Empowering Low-Light Image Enhancer through Customized Learnable Priors (ICCV2023)

October 12, 2023 · View on GitHub

Paper link, arXiv

Naishan Zheng*, Man Zhou*, Yanmeng Dong, Xiangyu Rui, Jie Huang, Chongyi Li, Feng Zhao

*Equal Contribution

University of Science and Technology of China, Xi’an Jiaotong University, S-Lab, Nanyang Technological University

How to test on LOL

  1. Update the paths of image sets and pre-trained models.
Updating the paths in configure files of /CUE/options/test/learnedPrior/LearnablePrior.yml
  1. Run the testing commands.
python test.py -opt /CUE/options/test/learnedPrior/LearnablePrior.yml

How to train CUE

Some steps require replacing your local paths.

  1. Training the learnable noise prior.
python train.py -opt /CUE/options/train/learnedPrior/MAE_refl_hog.yml
  1. Training the learnable illumination prior.
python train.py -opt /CUE/options/train/learnedPrior/UNet_illu_bil.yml
  1. Training the enhancement network.
python train.py -opt /CUE/options/train/learnedPrior/LearnablePrior.yml