai/cubie/\_model-zoo-yolo26-pose.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/_yolov8-hybrid.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/_yolov8-hybrid.md new file mode 100644 index 000000000..d06d5de33 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/_yolov8-hybrid.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 2 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8-hybrid.mdx +--- + +# YOLOv8 Hybrid + +import YOLOv8Hybrid from '../../../../../common/ai/cubie/\_model-zoo-yolov8-hybrid.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/densenet121-keras.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/densenet121-keras.md new file mode 100644 index 000000000..ce02f3a5c --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/densenet121-keras.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 22 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-densenet121-keras.mdx +--- + +# DenseNet121 + +import DenseNet121 from '../../../../../common/ai/cubie/\_model-zoo-densenet121-keras.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/lenet-caffe.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/lenet-caffe.md new file mode 100644 index 000000000..da2ea0ff6 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/lenet-caffe.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 24 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-lenet-caffe.mdx +--- + +# LeNet + +import LeNet from '../../../../../common/ai/cubie/\_model-zoo-lenet-caffe.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/mobilenetv1-tensorflow.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/mobilenetv1-tensorflow.md new file mode 100644 index 000000000..a552c39b2 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/mobilenetv1-tensorflow.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 18 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-mobilenetv1-tensorflow.mdx +--- + +# MobileNetV1 + +import MobileNetV1 from '../../../../../common/ai/cubie/\_model-zoo-mobilenetv1-tensorflow.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/mobilenetv2.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/mobilenetv2.md new file mode 100644 index 000000000..51ef4f03f --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/mobilenetv2.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 19 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-mobilenetv2.mdx +--- + +# MobileNetV2 + +import MobileNetV2 from '../../../../../common/ai/cubie/\_model-zoo-mobilenetv2.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/model-zoo-download.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/model-zoo-download.md new file mode 100644 index 000000000..8ce08e440 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/model-zoo-download.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 1 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-download.mdx +--- + +# Model Zoo Download + +import ModelZooDownload from '../../../../../common/ai/cubie/\_model-zoo-download.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/ppseg.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/ppseg.md new file mode 100644 index 000000000..fe48c4427 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/ppseg.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 17 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-ppseg.mdx +--- + +# PPSeg + +import PPSeg from '../../../../../common/ai/cubie/\_model-zoo-ppseg.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/resnet50-tflite.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/resnet50-tflite.md new file mode 100644 index 000000000..475aa0858 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/resnet50-tflite.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 20 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-resnet50-tflite.mdx +--- + +# ResNet50 TFLite + +import ResNet50TFLite from '../../../../../common/ai/cubie/\_model-zoo-resnet50-tflite.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/resnet50v2.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/resnet50v2.md new file mode 100644 index 000000000..274417b93 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/resnet50v2.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 21 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-resnet50v2.mdx +--- + +# ResNet50 V2 + +import ResNet50V2 from '../../../../../common/ai/cubie/\_model-zoo-resnet50v2.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/retinaface.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/retinaface.md new file mode 100644 index 000000000..ee3da187a --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/retinaface.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 15 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-retinaface.mdx +--- + +# RetinaFace + +import RetinaFace from '../../../../../common/ai/cubie/\_model-zoo-retinaface.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/squeezenet-pytorch.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/squeezenet-pytorch.md new file mode 100644 index 000000000..690bafe69 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/squeezenet-pytorch.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 23 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-squeezenet-pytorch.mdx +--- + +# SqueezeNet + +import SqueezeNet from '../../../../../common/ai/cubie/\_model-zoo-squeezenet-pytorch.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolo11-pose.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolo11-pose.md new file mode 100644 index 000000000..f1255f584 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolo11-pose.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 5 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo11-pose.mdx +--- + +# YOLO11 Pose + +import YOLO11Pose from '../../../../../common/ai/cubie/\_model-zoo-yolo11-pose.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolo11-seg.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolo11-seg.md new file mode 100644 index 000000000..672afce80 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolo11-seg.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 4 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo11-seg.mdx +--- + +# YOLO11 Seg + +import YOLO11Seg from '../../../../../common/ai/cubie/\_model-zoo-yolo11-seg.