Human3.6Mplus
收藏资源简介:
Human3.6Mplus是由德克萨斯大学圣安东尼奥分校研究团队构建的大型对齐数据集,旨在桥接三维人体姿态估计与生物力学分析。该数据集将Human3.6M视频与三维关键点序列,同其生物力学标签空间进行精确配对,涵盖七名受试者的三十种活动,实现了帧级准确的跨模态监督。其创建过程通过建立并验证两种模态坐标系间的解剖学对应关系,以共享骨盆根节点为锚点,确保生物力学模拟标签与视觉姿态数据的无缝对齐。该数据集主要应用于康复医学、运动科学及临床运动分析领域,旨在解决传统姿态估计缺乏生理学意义量化的问题,为无标记生物力学分析提供大规模监督数据支持。
Human3.6Mplus is a large-scale aligned dataset constructed by the research team from the University of Texas at San Antonio, with the goal of bridging the gap between 3D human pose estimation and biomechanical analysis. This dataset precisely pairs Human3.6M videos and 3D keypoint sequences with their matching biomechanical label spaces, covering thirty types of activities performed by seven subjects, thereby enabling frame-level accurate cross-modal supervision. In the process of creating this dataset, the anatomical correspondence between the two modal coordinate systems was established and validated, taking the shared pelvic root node as the anchor point to ensure seamless alignment between biomechanical simulation labels and visual pose data. This dataset is mainly applied in the fields of rehabilitation medicine, exercise science and clinical motion analysis, aiming to resolve the problem that traditional pose estimation approaches lack physiologically meaningful quantification, and provide large-scale supervised data support for markerless biomechanical analysis.




