URetinex-Net: Retinex-based Deep Unfolding Network for Low-light-Image-Enhancement

Official PyTorch implementation of URetinex-Net: Retinex-based Deep Unfolding Network for Low-light-Image-Enhancement in CVPR 2022.

[Paper] [Supplementary] [Video]

Requirements

1. Python == 3.7.6
2. PyTorch == 1.4.0
3. torchvision == 0.5.0

Test

If you only want to process a single image, just run like this (you can specify your image path)

python test.py --img_path "./demo/input/img.png"

Enhance results will be saved in ./demo/output if output_path is not specified!

Evaluate

If you want to evaluate using our provided pretrained model, please download the LOL test datasets. And arrange the dataset as ./test_data/LOLdataset/eval15. Then simply run

python evaluate.py

Citation

If you find our work useful, please cite our paper by the following:

@InProceedings{Wu_2022_CVPR,
    author    = {Wu, Wenhui and Weng, Jian and Zhang, Pingping and Wang, Xu and Yang, Wenhan and Jiang, Jianmin},
    title     = {URetinex-Net: Retinex-Based Deep Unfolding Network for Low-Light Image Enhancement},
    booktitle = {Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR)},
    month     = {June},
    year      = {2022},
    pages     = {5901-5910}
}

Noted that the code is only for non-commercial use! should you have any queries, contact me at wj1997s@163.com