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolo11.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolo11.md new file mode 100644 index 000000000..300d57295 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolo11.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 3 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo11.mdx +--- + +# YOLO11 + +import YOLO11 from '../../../../../common/ai/cubie/\_model-zoo-yolo11.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolo26.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolo26.md new file mode 100644 index 000000000..f9987edc1 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolo26.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 12 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo26.mdx +--- + +# YOLO26 + +import YOLO26 from '../../../../../common/ai/cubie/\_model-zoo-yolo26.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolov3-darknet.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolov3-darknet.md new file mode 100644 index 000000000..7d191f006 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolov3-darknet.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 9 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov3-darknet.mdx +--- + +# YOLOv3 + +import YOLOv3 from '../../../../../common/ai/cubie/\_model-zoo-yolov3-darknet.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolov5.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolov5.md new file mode 100644 index 000000000..cd33bd150 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolov5.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 10 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov5.mdx +--- + +# YOLOv5 + +import YOLOv5 from '../../../../../common/ai/cubie/\_model-zoo-yolov5.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolov8-pose.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolov8-pose.md new file mode 100644 index 000000000..d29446157 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolov8-pose.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 8 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8-pose.mdx +--- + +# YOLOv8 Pose + +import YOLOv8Pose from '../../../../../common/ai/cubie/\_model-zoo-yolov8-pose.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolov8-seg.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolov8-seg.md new file mode 100644 index 000000000..451da2b12 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolov8-seg.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 7 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8-seg.mdx +--- + +# YOLOv8 Seg + +import YOLOv8Seg from '../../../../../common/ai/cubie/\_model-zoo-yolov8-seg.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolov8.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolov8.md new file mode 100644 index 000000000..1eb3ebbd7 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolov8.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 6 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8.mdx +--- + +# YOLOv8 + +import YOLOv8 from '../../../../../common/ai/cubie/\_model-zoo-yolov8.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolox.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolox.md new file mode 100644 index 000000000..ed26d8739 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a5e/app-dev/npu-dev/model-zoo/yolox.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 14 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolox.mdx +--- + +# YOLOX + +import YOLOX from '../../../../../common/ai/cubie/\_model-zoo-yolox.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/cubie-lenet.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/_cubie-lenet.md similarity index 100% rename from i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/cubie-lenet.md rename to i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/_cubie-lenet.md diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/cubie-resnet50.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/_cubie-resnet50.md similarity index 100% rename from i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/cubie-resnet50.md rename to i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/_cubie-resnet50.md diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/cubie-yolact.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/_cubie-yolact.md similarity index 100% rename from i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/cubie-yolact.md rename to i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/_cubie-yolact.md diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/cubie-yolov5.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/_cubie-yolov5.md similarity index 100% rename from i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/cubie-yolov5.md rename to i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/_cubie-yolov5.md diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/README.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/README.md new file mode 100644 index 000000000..3c0300bea --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/README.md @@ -0,0 +1,9 @@ +--- +sidebar_position: 11 +--- + +# Model Zoo + +Here are some pre-prepared model deployment cases for your reference. + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/_LPRNet.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/_LPRNet.md new file mode 100644 index 000000000..1a2a23cdd --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/_LPRNet.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 25 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-LPRNet.mdx +--- + +# LPRNet + +import LPRNet from '../../../../../common/ai/cubie/\_model-zoo-LPRNet.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/_ppocr.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/_ppocr.md new file mode 100644 index 000000000..8b4a3a85e --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/_ppocr.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 16 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-ppocr.mdx +--- + +# PPOCR + +import PPOCR from '../../../../../common/ai/cubie/\_model-zoo-ppocr.