arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2301.06286cs.CV
收藏

基于Meta的人员身份再识别生成攻击

Meta Generative Attack on Person Reidentification

IIITD(印度德里英迪拉·甘地信息技术学院)

A V Subramanyam

摘要:

最近在重新鉴定人员身份时对对抗攻击进行了调查。这些攻击在跨数据集或跨模型设置下表现良好。然而,跨数据集跨模型场景中存在的挑战不允许这些模型实现类似的准确性。为此,我们提出了我们的方法,目标是针对不同的模型和跨数据集实现更好的可移植性。我们生成一个掩码来获得更好的跨模型性能,并使用Meta学习来提高具有挑战性的跨数据集跨模型设置的泛化能力。在Market-1501、DukeMTMC-reID和MSMT-17上的实验结果表明,与其他攻击相比,该攻击具有良好的效果。

英文摘要:

Adversarial attacks have been recently investigated in person re-identification. These attacks perform well under cross dataset or cross model setting. However, the challenges present in cross-dataset cross-model scenario does not allow these models to achieve similar accuracy. To this end, we propose our method with the goal of achieving better transferability against different models and across datasets. We generate a mask to obtain better performance across models and use meta learning to boost the generalizability in the challenging cross-dataset cross-model setting. Experiments on Market-1501, DukeMTMC-reID and MSMT-17 demonstrate favorable results compared to other attacks.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