WorldPose
收藏资源简介:
WorldPose是由苏黎世联邦理工学院等机构创建的一个大规模多人在野外的3D人体姿态估计数据集。该数据集基于2022年世界杯的录像,利用多视角摄像头捕捉了超过250万条3D姿态数据,覆盖了超过120公里的运动轨迹。数据集的创建过程包括静态摄像头的校准、3D人体姿态和形状的估计,以及广播摄像头的校准。该数据集的应用领域包括体育分析、团队动态研究等,旨在解决现有数据集在多人动态场景中的不足,特别是对于大范围、多人协调运动的捕捉和分析。
WorldPose is a large-scale 3D human pose estimation dataset for multi-person wild scenarios, created by institutions including ETH Zurich and other research organizations. This dataset is based on footage from the 2022 FIFA World Cup, capturing over 2.5 million 3D pose instances via multi-view cameras and covering a total motion trajectory of more than 120 kilometers. The dataset construction workflow includes calibration of static cameras, estimation of 3D human poses and shapes, as well as calibration of broadcast cameras. Its application domains include sports analytics, team dynamics research, and more. This dataset aims to address the limitations of existing datasets in multi-person dynamic scenes, particularly in the capture and analysis of large-scale, coordinated multi-person movements.

- 1WorldPose: A World Cup Dataset for Global 3D Human Pose Estimation苏黎世联邦理工学院, 国际足联, 阿姆斯特丹大学, 微软 · 2025年



