/DeepUPE

Underexposed Photo Enhancement Using Deep Illumination Estimation

Primary LanguagePython

Underexposed Photo Enhancement Using Deep Illumination Estimation

Ruixing Wang1, Qing Zhang2, Chi-Wing Fu1, Xiaoyong Shen3, Wei-Shi Zheng2, Jiaya Jia1,3

1The chinese university of hong kong 2Sun Yat-sen University 3Tencent Youtu Lab

Usage

  1. Clone the repository:

    git clone https://github.com/wangruixing/DeepUPE.git
  2. Install the Python dependencies, run:

    cd main
    pip install -r requirements.txt
    make
  3. Evaluation: The test set can be downloaded in https://drive.google.com/open?id=1FrlMdnwiUfHthtw0jHdp40IbOVXfsoZJ. It includes 500 pair images from MIT-Adobe FiveK 4500-5000. You can download this and run:

    python main/run.py checkpoints <input file path> <output file path>

PSNR evaluation code is in avg_psnr.m. Modify the related paths in 'avg_psnr.m', and run it.

Errata

We recently found an implementation bug in calculating PSNR. Fortunately, this bug doesn't affect any of the conclusions in our paper, we have corrected this bug in the Matlab code and updated the corresponding values in the revised paper. We apologize for the confusion to readers.

Bibtex

@InProceedings{Wang_2019_CVPR,
author = {Wang, Ruixing and Zhang, Qing and Fu, Chi-Wing and Shen, Xiaoyong and Zheng, Wei-Shi and Jia, Jiaya},
title = {Underexposed Photo Enhancement Using Deep Illumination Estimation},
booktitle = {The IEEE Conference on Computer Vision and Pattern Recognition (CVPR)},
month = {June},
year = {2019}
}