This repository contains the official code of DS-TransUNet: Dual Swin Transformer U-Net for Medical Image Segmentation

• Python==3.7.6
• Pytorch==1.8.0 && CUDA 11.1
• timm==0.4.5

1. Dataset
   • Downloading training dataset and move it into ./data, which can be found in this download link (Google Drive).
   • Downloading testing dataset and move it into ./data , which can be found in this download link (Google Drive).
2. Testing
   • Downloading our trained DS-TransUNet-B from Baidu Pan (dd79), and move it into ./checkpoints.
   • run test_kvasir.py
   • run crcriteria.py to get the DICE score, which uses EvaluateSegmentation. Or you can download our result images from Baidu Pan (dd79).
3. Training
   • downloading Swin-T and Swin-B from Swin-Transformer to ./checkpoints.
   • run train_kvasir.py

Code of other tasks will be comming soon.

Some of the codes in this repo are borrowed from:

• Swin-Transformer
• PraNet
• TransFuse

Please consider citing us if you find this work helpful:

@article{lin2022ds,
  title={DS-TransUNet: Dual Swin Transformer U-Net for Medical Image Segmentation},
  author={Lin, Ailiang and Chen, Bingzhi and Xu, Jiayu and Zhang, Zheng and Lu, Guangming and Zhang, David},
  journal={IEEE Transactions on Instrumentation and Measurement},
  year={2022},
  publisher={IEEE}
}

Please drop an email to [email protected]