Multi-view Dynamic Reflection Prior for Video Glass Surface Detection (AAAI 2024)

[AAAI 2024] Official code release of our paper "Multi-view Dynamic Reflection Prior for Video Glass Surface Detection"

Dataest

The training and testing dataset is available at Google drive.

Evaluation

Download the predicted results from the link and run the following command to evaluate the results.

python eval.py -pred ../results/pred  -gt ../VGD_dataset/test

Inference

Download the trained model from the VGSD.pth and run the following command to generate the predicted results.

python infer.py -pred ../results/ -exp ../checkpoints/VGSD.pth 

Training

1. Please inference the reflection maps for the glass regions in the training dataset (192 videos) from SIRR.
2. Download backbone weights from the resnext_101_32x4d.pth and run train.py to train the model.

Acknowledgement

This responsibility is based on the following repositories: GSD, VMD, BiFormer, SIRR. Thanks to the authors for their excellent work.

Contact

If you have any questions, please feel free to contact me via fawnliu2333@gmail.com.

Citation

@inproceedings{liu2024multi,
  title={Multi-View Dynamic Reflection Prior for Video Glass Surface Detection},
  author={Liu, Fang and Liu, Yuhao and Lin, Jiaying and Xu, Ke and Lau, Rynson WH},
  booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
  volume={38},
  number={4},
  pages={3594--3602},
  year={2024}
}