DehazeGAN: Sand-Dust Image Restoration using Transformer-based Denoiser and Adversarial Learning

how to start

1. install packages in requirements.txt
2. set pythonpath to the anaconda env dir and root dir of this project
3. (download SIE dataset from Google Drive to the Data directory)

data directory structure

Data/
└── paired/
|   ├── ground_truth/
|   │   ├── ...
|   │   ├── ...
|   │   └── ...
|   └── noisy/
|       ├── ...
|       ├── ...
|       └── ...
└── unpaired/
    ├── clear/
    │   ├── ...
    │   ├── ...
    │   └── ...
    └── noisy/
        ├── ...
        ├── ...
        └── ...