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/_yolo26-pose.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/_yolo26-pose.md new file mode 100644 index 000000000..e7d477529 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/_yolo26-pose.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 13 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo26-pose.mdx +--- + +# YOLO26 Pose + +import YOLO26Pose from '../../../../../common/ai/cubie/\_model-zoo-yolo26-pose.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/_yolov8-hybrid.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/_yolov8-hybrid.md new file mode 100644 index 000000000..d06d5de33 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/_yolov8-hybrid.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 2 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8-hybrid.mdx +--- + +# YOLOv8 Hybrid + +import YOLOv8Hybrid from '../../../../../common/ai/cubie/\_model-zoo-yolov8-hybrid.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/densenet121-keras.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/densenet121-keras.md new file mode 100644 index 000000000..ce02f3a5c --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/densenet121-keras.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 22 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-densenet121-keras.mdx +--- + +# DenseNet121 + +import DenseNet121 from '../../../../../common/ai/cubie/\_model-zoo-densenet121-keras.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/lenet-caffe.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/lenet-caffe.md new file mode 100644 index 000000000..da2ea0ff6 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/lenet-caffe.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 24 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-lenet-caffe.mdx +--- + +# LeNet + +import LeNet from '../../../../../common/ai/cubie/\_model-zoo-lenet-caffe.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/mobilenetv1-tensorflow.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/mobilenetv1-tensorflow.md new file mode 100644 index 000000000..a552c39b2 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/mobilenetv1-tensorflow.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 18 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-mobilenetv1-tensorflow.mdx +--- + +# MobileNetV1 + +import MobileNetV1 from '../../../../../common/ai/cubie/\_model-zoo-mobilenetv1-tensorflow.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/mobilenetv2.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/mobilenetv2.md new file mode 100644 index 000000000..51ef4f03f --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/mobilenetv2.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 19 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-mobilenetv2.mdx +--- + +# MobileNetV2 + +import MobileNetV2 from '../../../../../common/ai/cubie/\_model-zoo-mobilenetv2.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/model-zoo-download.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/model-zoo-download.md new file mode 100644 index 000000000..8ce08e440 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/model-zoo-download.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 1 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-download.mdx +--- + +# Model Zoo Download + +import ModelZooDownload from '../../../../../common/ai/cubie/\_model-zoo-download.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/ppseg.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/ppseg.md new file mode 100644 index 000000000..fe48c4427 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/ppseg.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 17 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-ppseg.mdx +--- + +# PPSeg + +import PPSeg from '../../../../../common/ai/cubie/\_model-zoo-ppseg.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/resnet50-tflite.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/resnet50-tflite.md new file mode 100644 index 000000000..475aa0858 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/resnet50-tflite.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 20 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-resnet50-tflite.mdx +--- + +# ResNet50 TFLite + +import ResNet50TFLite from '../../../../../common/ai/cubie/\_model-zoo-resnet50-tflite.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/resnet50v2.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/resnet50v2.md new file mode 100644 index 000000000..274417b93 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/resnet50v2.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 21 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-resnet50v2.mdx +--- + +# ResNet50 V2 + +import ResNet50V2 from '../../../../../common/ai/cubie/\_model-zoo-resnet50v2.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/retinaface.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/retinaface.md new file mode 100644 index 000000000..ee3da187a --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/retinaface.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 15 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-retinaface.mdx +--- + +# RetinaFace + +import RetinaFace from '../../../../../common/ai/cubie/\_model-zoo-retinaface.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/squeezenet-pytorch.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/squeezenet-pytorch.md new file mode 100644 index 000000000..690bafe69 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/squeezenet-pytorch.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 23 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-squeezenet-pytorch.mdx +--- + +# SqueezeNet + +import SqueezeNet from '../../../../../common/ai/cubie/\_model-zoo-squeezenet-pytorch.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolo11-pose.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolo11-pose.md new file mode 100644 index 000000000..f1255f584 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolo11-pose.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 5 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo11-pose.mdx +--- + +# YOLO11 Pose + +import YOLO11Pose from '../../../../../common/ai/cubie/\_model-zoo-yolo11-pose.