/CoTr

[MICCAI2021] CoTr: Efficiently Bridging CNN and Transformer for 3D Medical Image Segmentation

Primary LanguagePythonGNU General Public License v3.0GPL-3.0

CoTr: Efficient 3D Medical Image Segmentation by bridging CNN and Transformer

This is the official pytorch implementation of the CoTr:

Paper: CoTr: Efficient 3D Medical Image Segmentation by bridging CNN and Transformer.

Requirements

CUDA 11.0
Python 3.7
Pytorch 1.7
Torchvision 0.8.2

Usage

0. Installation

  • Install Pytorch1.7, nnUNet and CoTr as below
pip install torch==1.7.1+cu110 torchvision==0.8.2+cu110 torchaudio==0.7.2 -f https://download.pytorch.org/whl/torch_stable.html

cd nnUNet
pip install -e .

cd CoTr_package
pip install -e .

1. Data Preparation

  • Download BCV dataset
  • Preprocess the BCV dataset according to the uploaded nnUNet package.
  • Training and Testing ID are in data/splits_final.pkl.

2. Training

cd CoTr_package/CoTr/run

  • Run nohup python run_training.py -gpu='0' -outpath='CoTr' 2>&1 & for training.

3. Testing

  • Run nohup python run_training.py -gpu='0' -outpath='CoTr' -val --val_folder='validation_output' 2>&1 & for validation.

4. Citation

If this code is helpful for your study, please cite:

@article{xie2021cotr,
  title={CoTr: Efficiently Bridging CNN and Transformer for 3D Medical Image Segmentation},
  author={Xie, Yutong and Zhang, Jianpeng and Shen, Chunhua and Xia, Yong},
  booktitle={MICCAI},
  year={2021}
}
  

5. Acknowledgements

Part of codes are reused from the nnU-Net. Thanks to Fabian Isensee for the codes of nnU-Net.

Contact

Yutong Xie (xuyongxie@mail.nwpu.edu.cn)