Yangchao0053/SGTNet

Codes for MICCAI 2023 paper: 3D Dental Mesh Segmentation Using Semantics-Based Feature Learning with Graph-Transformer

3D-Dental-Mesh-Segmentation-Using-Semantics-Based-Feature-Learning-with-Graph-Transformer

Codes for MICCAI 2023 paper: 3D Dental Mesh Segmentation Using Semantics-Based Feature Learning with Graph-Transformer

Time to Open our code

7.21 when the camera-ready files are submitted.

Project dependencies

python~=3.7

pytorch==1.11.0+cu113

vtk==9.2.2

vedo==2022.4.1

The more detailed dependencies can be checked in the requirements.txt.

Project Configuration

First, you need to install all the libraries listed in the requirements.txt.

pip install -r requirements.txt

To train your network, you need to specify all the xxxxx in the train.py to specify your data loader and log directory. In detail, the input of our network is an $N\times24$ matrix for each mesh. The initial $N \times 12$ C-domain matrix composes of the 3D coordinates of the 3 vertices and the centroid of each cell, and the initial $N \times 12$ N-domain matrix composes of the normal vectors of the 3 vertices and the centroid of each cell. Then you can run

python train.py

and your network can be trained and the tested data will be listed in your specified directory.

Dataset

We are sorry, but due to our business agreement with our partners, we are unable to provide the data. Please prepare the data yourself for training the network.