https://github.com/JianghaiSCU/R2RNet

Official code of "R2RNet: Low-light Image Enhancement Via Real-low to Real-normal Network".
https://github.com/JianghaiSCU/R2RNet

Last synced: about 2 months ago
JSON representation

Official code of "R2RNet: Low-light Image Enhancement Via Real-low to Real-normal Network".

• Host: GitHub
• URL: https://github.com/JianghaiSCU/R2RNet
• Owner: JianghaiSCU
• Created: 2021-06-23T09:26:41.000Z (over 3 years ago)
• Default Branch: main
• Last Pushed: 2023-04-06T05:09:09.000Z (over 1 year ago)
• Last Synced: 2024-08-11T17:09:16.799Z (4 months ago)
• Language: Python
• Size: 11.5 MB
• Stars: 143
• Watchers: 1
• Forks: 17
• Open Issues: 5
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

• awesome-low-light-image-enhancement - github
• Awesome-Low-Light-Enhancement - Homepage (github project)

README

        
# R2RNet

Official code of "R2RNet: Low-light Image Enhancement via Real-low to Real-normal Network." Jiang Hai, Zhu Xuan, Ren Yang, Yutong Hao, Fengzhu Zou, Fang Lin, and Songchen Han. [[journal]](https://www.sciencedirect.com/science/article/pii/S1047320322002322)[[arxiv]](https://arxiv.org/abs/2106.14501)

## Citation

If you use this code or dataset from the paper for your research, please cite our paper:

```

@article{hai2023r2rnet,

  title={R2rnet: Low-light image enhancement via real-low to real-normal network},

  author={Hai, Jiang and Xuan, Zhu and Yang, Ren and Hao, Yutong and Zou, Fengzhu and Lin, Fang and Han, Songchen},

  journal={Journal of Visual Communication and Image Representation},

  volume={90},

  pages={103712},

  year={2023}

}

```

## Network Architecture

![fig3](https://user-images.githubusercontent.com/86350392/141397717-abff2d12-e810-4744-96e2-a1ce6af87002.jpeg)

The proposed R2RNet architecture. Our network consists of three subnets: a Decom-Net, a Denoise-Net, and a Enhance-Net, which perform decomposing, denoising, contrast enhancement and detail preservation, respectively. The Decom-Net decomposes the low-light image into an illumination map and a reflectance map based on the Retinex theory. The Denoise-Net aims to suppress the noise in the reflectance map. Subsequently, the illumination map obtained by Decom-Net and the reflectance map obtained by Denoise-Net are sent to the Relight-Net to improve image contrast and reconstruct details.

![fig4](https://user-images.githubusercontent.com/86350392/141397881-334d4764-5fe0-4412-9e87-fef882089c53.jpeg)

The proposed Relight-Net architecture. The Relight-Net consists of two modules: Contrast Enhancement Module (CEM) and Detail Reconstruction Module (DRM). CEM uses spatial information for contrast enhancement and DRM uses frequency information to preserve image details.

# Pytorch

This is a Pytorch implementation of R2RNet.

## Requirements

1. Python 3.x 

2. Pytorch == 1.9.0 (We used torch.fft.fftn(ifftn) and torch.fft.rfftn(irfftn) in our code). 

## Dataset

We have fixed the image naming bugs, you can download the LSRW dataset from: https://pan.baidu.com/s/1XHWQAS0ZNrnCyZ-bq7MKvA (code: wmrr). 

(Note: Some outdoor image pairs are not pixel-to-pixel aligned, there may be some local offsets between two images. In fact, we think that this has a limited impact on the LLIE task. If you think it has a great impact on your task, please choose the appropriate training strategy and evaluation metric by yourself.)

## Pre-trained model

You can download pre-trained models from：https://pan.baidu.com/s/1fYBAvzCuuzmaFmDDAlsCWA (code: wmr1), then put the pre-trained models into Decom, Denoise, Relight, respectively. 

The pre-trained VGG-16 model can be downloaded from:https://pan.baidu.com/s/1kf1uLjLaAMbfji0fPZKtEQ (code: wmrr).

## Testing Usage

python predict.py

## Training Usage

python trian.py

# Reference

Code borrows heavily from https://github.com/aasharma90/RetinexNet_PyTorch.