caffeandtf/aod-net-pytorch

Pytorch implementation of AOD-Net : All-in-One Network for Dehazing - ICCV2017

AOD-Net

Introducion

This project is a pytorch implementation of AOD-Net : All-in-One Network for Dehazing. Recently, there are a number of good implementations:

• Boyiliee/AOD-Net the original author's project, developed based on pycaffe, only have pre-trained model and inference code
• TheFairBear/PyTorch-Image-Dehazing developed based on pytorch
• weber0522bb/AODnet-by-pytorch developed based on pytorch
• https://github.com/sachinpuranik99/AOD_Net developed based on keras

During our implementing, we referred the above implementations. However, our implementation has several unique and new features compared with the above implementations:

• It has both train and test code We provided both train and test code and pre-trained pytorch pickle file
• It has more clearly code structure We refactore the code structure of the dataloader to make it more consistent with pytorch dataset
• It has better logging decorator We add logging decorator for network pipeline function calling

Preparation

First of all, clone the code

git clone https://github.com/walsvid/AOD-Net.pytorch.git

Then, install prerequisites

pip install -r requirements.txt

Data Preparation

Please download the training images and original images from author's web page.

Then make a directory for data, change the parameters about data directories.

Pre-trained model

Please download the pretrained model from this download link.

Train

chmod +x run_train.sh
./run_train.sh

You can change the parameter in train bash script to satisfied your project.

Test

chmod +x run_test.sh
./run_test.sh

Demo

This is the dehazing result image comparison. Left image is haze image, right image is clean image processed by AOD-Net.

TODO

• Integrate existing models with detection tasks

Citation

If you using this project in your work, please don't forget to cite the original author's paper.

@inproceedings{li2017aod,
  title={Aod-net: All-in-one dehazing network},
  author={Li, Boyi and Peng, Xiulian and Wang, Zhangyang and Xu, Jizheng and Feng, Dan},
  booktitle={Proceedings of the IEEE International Conference on Computer Vision},
  volume={1},
  number={4},
  pages={7},
  year={2017}
}