English(EN) Privacy-Preserving Person Re-Identification from Temporal Sequences with Transformer and Hungarian Optimization

隐私保护行人重识别利用深度图像和Transformer

作者 PulseAugur 编辑部 · [1 个来源] · 2026-06-22 12:18

研究人员开发了一种新颖的行人重识别系统，该系统通过利用深度图像而非传统的RGB数据来优先保护隐私。该方法采用Transformer编码器处理时序序列，捕捉RGB和深度模态的动态运动模式。Hungarian算法用于优化不同视图之间的匹配过程，而batch hard triplet loss则增强了特征学习。在TVPR2和BIWI RGBD-ID等数据集上的评估表明，仅使用深度数据的方案在保护用户隐私的同时实现了具有竞争力的性能。 AI

排序理由该集群包含一篇详细介绍新颖行人重识别方法的论文。[lever_c_demoted from research: ic=1 ai=1.0]

在 arXiv cs.CV 阅读 →

AI 生成摘要 · Google Gemini · 来自 1 个来源。我们如何撰写摘要 →

报道来源 [1]

arXiv cs.CV TIER_1 English(EN) · Laurent Guimas · 2026-06-22 12:18

Privacy-Preserving Person Re-Identification from Temporal Sequences with Transformer and Hungarian Optimization

Person re-identification (Re-ID) is a crucial task in surveillance and human behavior analysis, often used in public spaces such as transport hubs. Traditional RGB-based Re-ID methods raise privacy concerns and are highly sensitive to lighting variations and occlusion. In this pa…

报道来源 [1]

Privacy-Preserving Person Re-Identification from Temporal Sequences with Transformer and Hungarian Optimization

相关实体

相关话题