MMVP
收藏资源简介:
MMVP是一个由北京航空航天大学等机构创建的多模态人体运动捕捉数据集,集成了视觉和压力传感器数据。该数据集包含超过44,000帧同步的RGBD视频和足底压力数据,特别适用于大范围和快速人体运动的精确形状估计、稳健姿态拟合和准确全局平移跟踪。数据集涵盖了跑步、跳远、舞蹈等多种运动类型,主要由青少年参与,旨在解决现有数据集在快速和大范围运动中足部接触标注不精确的问题。通过引入精确的足底压力信号,MMVP能够提供更密集和准确的足部接触标注,适用于增强现实/虚拟现实、疾病诊断、机器人操作和体育训练等多个领域的研究。
MMVP is a multimodal human motion capture dataset developed by institutions including Beihang University, which integrates visual and pressure sensor data. This dataset contains over 44,000 synchronized frames of RGBD video and plantar pressure data, and is particularly tailored for accurate shape estimation, robust pose fitting and precise global translation tracking of large-scale and fast human movements. Covering diverse motion types such as running, long jump, dance and more, the dataset is primarily collected from adolescent participants. It aims to address the issue of imprecise foot contact annotations in existing datasets for fast and large-scale motions. By leveraging precise plantar pressure signals, MMVP can deliver denser and more accurate foot contact annotations, making it suitable for research across multiple fields including augmented reality/virtual reality, disease diagnosis, robotic manipulation and sports training.




