PIE: Physics-inspired Contrastive Learning for Low-light Enhancement

Experiment

PyTorch implementation of PIE

Requirements

Python 3.8
PyTorch 1.4.0
opencv
torchvision
numpy
pillow
scikit-learn
tqdm
matplotlib
visdom

SCL-LLE does not need special configurations. Just basic environment.

Folder structure

The following shows the basic folder structure.

├── datasets
│   ├── data
│   │   ├── cityscapes
│   │   └── Contrast
|   ├── test_data
│   ├── cityscapes.py
|   └── util.py
├── lowlight_test.py # low-light image enhancement testing code
├── train.py # training code
├── lowlight_model.py
├── Myloss.py
├── CR.py
├── curves.py
├── checkpoints
│   ├── PIE.pth #  A pre-trained PiLLE model

Test

cd PIE

python lowlight_test.py

The script will process the images in the sub-folders of "test_data" folder and make a new folder "result" in the "datasets". You can find the enhanced images in the "result" folder.

Train

cd PIE
download the Cityscapes dataset
download the cityscapes training data google drive and contrast training data google drive
unzip and put the downloaded "train" folder and "Contrast" folder to "datasets/data/cityscapes/leftImg8bit" folder and "datasets/data" folder

python train.py

Contact

If you have any question, please contact xuzhengyan@nuaa.edu.cn

liuguoyou / PIE