Skip to content

"We employ the SSD object detection algorithm to determine whether the driver is drowsy (by analyzing eye closure or mouth opening) or engaged in distracting behaviors like smoking or using a phone. Additionally, the PERCLOS metric is used to assess driver fatigue.

License

Notifications You must be signed in to change notification settings

yejun688/Dangerous_driving_behavior_detection

Repository files navigation

运行环境:

1.python 3.7.4 2.pytorch 1.4.0 3.python-opencv

说明 预训练的权重文件[vgg_16] 具体的配置文件请看Config.py文件 训练运行python Train.py 单张测试 python Test.py

##目前进度: 1、PERCLOS计算 DONE 2、眨眼频率计算 DONE 3、打哈欠检测及计算 DONE 4、疲劳检测 DONE 5、人脸情绪检测 DONE 6、口罩检测Done

网络检测性能:准确率82.18%

主要文件说明: ssd_net_vgg.py 定义class SSD的文件 Train.py 训练代码 voc0712.py 数据集处理代码(没有改文件名,改的话还要改其他代码) loss_function.py 损失函数 detection.py 检测结果的处理代码,将SSD返回结果处理为opencv可以处理的形式 test.py 单张图片测试代码 Ps:没写参数接口,所以要改测试的图片就要手动改代码内部文件名了 l2norm.py l2正则化 Config.py 配置参数 utils.py 工具类 camera_detection.py 摄像头检测代码V1,V2 augmentations.py 生成识别框 Run.py 主程序运行文件 MainWindow.py UI界面布局 /Emoji 人脸情绪识别部分 /mtcnn 人脸定位检测 /test 测试视频、图片素材 /Facemask口罩检测相关结构

数据集结构(没有给出,需要的话咨询[email protected]或者[email protected]): /dataset: /Annotations 存放含有目标信息的xml文件 /ImageSets/Main 存放图片名的文件 /JPEGImages 存放图片 /txt.py 生成ImageSets文件的代码

权重文件存放路径: /weights

带训练和测试代码版本请看https://github.com/DohaerisT/DangerousDrivingDetector 参考资料: https://blog.csdn.net/zxd52csx/article/details/82795104

About

"We employ the SSD object detection algorithm to determine whether the driver is drowsy (by analyzing eye closure or mouth opening) or engaged in distracting behaviors like smoking or using a phone. Additionally, the PERCLOS metric is used to assess driver fatigue.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages