The official pytorch implementation of the paper Revitalizing Convolutional Network for Image Restoration
All resulting images and pre-trained models are available in the provided links.
11/26/2024 Code for real haze and haze4k are released.
07/22/2024 We release the code for dehazing (ITS/OTS), desnowing, deraining, and motion deblurring.
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
Please use the pillow package downloaded by Conda rather than pip.
Install warmup scheduler:
cd pytorch-gradual-warmup-lr/
python setup.py install
cd ..
Please refer to respective directories.
| Model | Parameters | FLOPs |
|---|---|---|
| ConvIR-S (small) | 5.53M | 42.1G |
| ConvIR-B (base) | 8.63M | 71.22G |
| (追記) ConvIR-L (large) (追記ここまで) | 14.83M | 129.34G |
| Task | Dataset | PSNR | SSIM |
|---|---|---|---|
| Image Dehazing | SOTS-Indoor | 41.53/42.72 | 0.996/0.997 |
| SOTS-Outdoor | 37.95/39.42 | 0.994/0.996 | |
| Haze4K | 33.36/34.15/(追記) 34.50 (追記ここまで) | 0.99/0.99/(追記) 0.99 (追記ここまで) | |
| Dense-Haze | 17.45/16.86 | 0.648/0.621 | |
| NH-HAZE | 20.65/20.66 | 0.807/0.802 | |
| O-HAZE | 25.25/25.36 | 0.784/0.780 | |
| I-HAZE | 21.95/22.44 | 0.888/0.887 | |
| SateHaze-1k-Thin/Moderate/Thick | 25.11/26.79/22.65 | 0.978/0.978/0.950 | |
| NHR | 28.85/29.49 | 0.981/0.983 | |
| GTA5 | 31.68/31.83 | 0.917/0.921 | |
| Image Desnowing | CSD | 38.43/39.10 | 0.99/0.99 |
| SRRS | 32.25/32.39 | 0.98/0.98 | |
| Snow100K | 33.79/33.92 | 0.95/0.96 | |
| Image Deraining | Test100 | (追記) 31.40 (追記ここまで) | (追記) 0.919 (追記ここまで) |
| Test2800 | (追記) 33.73 (追記ここまで) | (追記) 0.937 (追記ここまで) | |
| Defocus Deblurring | DPDD | 26.06/26.16/(追記) 26.36 (追記ここまで) | 0.810/0.814/(追記) 0.820 (追記ここまで) |
| Motion Deblurring | GoPro | (追記) 33.28 (追記ここまで) | (追記) 0.963 (追記ここまで) |
| RSBlur | (追記) 34.06 (追記ここまで) | (追記) 0.868 (追記ここまで) |
@article{cui2024revitalizing,
title={Revitalizing Convolutional Network for Image Restoration},
author={Cui, Yuning and Ren, Wenqi and Cao, Xiaochun and Knoll, Alois},
journal={IEEE Transactions on Pattern Analysis and Machine Intelligence},
year={2024},
publisher={IEEE}
}
@inproceedings{cui2023irnext,
title={IRNeXt: Rethinking Convolutional Network Design for Image Restoration},
author={Cui, Yuning and Ren, Wenqi and Yang, Sining and Cao, Xiaochun and Knoll, Alois},
booktitle={International Conference on Machine Learning},
pages={6545--6564},
year={2023},
organization={PMLR}
}
Should you have any problem, please contact Yuning Cui.