q-zh / absorption

Single Image Reflection Removal with Absorption Effect

The paper and supplementatry material can be found from here.

Dependencies

• Python 3.5+
• PyTorch 0.4.0+

Training

# Prepare the training data
Please use the provided matlab code (https://github.com/q-zh/absorption/tree/main/matlab).
# Download the VGG model
VGG-16 from http://cs.stanford.edu/people/jcjohns/fast-neural-style/models/vgg16.t7
# Run for training
python main.py 

Testing

Please find pre-traind model here with code: cvpr.

# Prepare the test set 
The data folder should include the reflection-contaminated images
# Run for test
python test.py 

Citation

If you find our code is useful, please cite our paper. If you have any problem of implementation or running the code, please contact us: [e-mail].

@inproceedings{CVPR2021_zheng_single,
  title={Single Image Reflection Removal with Absorption Effect},
  author={Zheng, Qian and Shi, Boxin and Chen, Jinnan and Jiang, Xudong and Duan, Ling-Yu and Kot, Alex C},
  booktitle={Proceedings of  Computer Vision and Pattern Recognition},
  year={2021}
}