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolo11-seg.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolo11-seg.md new file mode 100644 index 000000000..672afce80 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolo11-seg.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 4 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo11-seg.mdx +--- + +# YOLO11 Seg + +import YOLO11Seg from '../../../../../common/ai/cubie/\_model-zoo-yolo11-seg.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolo11.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolo11.md new file mode 100644 index 000000000..300d57295 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolo11.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 3 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo11.mdx +--- + +# YOLO11 + +import YOLO11 from '../../../../../common/ai/cubie/\_model-zoo-yolo11.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolo26.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolo26.md new file mode 100644 index 000000000..f9987edc1 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolo26.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 12 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo26.mdx +--- + +# YOLO26 + +import YOLO26 from '../../../../../common/ai/cubie/\_model-zoo-yolo26.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolov3-darknet.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolov3-darknet.md new file mode 100644 index 000000000..7d191f006 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolov3-darknet.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 9 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov3-darknet.mdx +--- + +# YOLOv3 + +import YOLOv3 from '../../../../../common/ai/cubie/\_model-zoo-yolov3-darknet.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolov5.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolov5.md new file mode 100644 index 000000000..cd33bd150 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolov5.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 10 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov5.mdx +--- + +# YOLOv5 + +import YOLOv5 from '../../../../../common/ai/cubie/\_model-zoo-yolov5.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolov8-pose.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolov8-pose.md new file mode 100644 index 000000000..d29446157 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolov8-pose.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 8 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8-pose.mdx +--- + +# YOLOv8 Pose + +import YOLOv8Pose from '../../../../../common/ai/cubie/\_model-zoo-yolov8-pose.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolov8-seg.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolov8-seg.md new file mode 100644 index 000000000..451da2b12 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolov8-seg.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 7 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8-seg.mdx +--- + +# YOLOv8 Seg + +import YOLOv8Seg from '../../../../../common/ai/cubie/\_model-zoo-yolov8-seg.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolov8.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolov8.md new file mode 100644 index 000000000..1eb3ebbd7 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolov8.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 6 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8.mdx +--- + +# YOLOv8 + +import YOLOv8 from '../../../../../common/ai/cubie/\_model-zoo-yolov8.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolox.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolox.md new file mode 100644 index 000000000..ed26d8739 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7a/app-dev/npu-dev/model-zoo/yolox.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 14 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolox.mdx +--- + +# YOLOX + +import YOLOX from '../../../../../common/ai/cubie/\_model-zoo-yolox.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/cuibie-lenet.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/_cubie-lenet.md similarity index 100% rename from i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/cuibie-lenet.md rename to i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/_cubie-lenet.md diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/cubie-resnet50.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/_cubie-resnet50.md similarity index 100% rename from i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/cubie-resnet50.md rename to i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/_cubie-resnet50.md diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/cubie-yolact.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/_cubie-yolact.md similarity index 100% rename from i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/cubie-yolact.md rename to i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/_cubie-yolact.md diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/cubie-yolov5.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/_cubie-yolov5.md similarity index 100% rename from i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/cubie-yolov5.md rename to i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/_cubie-yolov5.md diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/cubie-acuity-env.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/cubie-acuity-env.md index 54d4059f6..624ca7c1c 100644 --- a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/cubie-acuity-env.md +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/cubie-acuity-env.md @@ -1,5 +1,5 @@ --- -sidebar_position: 1 +sidebar_position: 2 doc_kind: wrapper source_of_truth: common diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/README.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/README.md new file mode 100644 index 000000000..3c0300bea --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/README.md @@ -0,0 +1,9 @@ +--- +sidebar_position: 11 +--- + +# Model Zoo + +Here are some pre-prepared model deployment cases for your reference. + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/_LPRNet.