Plumess / A-Cross-scale-Framework-for-Low-light-Image-Enhancement-Using-Spatial-spectral-Infomation

A Cross-scale Framework for Low-light Image Enhancement Using Spatial-spectral Infomation
MIT License
1 stars 0 forks source link

README.md

A Cross-scale Framework for Low-light Image Enhancement Using Spatial-spectral Infomation

This is a pytorch implementation of the CE Paper in Computers & Electrical Engineering 2023 and the VCIP Paper in 2022 IEEE VCIP

This is the link of pre-trained model in GoogleDrive or BaiduNetDisk(23tc).

This is the link of a selection of test images used to verify that the code has been deployed correctly, including images from the Sony, Fuji and LOL datasets in GoogleDrive or BaiduNetDisk(9pru).


Requirements


Description

Additionally:

Since 'window attention' is in a square window, the original image is cropped to a fixed size square patch for both testing and training in the code. If you wish to train and test at full size, please refer to the modifications to the SpectralTransform, LeFF, LeWinTransformer, Input, Output and ResLeWinTransformerLayer's forward methods in model_fullSize.py, And the hiding of the crop operation in dataset_sony_fullSize.py.


Test

  1. Download the pre-trained model
  2. Prepare test dataset
  3. Run the corresponding main.py

Train

  1. Prepare the dataset
  2. Train the network. That's all.

Citation

If you use our code, please cite our paper.


Notes