The code reproduce the results of our solutions on both tracks for Meta AI Video Similarity Challenge (CVPR 2023 Workshop).
Our solutions got the first place on both tracks, including Descriptor Track and Matching Track.
We provide DockerFiles to help you build a runnable environment directly. Or you can run setup.sh to install each package on your runnable environment.
Cause our matching solution partialy depend on the results of descriptor solution, you should reproduce the results of descriptor track first.
Please refer to VSC22-Descriptor-Track-1st and VSC22-Matching-Track-1st for more details.
If you find this code helpful, please cite our paper,
@article{wang2023dual,
title={A Dual-level Detection Method for Video Copy Detection},
author={Wang, Tianyi and Ma, Feipeng and Liu, Zhenhua and Rao, Fengyun},
journal={arXiv preprint arXiv:2305.12361},
year={2023}
}
@article{liu2023similarity,
title={A Similarity Alignment Model for Video Copy Segment Matching},
author={Liu, Zhenhua and Ma, Feipeng and Wang, Tianyi and Rao, Fengyun},
journal={arXiv preprint arXiv:2305.15679},
year={2023}
}