DLEN

PyTorch code for our paper" DLEN: Deep Laplacian Enhancement Networks for Low-Light Images "

Visual comparisons on LOLv1 and LOLv2

Dependencies

python==3.7.5
torch==1.7.1
torchvision==0.8.2
timm==0.2.4
matplotlib==3.4.3
tensorboard==2.5.0
numpy==1.19.5
scipy==1.7.1
opencv-python==4.2.0.34

cd DLEN 
pip install -r ./requirements.txt

Folder structure

Download the datasets and pretrained models first. Please prepare the basic folder structure as follows.

  /DLEN
    /src     # config files for datasets and PSNR and SSIM code
    /models   # python files for DLEN
    /pretrained_models  # folder for pretrained models
    requirements.txt
    README.md

Test

For the evaluation on LOL-v1 and LOL-v2, you should change input_dir (Input image directory) with your datasets path on the test.py

# put datasets and pretrained model in the corresponding directory 
cd DLEN 
python test.py

Train

The source code for training our DLEN will be available after the publication of the paper.