lee-ff / recursive-image-dehazing-via-perceptually-optimized-generative-adversarial-network-pogan

Recursive Image Dehazing via Perceptually Optimized Generative Adversarial Network

Recursive-Image-Dehazing-via-Perceptually-Optimized-Generative-Adversarial-Network-POGAN

Here is the code for our paper entitled "Recursive Image Dehazing via Perceptually Optimized Generative Adversarial Network (POGAN)".

The code has been tested on Tensorflow 1.4.0.

To do testing:

python main.py --mode=e

To do training, simply put original and hazy images in folder data/train_ori and data/train_haze respectively, then run:

python main.py

If you find the code useful, please consider cite our work:

@inproceedings{du2019recursive,
  title={Recursive Image Dehazing via Perceptually Optimized Generative Adversarial Network (POGAN)},
  author={Du, Yixin and Li, Xin},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition Workshops},
  year={2019}
}

If you have any question, please contact [email protected]

We acknowledge and thank the author of SRGAN for sharing their source code: https://github.com/tensorlayer/srgan