X-CAUNET: Cross-Channel Attention with Underwater Image-Enhancement Transformer

• This paper has been accepted in ICASSP 2024.

Checkpoints

• You can download the trained models from here.

Datasets

• UIEB.
• SUIM-E.
• Please modify the line no. 31 in dataset.py for differnet datasets.

self.filesA, self.filesB = self.get_file_paths(self.data_path, 'UIEB') ---> UIEB or SUIM

• Dataset file structure should be as follows:

├── UIEB/SUIM
    ├── Train
        ├── inp
            ├── *.jpg/*.png
            ├── *.jpg/*.png
            └── ...
        ├── gt
            ├── *.jpg/*.png
            ├── *.jpg/*.png
            └── ...
    ├── Test
        ├── inp
            ├── *.jpg/*.png
            ├── *.jpg/*.png
            └── ...
        ├── gt
            ├── *.jpg/*.png
            ├── *.jpg/*.png
            └── ...

Train

python train.py

Test

-Please modify the line no. 28 in test.py with uieb.pt or suim.pt for different test data.

python test.py

Citation

@INPROCEEDINGS{10445832,
  author={Pramanick, Alik and Sarma, Sandipan and Sur, Arijit},
  booktitle={ICASSP 2024 - 2024 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)}, 
  title={X-CAUNET: Cross-Color Channel Attention with Underwater Image-Enhancing Transformer}, 
  year={2024},
  volume={},
  number={},
  pages={3550-3554},
  keywords={Correlation;Image color analysis;Message passing;Speech enhancement;Transformers;Colored noise;Image enhancement;Cross-attention;transformer;underwater image enhancement;deep learning},
  doi={10.1109/ICASSP48485.2024.10445832}}

Acknowledgements

• https://github.com/pksvision/Deep-WaveNet-Underwater-Image-Restoration