MoVi
收藏资源简介:
MoVi数据集是由约克大学创建的大型多用途人体运动和视频数据集,包含90名男女演员执行20种日常动作和体育运动的视频记录。数据集通过光学运动捕捉系统、视频摄像机和惯性测量单元(IMU)进行五轮捕捉,总时长包括9小时的动捕数据、17小时的视频数据和6.6小时的IMU数据。数据集创建过程中,演员穿着不同服装进行动作捕捉,确保数据的多样性和准确性。MoVi数据集适用于人体姿态估计、动作识别、运动建模、步态分析和身体形状重建等多个领域,旨在解决基于视频的人体运动分析中的关键问题。
The MoVi Dataset is a large-scale, multi-purpose human motion and video dataset developed by York University. It encompasses video recordings of 90 male and female actors performing 20 daily activities and sports movements. The dataset was collected across five capture sessions using optical motion capture systems, video cameras, and inertial measurement units (IMUs). The total aggregated duration includes 9 hours of motion capture data, 17 hours of video data, and 6.6 hours of IMU data. During the dataset's development, actors wore varied outfits during motion capture to ensure data diversity and accuracy. The MoVi Dataset is applicable to multiple research fields including human pose estimation, action recognition, motion modeling, gait analysis, and body shape reconstruction, and it aims to address key challenges in video-based human motion analysis.




