Skip to content

Coherent Event Guided Low-Light Video Enhancement

License

Notifications You must be signed in to change notification settings

sherrycattt/EvLowLight

Repository files navigation

Coherent Event Guided Low-Light Video Enhancement

Paper | Project Page

Jinxiu Liang1, Yixin Yang1, Boyu Li1, Peiqi Duan1, Yong Xu2, Boxin Shi1

1Peking University
2South China University of Technology


⭐If EvLowLight is helpful for you, please help star this repo. Thanks!🤗

Table Of Contents

TODO

  • Release inference code and pretrained models.
  • Update links to paper and project page.
  • Provide a runtime environment Docker image.
  • Release train code and training set.

Installation

  1. Clone this repo using git:

    git clone https://github.com/sherrycattt/EvLowLight.git
  2. Create environment:

    Option 1: Using pip

    cd EvLowLight
    conda create -n evlowlight python=3.8
    conda activate evlowlight
    pip install -r requirements.txt

    Option 2: Using docker

    docker run --runtime=nvidia --gpus all --ipc=host --network=host  --rm -it \
      --ulimit memlock=-1 --ulimit stack=67108864 \
      -v `pwd`/EvLowLight:/workspace \
      -v `pwd`/timelens:/datasets/timelens \
      sherrycat/evlowlight

    Note the installation is only compatible with Linux users.

Inference

We provide an example for inference, check options/**_option.yml for more arguments.

python inference.py -opt options/timelens_option.yml

Data Preparation

We provide example test data converted from the TimeLens for demo, which can be downloaded from Link (extracted code: Y9CN). Please place the dataset in the ../datasets folder. The dataset structure should be organized as follows:

├── timelens
│   └── events
│       ├── paprika_1000_gain_control_02
│       │   ├── events.txt
│       │   └── timestamp.txt
│       ├── pen_03
│       │   ├── events.txt
│       │   └── timestamp.txt
│       ...
│   └── low
│       ├── paprika_1000_gain_control_02
│       │   ├── 000000.png
│       │   └── 000001.png
│       │   ...
│       ├── pen_03
│       │   ├── 000000.png
│       │   └── 000001.png
│       │   ...
│       ...
│  ...

Each subfolder in the low folder contains image files with template filename %06d.png, and the file in the events subfolder contains events corresponding to the image subfolder with template filename events.txt defined as ev_file_ext in the option configuration file. Moreover, events also contains timestamp.txt where image timestamps are stored. The image stamps in timestamp.txt should match with the image files .

Citation

Please cite us if our work is useful for your research.

@inproceedings{liang2023evlowlight,
  title = {Coherent Event Guided Low-Light Video Enhancement},
  booktitle = {Proceedings of the IEEE/CVF International Conference on Computer Vision (ICCV)},
  author = {Liang, Jinxiu and Yang, Yixin and Li, Boyu and Duan, Peiqi and Xu, Yong and Shi, Boxin},
  year = {2023},
  pages = {10615--10625},
}

License

This project is released under the Apache 2.0 license.

Acknowledgement

This project is based on BasicSR. Thanks for their awesome work.

Contact

If you have any questions, please feel free to contact with me at cssherryliang@pku.edu.cn.

About

Coherent Event Guided Low-Light Video Enhancement

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages