Dual-domain strip attention for image restoration

Yuning Cui, Alois Knoll

Installation

The project is built with PyTorch 3.8, PyTorch 1.8.1. CUDA 10.2, cuDNN 7.6.5 For installing, follow these instructions:

conda install pytorch=1.8.1 torchvision=0.9.1 -c pytorch
pip install tensorboard einops scikit-image pytorch_msssim opencv-python

Install warmup scheduler:

cd pytorch-gradual-warmup-lr/
python setup.py install
cd ..

Pre-trained models here

Results (DSANet)

|Task|Dataset|PSNR|SSIM| |----|------|-----|----| |Image Dehazing|ITS|41.36|0.997| ||OTS|38.39|0.995| ||Dense-Haze|16.70|0.607| ||NH-HAZE|20.51|0.801| ||O-HAZE|25.79|0.94| ||NH-HAZE2|21.63|0.841| ||NHR|28.08|0.978| |Image Desnowing|CSD|38.09|0.99| ||SRRS|31.97|0.98| ||Snow100K|33.70|0.95|

Citation

If you find this project useful for your research, please consider citing:

@article{cui2024dual,
  title={Dual-domain strip attention for image restoration},
  author={Cui, Yuning and Knoll, Alois},
  journal={Neural Networks},
  volume={171},
  pages={429--439},
  year={2024},
  publisher={Elsevier}
}

Contact

Should you have any question, please contact Yuning Cui.