DM2F-Net-improve

By Yuyan Chen.

Results

	算法	MSE	PNSR	SSIM	CIEDE2000
O-Haze	baseline	0.0038	24.304	0.7192	4.7643
	improve	0.0037	24.436	0.7242	4.7547
HazeRD	baseline	0.0679	14.481	0.8314	16.161
	improve	0.0695	14.584	0.8315	16.036

The dehazing results can be found at Baidu Wangpan.

Installation & Preparation

Make sure you have Python>=3.7 installed on your machine.

Environment setup:

1. Create conda environment

   conda create -n dm2f
   conda activate dm2f
   

2. Install dependencies (test with PyTorch 1.8.0):

   1. Install pytorch==1.8.0 torchvision==0.9.0 (via conda, recommend).

   2. Install other dependencies

      pip install -r requirements.txt
      

• Prepare the dataset

  • Download the RESIDE dataset from the official webpage.

  • Download the O-Haze dataset from the official webpage.

  • Make a directory ./data and create a symbolic link for uncompressed data, e.g., ./data/RESIDE.

Training

1. Set the path of datasets in tools/config.py
2. Run by python train.py

Use pretrained ResNeXt (resnext101_32x8d) from torchvision.

Training a model on a single RTX 3080 Ti(12GB) GPU takes about 6 hours.

Testing

1. Set the path of five benchmark datasets in tools/config.py.
2. Put the trained model in ./ckpt/.
3. Run by python test.py (for O-Haze/HazeRD/RESIDE) or python output.py (for own pictures)

Settings of testing were set at the top of test.py, and you can conveniently change them as you need.

License

DM2F-Net is released under the MIT license.

Citation

If you find the paper or the code helpful to your research, please cite the project.

@inproceedings{deng2019deep,
  title={Deep multi-model fusion for single-image dehazing},
  author={Deng, Zijun and Zhu, Lei and Hu, Xiaowei and Fu, Chi-Wing and Xu, Xuemiao and Zhang, Qing and Qin, Jing and Heng, Pheng-Ann},
  booktitle={Proceedings of the IEEE/CVF International Conference on Computer Vision},
  pages={2453--2462},
  year={2019}
}