PoseTrack
收藏资源简介:
PoseTrack是一个大规模的基准数据集,由德国萨尔布吕肯马克斯普朗克信息学研究所创建,专注于视频中的人体姿态估计和关节跟踪。该数据集包含超过150,000个标注姿态和超过22,000个标注帧,涵盖了多种动态活动和人群密集场景。数据集的创建过程涉及从MPII人体姿态基准中选取关键帧,并围绕这些关键帧添加约五秒的视频片段,进行密集标注。PoseTrack的应用领域主要集中在解决视频中多人姿态估计和关节跟踪的挑战,旨在推动相关技术的发展和评估。
PoseTrack is a large-scale benchmark dataset created by the Max Planck Institute for Informatics in Saarbrücken, Germany, focusing on human pose estimation and joint tracking in videos. This dataset contains over 150,000 annotated poses and more than 22,000 annotated frames, covering a variety of dynamic activities and dense crowd scenes. The dataset construction process involves selecting key frames from the MPII Human Pose Benchmark, then adding approximately five-second video segments around these key frames for dense annotation. The primary application scenarios of PoseTrack focus on addressing the challenges of multi-person pose estimation and joint tracking in videos, aiming to promote the development and evaluation of related technologies.
- 1PoseTrack: A Benchmark for Human Pose Estimation and Tracking德国萨尔布吕肯马克斯普朗克信息学研究所 · 2018年



