- python 3.7
- CUDA==11.2
- faiss-gpu==1.6.0
- There is a processed tar file in BaiduYun (Password: 9pmk) with all needed files.
See run.sh for more information.
If you find this code useful in your research, please cite:
@article{gong2024cross,
title={Cross-modality perturbation synergy attack for person re-identification},
author={Gong, Yunpeng and Zhong, Zhun and Qu, Yansong and Luo, Zhiming and Ji, Rongrong and Jiang, Min},
journal={Advances in Neural Information Processing Systems},
volume={37},
pages={23352--23377},
year={2024}
}
The code is based on LTA, Random Color Erasing and Mata_attack.
Our Related Code (The code is included in an open-source project with more than 4.4k stars. Reference RandomGrayscaleErasing.py)
- Multiform Attack for Transferable Cross-Modal Person Re-Identification
- Person re-identification method based on color attack and joint defence
If you use the code, please kindly cite the paper.
@inproceedings{colorAttack2022,
title={Person re-identification method based on color attack and joint defence},
author={Gong, Yunpeng and Huang, Liqing and Chen, Lifei},
booktitle={Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition},
pages={4313--4322},
year={2022}
}
@inproceedings{gong2026theory,
title={A theory-inspired framework for few-shot cross-modal sketch person re-identification},
author={Gong, Yunpeng and Hou, Yongjie and Shi, Jiangming and Diep, Kim Long and Jiang, Min},
booktitle={Proceedings of the AAAI Conference on Artificial Intelligence},
volume={40},
number={6},
pages={4284--4292},
year={2026}
1.Recently, some users have reported that running the same code now produces results that differ significantly from those obtained a few months ago. We suspect this is most likely caused by updates to packages or the runtime environment. It is a common issue that significant discrepancies in results arise due to updates in the code package and its environment. At the moment, there is no good solution, and the author does not have the time or capacity to maintain it. Please make sure that your environment and package versions are configured according to the versions used when the Mata_attack code was released.
2.In addition, questions regarding the model accuracy of AGW and DDAG are frequently asked. For such inquiries, it is recommended to consult the respective authors of the relevant papers.
The code was organized and released within a relatively short timeframe after the experiments were completed, so some details may not have been fully standardized. In fact, the released code already includes the main functionalities required for implementing the proposed method and conducting the relevant experiments. However, during our research, certain parts of the code, parameter configurations, and experimental procedures were temporarily modified to accommodate specific experiments and testing purposes. Some of these changes were not fully reverted or consolidated before the code was released. As a result, a few implementation details may not perfectly align with the descriptions in the paper or the standard experimental settings. These discrepancies are mainly related to code organization and experimental configurations and can be addressed with minor adjustments.
Email: fmonkey625@gmail.com