Low-Light-Video-Enhancement-by-Learning-on-Static-Videos-with-Cross-Frame-Attention

This is a pythorch implementation of “Low-Light-Video-Enhancement-by-Learning-on-Static-Videos-with-Cross-Frame-Attention” in BMVC 2022, by Shivam Chhirolya, Sameer Malik, and Rajiv Soundararajan.

Paper

Requirements

pytorch 1.10.0
cudatoolkit 10.2
numpy 2.7
opencv-python 4.5
einops 0.4.1
skimage

Checkpoints

Please download the checkpoints from below link and put those checkpoints in checkpoints directory

https://drive.google.com/drive/folders/1fdQVKbYXXpc4icS4GvuJJCY92T7ejmag?usp=sharing

Usage

Testing

1. To test our model on DAVIS dataset run " test_DAVIS.py "
2. To test our model on DRV RGB dataset run " test_DRV.py "

Training

1. To train our model on DAVIS dataset run " train_DAVIS.py "
2. To train our model on DRV RGB dataset run " train_DRV.py "

Model

1. Davis_901.pth (This model is trained using synthetic low light Davis dataset)
2. DRV_901.pth (This model is trained using RGB form of DRV dataset)