/ELIE_STR

Primary LanguagePython

Extremely Low-light Image Enhancement with Scene Text Restoration

This repository is the official implementation of our work, which is accepted to ICPR 2022 as an oral presentation.

Pre-trained Models

Please download the data here: SID-Sony files and then decompress it to the "/data" folder.

Requirements

To create the conda environment, run:

conda env create -n elie_str --file elie_str.yml

Training

To train the SID-Sony model, run:

python train_Sony_SID.py

Evaluation

To evaluate the trained model, run:

python test_Sony_SID.py

Citation

If you use this in your research, please cite our work by using the following BibTeX entry:

 @article{hsu2022extremely,
  title={Extremely Low-light Image Enhancement with Scene Text Restoration},
  author={Hsu, Pohao and Lin, Che-Tsung and Ng, Chun Chet and Kew, Jie-Long and Tan, Mei Yih and Lai, Shang-Hong and Chan, Chee Seng and Zach, Christopher},
  journal={arXiv preprint arXiv:2204.00630},
  year={2022}
}