Home

Awesome

Recurrent Exposure Generation for Low-Light Face Detection

PyTorch implementation of the paper Recurrent Exposure Generation for Low-Light Face Detection in TMM 2021.

<img src="method.png" alt="alt text" style="zoom: 50%;" />

Dataset Preparation

The DARK FACE dataset can be downloaded from https://flyywh.github.io/CVPRW2019LowLight/.

Please replace _C.FACE.DSET_DIR in the file data/config.py with corresponding directory.

Pretrained Model

This implementation uses PyramidBox as base detector: Checkpoint

Usage

bash scripts/test_REGDet_pyramidbox.sh

Dependencies

Citation

If you use REGDet for your research, please cite:

@article{liang2021recurrent,
  ids = {liang2020recurrent},
  title = {Recurrent {{Exposure Generation}} for {{Low}}-{{Light Face Detection}}},
  author = {Liang, Jingxiu and Wang, Jingwen and Quan, Yuhui and Chen, Tianyi and Liu, Jiaying and Ling, Haibin and Xu, Yong},
  year = {2021},
  issn = {1941-0077},
  doi = {10.1109/tmm.2021.3068840},
  journal = {IEEE Transactions on Multimedia},
}