ImgRestoration for Divers!

• This repository is an easy toolkit to restore the underwater image/video using Deep Learning model!
• It uses the model from the paper Underwater Ranker: Learn Which is better and How to be better [AAAI 2023 Oral]

Get Started

• Install python>=3.6
• Git clone this repository by typing this command in the command line:
  • git clone https://github.com/AllenEdgarPoe/UnderwaterImgRestore4Diver.git
• Create python virtual environment by using the below command line:
  • python -m venv venv
• Activate the virtual environment:
  • venv\Scripts\activate
• Install required libraries:
  • pip install -r requirements.txt
• Download models from models
  • get NU2Net_ckpt.pth and put it in ./checkpoints/NU2Net_ckpt.pth
• Start the program!
  • python main.py

Functions

• Convert the single image/video file by clicking Single file tab
• Convert the whole image/video files by clicking Batch tab.
• You can set the save directory by designating it through output_path tab. If you don't, the default saving path is ./results
• Comparison results will be saved in [save_path]/compare
• Enjoy the beautiful comparison results in the below tab!

Examples

Dialog

Reulst (Image comparison)

All of the pictures were taken by me @ Gili.T (Indonesia)

Result (Video comparison)

dive.mp4

Acknowledgement

• The code for model architecture is from UnderwaterRanker