Skip to content

Latest commit

 

History

History
43 lines (32 loc) · 1.34 KB

README.md

File metadata and controls

43 lines (32 loc) · 1.34 KB

CasDyF-Net: Image Dehazing via Cascaded Dynamic Filters

Authors: Yinglong Wang, Bin He

Paper link(arxiv): Link of paper

Installation

The project is built with Python 3.8, PyTorch 1.10.0, CUDA 11.3,

For installing, follow these instructions:

pip install -r requirements.txt

If you're having problems with your version, you can try installing each package one by one using pip alone

Training and Evaluation

To Train ,you can run this cmd in your terminal:

python main.py --mode train --dataset yourpath --batch_size 8

You can also change the parameters in the main.py file.Saving your changes and running main.py will have the same effect

Results

You can download the pre-trained models from this link:

Download the pre-trained models

If you just need the output images of the models, you can download the images from this link:

Download the images

Dataset PSNR SSIM
SOTS-Indoor 43.21 0.997
SOTS-Outdoor 38.86 0.995
Dense-Haze 17.56 0.658
O-HAZE 25.44 0.936
Haze4K 35.73 0.99

Citation

Contact

e-mail: wangyinglong2023@gmail.com

wechat: dauing2023