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/_LPRNet.md new file mode 100644 index 000000000..1a2a23cdd --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/_LPRNet.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 25 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-LPRNet.mdx +--- + +# LPRNet + +import LPRNet from '../../../../../common/ai/cubie/\_model-zoo-LPRNet.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/_ppocr.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/_ppocr.md new file mode 100644 index 000000000..8b4a3a85e --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/_ppocr.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 16 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-ppocr.mdx +--- + +# PPOCR + +import PPOCR from '../../../../../common/ai/cubie/\_model-zoo-ppocr.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/_yolo26-pose.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/_yolo26-pose.md new file mode 100644 index 000000000..e7d477529 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/_yolo26-pose.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 13 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo26-pose.mdx +--- + +# YOLO26 Pose + +import YOLO26Pose from '../../../../../common/ai/cubie/\_model-zoo-yolo26-pose.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/_yolov8-hybrid.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/_yolov8-hybrid.md new file mode 100644 index 000000000..d06d5de33 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/_yolov8-hybrid.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 2 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8-hybrid.mdx +--- + +# YOLOv8 Hybrid + +import YOLOv8Hybrid from '../../../../../common/ai/cubie/\_model-zoo-yolov8-hybrid.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/densenet121-keras.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/densenet121-keras.md new file mode 100644 index 000000000..ce02f3a5c --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/densenet121-keras.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 22 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-densenet121-keras.mdx +--- + +# DenseNet121 + +import DenseNet121 from '../../../../../common/ai/cubie/\_model-zoo-densenet121-keras.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/lenet-caffe.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/lenet-caffe.md new file mode 100644 index 000000000..da2ea0ff6 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/lenet-caffe.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 24 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-lenet-caffe.mdx +--- + +# LeNet + +import LeNet from '../../../../../common/ai/cubie/\_model-zoo-lenet-caffe.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/mobilenetv1-tensorflow.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/mobilenetv1-tensorflow.md new file mode 100644 index 000000000..a552c39b2 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/mobilenetv1-tensorflow.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 18 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-mobilenetv1-tensorflow.mdx +--- + +# MobileNetV1 + +import MobileNetV1 from '../../../../../common/ai/cubie/\_model-zoo-mobilenetv1-tensorflow.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/mobilenetv2.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/mobilenetv2.md new file mode 100644 index 000000000..51ef4f03f --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/mobilenetv2.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 19 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-mobilenetv2.mdx +--- + +# MobileNetV2 + +import MobileNetV2 from '../../../../../common/ai/cubie/\_model-zoo-mobilenetv2.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/model-zoo-download.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/model-zoo-download.md new file mode 100644 index 000000000..8ce08e440 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/model-zoo-download.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 1 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-download.mdx +--- + +# Model Zoo Download + +import ModelZooDownload from '../../../../../common/ai/cubie/\_model-zoo-download.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/ppseg.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/ppseg.md new file mode 100644 index 000000000..fe48c4427 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/ppseg.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 17 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-ppseg.mdx +--- + +# PPSeg + +import PPSeg from '../../../../../common/ai/cubie/\_model-zoo-ppseg.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/resnet50-tflite.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/resnet50-tflite.md new file mode 100644 index 000000000..475aa0858 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/resnet50-tflite.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 20 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-resnet50-tflite.mdx +--- + +# ResNet50 TFLite + +import ResNet50TFLite from '../../../../../common/ai/cubie/\_model-zoo-resnet50-tflite.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/resnet50v2.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/resnet50v2.md new file mode 100644 index 000000000..274417b93 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/resnet50v2.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 21 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-resnet50v2.mdx +--- + +# ResNet50 V2 + +import ResNet50V2 from '../../../../../common/ai/cubie/\_model-zoo-resnet50v2.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/retinaface.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/retinaface.md new file mode 100644 index 000000000..ee3da187a --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/retinaface.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 15 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-retinaface.mdx +--- + +# RetinaFace + +import RetinaFace from '../../../../../common/ai/cubie/\_model-zoo-retinaface.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/squeezenet-pytorch.