P-DESTRE
收藏资源简介:
P-DESTRE数据集是由葡萄牙的University of Beira Interior和印度的JSS Science and Technology University联合创建的,旨在支持无人机(UAV)基于的行人搜索研究。该数据集包含多天的视频数据,提供了跨多天的连续ID标注,适用于无法依赖服装外观特征的行人搜索问题。数据集包含超过1400万个边界框,覆盖了261个已知身份,支持行人检测、跟踪、再识别和软生物识别等多种研究。
The P-DESTRE dataset was jointly created by the University of Beira Interior of Portugal and JSS Science and Technology University of India, aiming to support unmanned aerial vehicle (UAV)-based pedestrian search research. This dataset contains multi-day video data, providing continuous ID annotations across multiple days, which is suitable for pedestrian search scenarios where clothing appearance features cannot be relied upon. It includes over 14 million bounding boxes, covering 261 known identities, and supports multiple research directions such as pedestrian detection, tracking, re-identification, and soft biometrics.

- P-DESTRE数据集首次发表于《IEEE Transactions on Pattern Analysis and Machine Intelligence》期刊,标志着该数据集的正式诞生。
- P-DESTRE数据集首次应用于计算机视觉领域的深度学习研究,特别是在行人检测和姿态估计方面,展示了其强大的数据支持能力。
- P-DESTRE数据集被广泛应用于多个国际会议和研讨会,成为评估和比较行人检测算法性能的标准数据集之一。
- P-DESTRE数据集的扩展版本发布,增加了更多的标注信息和多样化的场景,进一步提升了其在复杂环境下行人分析任务中的应用价值。



