Skip to content

ayanglab/SwinGANMR

Repository files navigation

SwinGANMR

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

This is the official implementation of our proposed ST-GAN, EES-GAN and TES-GAN:

Fast MRI Reconstruction: How Powerful Transformers Are?

Please cite:

@ARTICLE{2022arXiv220109400H,
       author = {{Huang}, Jiahao and {Wu}, Yinzhe and {Wu}, Huanjun and {Yang}, Guang},
        title = "{Fast MRI Reconstruction: How Powerful Transformers Are?}",
      journal = {arXiv e-prints},
     keywords = {Electrical Engineering and Systems Science - Image and Video Processing, Computer Science - Computer Vision and Pattern Recognition, Computer Science - Machine Learning},
         year = 2022,
        month = jan,
          eid = {arXiv:2201.09400},
        pages = {arXiv:2201.09400},
archivePrefix = {arXiv},
       eprint = {2201.09400},
 primaryClass = {eess.IV},
}

Overview_of_SwinGANMR

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 ST-GAN on CC:

python main_train_stganmr.py --opt ./options/STGAN/example/train_stganmr_CCnpi_G1D30.json

To train EES-GAN on CC:

python main_train_eesganmr.py --opt ./options/EESGAN/example/train_eesganmr_CCnpi_G1D30.json

To train TES-GAN on CC:

python main_train_tesganmr.py --opt ./options/TESGAN/example/train_tesganmr_CCnpi_G1D30.json

To test ST-GAN on CC:

python main_test_stganmr_CC.py --opt ./options/STGAN/example/test/test_stganmr_CCnpi_G1D30.json

To test EES-GAN on CC:

python main_test_eesganmr_CC.py --opt ./options/EESGAN/example/test/test_eesganmr_CCnpi_G1D30.json

To test TES-GAN on CC:

python main_test_tesganmr_CC.py --opt ./options/TESGAN/example/test/test_tesganmr_CCnpi_G1D30.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).