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/squeezenet-pytorch.md new file mode 100644 index 000000000..690bafe69 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/squeezenet-pytorch.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 23 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-squeezenet-pytorch.mdx +--- + +# SqueezeNet + +import SqueezeNet from '../../../../../common/ai/cubie/\_model-zoo-squeezenet-pytorch.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolo11-pose.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolo11-pose.md new file mode 100644 index 000000000..f1255f584 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolo11-pose.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 5 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo11-pose.mdx +--- + +# YOLO11 Pose + +import YOLO11Pose from '../../../../../common/ai/cubie/\_model-zoo-yolo11-pose.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolo11-seg.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolo11-seg.md new file mode 100644 index 000000000..672afce80 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolo11-seg.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 4 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo11-seg.mdx +--- + +# YOLO11 Seg + +import YOLO11Seg from '../../../../../common/ai/cubie/\_model-zoo-yolo11-seg.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolo11.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolo11.md new file mode 100644 index 000000000..300d57295 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolo11.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 3 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo11.mdx +--- + +# YOLO11 + +import YOLO11 from '../../../../../common/ai/cubie/\_model-zoo-yolo11.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolo26.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolo26.md new file mode 100644 index 000000000..f9987edc1 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolo26.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 12 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo26.mdx +--- + +# YOLO26 + +import YOLO26 from '../../../../../common/ai/cubie/\_model-zoo-yolo26.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolov3-darknet.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolov3-darknet.md new file mode 100644 index 000000000..7d191f006 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolov3-darknet.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 9 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov3-darknet.mdx +--- + +# YOLOv3 + +import YOLOv3 from '../../../../../common/ai/cubie/\_model-zoo-yolov3-darknet.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolov5.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolov5.md new file mode 100644 index 000000000..cd33bd150 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolov5.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 10 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov5.mdx +--- + +# YOLOv5 + +import YOLOv5 from '../../../../../common/ai/cubie/\_model-zoo-yolov5.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolov8-pose.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolov8-pose.md new file mode 100644 index 000000000..d29446157 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolov8-pose.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 8 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8-pose.mdx +--- + +# YOLOv8 Pose + +import YOLOv8Pose from '../../../../../common/ai/cubie/\_model-zoo-yolov8-pose.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolov8-seg.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolov8-seg.md new file mode 100644 index 000000000..451da2b12 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolov8-seg.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 7 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8-seg.mdx +--- + +# YOLOv8 Seg + +import YOLOv8Seg from '../../../../../common/ai/cubie/\_model-zoo-yolov8-seg.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolov8.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolov8.md new file mode 100644 index 000000000..1eb3ebbd7 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolov8.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 6 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8.mdx +--- + +# YOLOv8 + +import YOLOv8 from '../../../../../common/ai/cubie/\_model-zoo-yolov8.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolox.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolox.md new file mode 100644 index 000000000..ed26d8739 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7s/app-dev/npu-dev/model-zoo/yolox.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 14 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolox.mdx +--- + +# YOLOX + +import YOLOX from '../../../../../common/ai/cubie/\_model-zoo-yolox.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/cubie-lenet.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/_cubie-lenet.md similarity index 100% rename from i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/cubie-lenet.md rename to i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/_cubie-lenet.md diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/cubie-resnet50.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/_cubie-resnet50.md similarity index 100% rename from i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/cubie-resnet50.md rename to i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/_cubie-resnet50.md diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/cubie-yolact.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/_cubie-yolact.md similarity index 100% rename from i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/cubie-yolact.md rename to i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/_cubie-yolact.md diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/cubie-yolov5.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/_cubie-yolov5.md similarity index 100% rename from i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/cubie-yolov5.md rename to i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/_cubie-yolov5.md diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/README.