Low-Light Image Enhancement with Multi-stage Residue Quantization and Brightness-aware Attention (ICCV2023)

This repository contains the Pytorch codes for paper Low-Light Image Enhancement with Multi-stage Residue Quantization and Brightness-aware Attention (ICCV (2023)). [paper]

Overview

In this paper, we propose a brightness-aware network with normal-light priors based on brightness-aware attention and residualquantized codebook. To achieve a more natural and realistic enhancement, we design a query module to obtain more reliable normal-light features and fuse them with lowlight features by a fusion branch. In addition, we propose a brightness-aware attention module to further improve the robustness of the network to the brightness. Extensive experimental results on both real-captured and synthetic data show that our method outperforms existing state-of-the-art methods.

Architecture

Figure 2: Architectures of the proposed three-stage framework for low-light image enhancement.

Results

Results on LOLv1, LOLv2-real and LOLv2-synthetic dataset can be downloaded from [Google Drive]

Usage

Download the RQ-LLIE repository

0. Requirements are Python 3 and PyTorch 1.8.0.
1. Download this repository via git

git clone https://github.com/LiuYunlong99/RQ-LLIE

or download the zip file manually.

Prepare the Dataset

Download the following datasets:

LOLv1: [Google Drive]

LOLv2: [Google Drive]

Evaluation

Download the pretrained models from [Google Drive]. And put them in the folder ./pretrained_models .

# LOLv1
python test_LOLv1_v2_real.py -opt options/test/LOLv1.yml

# LOLv2-Real
python test_LOLv1_v2_real.py -opt options/test/LOLv2_real.yml

# LOLv2-Synthetic
python test_LOLv2_synthetic.py -opt options/test/LOLv2_synthetic.yml

Note you need to change the dataroot_GT and dataroot_LQ to your path in the option file.

Training

Train the model on the corresponding dataset using the train config. For example, the training on LOLv1:

python -m torch.distributed.launch --nproc_per_node 1 --master_port 4320 train.py -opt options/train/LOLv1.yml --launcher pytorch

Acknowledgements

This source code is inspired by SNR(CVPR22).

Citation Information

If you find our work useful for your research, please consider giving this project a star and citing the following papers :)

@InProceedings{Liu_2023_ICCV,
    author    = {Liu, Yunlong and Huang, Tao and Dong, Weisheng and Wu, Fangfang and Li, Xin and Shi, Guangming},
    title     = {Low-Light Image Enhancement with Multi-Stage Residue Quantization and Brightness-Aware Attention},
    booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
    month     = {October},
    year      = {2023},
    pages     = {12140-12149}
}