RUAS
This is the official code for the paper "Learning with Nested Scene Modeling and Cooperative Architecture Search for Low-Light Vision"
A preliminary version of this work has been published in CVPR 2021 [1]. RUAS In the conference work, we proposed a new method to integrate the principled optimization unrolling technique with a cooperative prior architecture search strategy for designing an effective yet lightweight low-light image enhancement network. In this journal submission, a series of substantial extensions have been made to improve our conference work.
Environment Preparing
python 3.6
pytorch 1.8.0
Testing Enhancement
We provide different models which are trained from different datasets, the models are saved in './weights/'. lol.pt is trained from LOL dataset. mit.pt is trained from MIT5K dataset. darkface.pt is trained from DarkFace dataset. Finally, run test_enhancement.py, the results will be saved in ./result/
python test_enhancement.py
--data_path #The folder path of the picture you want to test
'./data/enhance_test_data/lol/test'
--model #The checkpoint which you want use
'weights/lol.pt'
--save_path #The save path of the picture processed
./result/
Testing Detection
We provide model which is trained from DarkFACE dataset, please download the model from Google Drive or Baiduyun (Extraction code: ruas) and put it in './weights/'. Run test_detection.py, the results will be saved in ./result/
python test_detection.py
--data_path #The folder path of the picture you want to test
'./data/detection_test_data'
--model #The checkpoint for detection
'weights/detection.pth'
--save_path #The save path of the picture processed
./result/
Reference
If you find our work useful in your research please consider citing our paper:
@inproceedings{liu2021ruas,
title = {Retinex-inspired Unrolling with Cooperative Prior Architecture Search for Low-light Image Enhancement},
author = {Risheng, Liu and Long, Ma and Jiaao, Zhang and Xin, Fan and Zhongxuan, Luo},
booktitle = {Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
year = {2021}
}
A great thanks to DARTS for providing the basis for this code.