HybrUR

A Hybrid Physical-Neural Solution for Unsupervised Underwater Image Restoration

Requirements

Python 3
Pytorch (with CUDA)
Opencv-Python
Pyyaml

Project directory

checkpoints -- the trained network parameters to reload the HybrUR model
config -- deploy the testing parameters
model -- methodology implementation
results -- the storage path of generated images
samples -- network input (underwater and in-air images)
utils -- utility functions

Usage

cd `project root directory`
python test.py

The generated results will be saved in the results folder