MeshNet: Mesh Neural Network for 3D Shape Representation

This repository is not an original official implementation of the work, but a refactored codebase of https://github.com/iMoonLab/MeshNet. Performed within the FSE coursework at Skoltech.

Installation

First, you have to clone repository:

git clone https://github.com/KotikNikita/MeshNet

Run docker, you have 2 options:

1. From DockerHub:

docker pull poliik/meshnet_docker:latest
docker run -it poliik/meshnet_docker:latest

2. Build image locally:

docker build -t meshnet_docker -f Dockerfile .
docker run -it meshnet_docker

To run with GPU:

1. Install nvidia-container-toolkit.
2. Specify --gpus all flag when running docker run ....
3. Set 'cuda_devices' to '0' in config/train_config.yaml and config/test_config.yaml.

Usage

Data Preparation

First, you should download the reorganized ModelNet40 dataset. To download train and test dataset, run the following commnd

bash download.sh

To train model, run:

bash train.sh

To test model, run:

bash test.sh

To run unit and integration tests, run:

bash tests.sh

(some tests require a dataset to pass)

Modifications

For each data file XXX.off in ModelNet, we reorganize it to the format required by MeshNet and store it into XXX.npz. The reorganized file includes two parts of data:

• The "face" part contains the center position, vertices' positions and normal vector of each face.
• The "neighbor_index" part contains the indices of neighbors of each face.

If you wish to create and use your own dataset, simplify your models and organize the .off files similar to the ModelNet dataset. Then use the code in data/preprocess.py to transform them into the required .npz format. Notice that the parameter max_faces in config files should be maximum number of faces among all of your simplified mesh models.

You can modify the configuration in the config/train_config.yaml for your own training, including the CUDA devices to use, the flag of data augmentation and the hyper-parameters of MeshNet.

The pretrained MeshNet model weights are stored in pretrained model. You can download it and configure the "load_model" in config/test_config.yaml with your path to the weight file.