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/README.md new file mode 100644 index 000000000..3c0300bea --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/README.md @@ -0,0 +1,9 @@ +--- +sidebar_position: 11 +--- + +# Model Zoo + +Here are some pre-prepared model deployment cases for your reference. + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/_LPRNet.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/_LPRNet.md new file mode 100644 index 000000000..1a2a23cdd --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/_LPRNet.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 25 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-LPRNet.mdx +--- + +# LPRNet + +import LPRNet from '../../../../../common/ai/cubie/\_model-zoo-LPRNet.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/_ppocr.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/_ppocr.md new file mode 100644 index 000000000..8b4a3a85e --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/_ppocr.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 16 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-ppocr.mdx +--- + +# PPOCR + +import PPOCR from '../../../../../common/ai/cubie/\_model-zoo-ppocr.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/_yolo26-pose.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/_yolo26-pose.md new file mode 100644 index 000000000..e7d477529 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/_yolo26-pose.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 13 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo26-pose.mdx +--- + +# YOLO26 Pose + +import YOLO26Pose from '../../../../../common/ai/cubie/\_model-zoo-yolo26-pose.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/_yolov8-hybrid.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/_yolov8-hybrid.md new file mode 100644 index 000000000..d06d5de33 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/_yolov8-hybrid.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 2 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8-hybrid.mdx +--- + +# YOLOv8 Hybrid + +import YOLOv8Hybrid from '../../../../../common/ai/cubie/\_model-zoo-yolov8-hybrid.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/densenet121-keras.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/densenet121-keras.md new file mode 100644 index 000000000..ce02f3a5c --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/densenet121-keras.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 22 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-densenet121-keras.mdx +--- + +# DenseNet121 + +import DenseNet121 from '../../../../../common/ai/cubie/\_model-zoo-densenet121-keras.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/lenet-caffe.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/lenet-caffe.md new file mode 100644 index 000000000..da2ea0ff6 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/lenet-caffe.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 24 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-lenet-caffe.mdx +--- + +# LeNet + +import LeNet from '../../../../../common/ai/cubie/\_model-zoo-lenet-caffe.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/mobilenetv1-tensorflow.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/mobilenetv1-tensorflow.md new file mode 100644 index 000000000..a552c39b2 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/mobilenetv1-tensorflow.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 18 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-mobilenetv1-tensorflow.mdx +--- + +# MobileNetV1 + +import MobileNetV1 from '../../../../../common/ai/cubie/\_model-zoo-mobilenetv1-tensorflow.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/mobilenetv2.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/mobilenetv2.md new file mode 100644 index 000000000..51ef4f03f --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/mobilenetv2.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 19 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-mobilenetv2.mdx +--- + +# MobileNetV2 + +import MobileNetV2 from '../../../../../common/ai/cubie/\_model-zoo-mobilenetv2.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/model-zoo-download.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/model-zoo-download.md new file mode 100644 index 000000000..8ce08e440 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/model-zoo-download.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 1 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-download.mdx +--- + +# Model Zoo Download + +import ModelZooDownload from '../../../../../common/ai/cubie/\_model-zoo-download.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/ppseg.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/ppseg.md new file mode 100644 index 000000000..fe48c4427 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/ppseg.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 17 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-ppseg.mdx +--- + +# PPSeg + +import PPSeg from '../../../../../common/ai/cubie/\_model-zoo-ppseg.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/resnet50-tflite.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/resnet50-tflite.md new file mode 100644 index 000000000..475aa0858 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/resnet50-tflite.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 20 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-resnet50-tflite.mdx +--- + +# ResNet50 TFLite + +import ResNet50TFLite from '../../../../../common/ai/cubie/\_model-zoo-resnet50-tflite.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/resnet50v2.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/resnet50v2.md new file mode 100644 index 000000000..274417b93 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/resnet50v2.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 21 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-resnet50v2.mdx +--- + +# ResNet50 V2 + +import ResNet50V2 from '../../../../../common/ai/cubie/\_model-zoo-resnet50v2.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/retinaface.