SDAUT

by Jiahao Huang (j.huang21@imperial.ac.uk)

This is the official implementation of our proposed SDAUT:

Swin Deformable Attention U-Net Transformer (SDAUT) for Explainable Fast MRI

Please cite:

@ARTICLE{2022arXiv220702390H,
       author = {{Huang}, Jiahao and {Xing}, Xiaodan and {Gao}, Zhifan and {Yang}, Guang},
        title = "{Swin Deformable Attention U-Net Transformer (SDAUT) for Explainable Fast MRI}",
      journal = {arXiv e-prints},
     keywords = {Computer Science - Computer Vision and Pattern Recognition, Computer Science - Machine Learning, Electrical Engineering and Systems Science - Image and Video Processing},
         year = 2022,
        month = jul,
          eid = {arXiv:2207.02390},
        pages = {arXiv:2207.02390},
archivePrefix = {arXiv},
       eprint = {2207.02390},
 primaryClass = {cs.CV},
       adsurl = {https://ui.adsabs.harvard.edu/abs/2022arXiv220702390H},
      adsnote = {Provided by the SAO/NASA Astrophysics Data System}
}

The structure of SDAUT:

Our proposed swin deformable self-attention:

Requirements

matplotlib==3.3.4

opencv-python==4.5.3.56

Pillow==8.3.2

pytorch-fid==0.2.0

scikit-image==0.17.2

scipy==1.5.4

tensorboardX==2.4

timm==0.4.12

torch==1.9.0

torchvision==0.10.0

Training and Testing

Use different options (json files) to train different networks.

Calgary Campinas multi-channel dataset (CC)

To train SDAUT on CC:

python main_train_sdaut.py --opt ./options/SDAUT/example/train_sdaut_CCsagnpi_G1D30_kkddkk_offset_ps2_res256.json

To test SDAUT on CC:

python main_test_sdaut_CC.py --opt ./options/SDAUT/example/test/test_sdaut_CCsagnpi_G1D30_kkddkk_offset_ps2_res256.json

This repository is based on:

Swin Transformer for Fast MRI (code and paper);

SwinIR: Image Restoration Using Swin Transformer (code and paper);

Swin Transformer: Hierarchical Vision Transformer using Shifted Windows (code and paper).

Vision Transformer with Deformable Attention (code and [paper](h