Rosie-Lab/compas3d

CoMPAS3D是一个大规模的运动捕捉数据集，旨在支持通过舞蹈进行非语言、身体交流的研究。它包含18位舞者在初学者、中级和专业人士技能水平上的即兴萨尔萨双人舞表演超过3小时。每个序列都包含高保真的3D运动数据，形式为SMPL-X (.npz)文件，以及视频可视化和同步音频(.mp4)。我们为50%的序列提供了详细的帧级标注，包括动作类型、风格变化和执行错误。该数据集是在一个受控的工作室中使用20台Vicon系统以120fps的速度收集的。数据集包括72个长格式序列（每个2.5分钟）的自然领导-跟随互动。参与者来自不同的舞蹈背景，所有参与者都提供了数据发布的信息同意书。CoMPAS3D为研究具身对话、社交流畅性和AI和机器人中的多代理协调提供了一个丰富的测试平台。

CoMPAS3D (Complex Multi-Level Person-Interaction Annotated Salsa Dataset) is a large-scale motion capture dataset designed to support research on nonverbal, physical communication through dance. It contains over 3 hours of improvised salsa duet performances by 18 dancers across beginner, intermediate, and professional skill levels. Each sequence features high-fidelity 3D motion data in the form of SMPL-X (.npz) files, along with video visualizations and synchronized audio (.mp4). We provide detailed frame-level annotations of move types, stylistic variation, and execution errors for 50% of the sequences (.txt). Collected in a controlled studio using a 20-camera Vicon system at 120fps, the dataset includes 72 long-form sequences (each 2.5 minutes) of naturalistic leader-follower interaction. Participants were drawn from diverse dance backgrounds, and all provided informed consent for data release. CoMPAS3D offers a rich testbed for studying embodied dialogue, social fluency, and multi-agent coordination in AI and robotics.