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/retinaface.md new file mode 100644 index 000000000..ee3da187a --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/retinaface.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 15 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-retinaface.mdx +--- + +# RetinaFace + +import RetinaFace from '../../../../../common/ai/cubie/\_model-zoo-retinaface.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/squeezenet-pytorch.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/squeezenet-pytorch.md new file mode 100644 index 000000000..690bafe69 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/squeezenet-pytorch.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 23 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-squeezenet-pytorch.mdx +--- + +# SqueezeNet + +import SqueezeNet from '../../../../../common/ai/cubie/\_model-zoo-squeezenet-pytorch.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolo11-pose.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolo11-pose.md new file mode 100644 index 000000000..f1255f584 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolo11-pose.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 5 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo11-pose.mdx +--- + +# YOLO11 Pose + +import YOLO11Pose from '../../../../../common/ai/cubie/\_model-zoo-yolo11-pose.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolo11-seg.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolo11-seg.md new file mode 100644 index 000000000..672afce80 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolo11-seg.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 4 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo11-seg.mdx +--- + +# YOLO11 Seg + +import YOLO11Seg from '../../../../../common/ai/cubie/\_model-zoo-yolo11-seg.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolo11.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolo11.md new file mode 100644 index 000000000..300d57295 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolo11.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 3 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo11.mdx +--- + +# YOLO11 + +import YOLO11 from '../../../../../common/ai/cubie/\_model-zoo-yolo11.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolo26.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolo26.md new file mode 100644 index 000000000..f9987edc1 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolo26.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 12 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolo26.mdx +--- + +# YOLO26 + +import YOLO26 from '../../../../../common/ai/cubie/\_model-zoo-yolo26.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolov3-darknet.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolov3-darknet.md new file mode 100644 index 000000000..7d191f006 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolov3-darknet.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 9 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov3-darknet.mdx +--- + +# YOLOv3 + +import YOLOv3 from '../../../../../common/ai/cubie/\_model-zoo-yolov3-darknet.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolov5.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolov5.md new file mode 100644 index 000000000..cd33bd150 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolov5.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 10 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov5.mdx +--- + +# YOLOv5 + +import YOLOv5 from '../../../../../common/ai/cubie/\_model-zoo-yolov5.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolov8-pose.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolov8-pose.md new file mode 100644 index 000000000..d29446157 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolov8-pose.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 8 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8-pose.mdx +--- + +# YOLOv8 Pose + +import YOLOv8Pose from '../../../../../common/ai/cubie/\_model-zoo-yolov8-pose.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolov8-seg.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolov8-seg.md new file mode 100644 index 000000000..451da2b12 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolov8-seg.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 7 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8-seg.mdx +--- + +# YOLOv8 Seg + +import YOLOv8Seg from '../../../../../common/ai/cubie/\_model-zoo-yolov8-seg.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolov8.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolov8.md new file mode 100644 index 000000000..1eb3ebbd7 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolov8.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 6 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolov8.mdx +--- + +# YOLOv8 + +import YOLOv8 from '../../../../../common/ai/cubie/\_model-zoo-yolov8.mdx'; + + diff --git a/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolox.md b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolox.md new file mode 100644 index 000000000..ed26d8739 --- /dev/null +++ b/i18n/en/docusaurus-plugin-content-docs/current/cubie/a7z/app-dev/npu-dev/model-zoo/yolox.md @@ -0,0 +1,14 @@ +--- +sidebar_position: 14 + +doc_kind: wrapper +source_of_truth: common +imports_resolve_to: + - docs/common/ai/cubie/_model-zoo-yolox.mdx +--- + +# YOLOX + +import YOLOX from '../../../../../common/ai/cubie/\_model-zoo-yolox.mdx'; + + diff --git a/static/img/cubie/allwinnertech-login.webp b/static/img/cubie/allwinnertech-login.webp new file mode 100644 index 000000000..f92ebee5c Binary files /dev/null and b/static/img/cubie/allwinnertech-login.webp differ diff --git a/static/img/cubie/allwinnertech-model-zoo-download-guide1.webp b/static/img/cubie/allwinnertech-model-zoo-download-guide1.webp new file mode 100644 index 000000000..e3b3ed3c4 Binary files /dev/null and b/static/img/cubie/allwinnertech-model-zoo-download-guide1.webp differ diff --git a/static/img/cubie/allwinnertech-model-zoo-download-guide2.webp b/static/img/cubie/allwinnertech-model-zoo-download-guide2.webp new file mode 100644 index 000000000..a3412d63e Binary files /dev/null and b/static/img/cubie/allwinnertech-model-zoo-download-guide2.webp differ