UTD-MHAD 人体动作识别数据集

UTD 全称 University of Texas at Dallas，MHAD 全称 Multimodal Human Action Dataset 。这个数据集由 8 名受试者 27 个动作的视频组成。每个受试者重复一个动作 4 次，总共产生 861 个动作序列（有 3 个动作序列因受损被删除）。该数据集有四种时间同步的数据模式，分别是 RGB 视频、深度视频、骨骼位置、来自 Kinect 摄像头和可穿戴惯性传感器的惯性信号。

UTD stands for University of Texas at Dallas, while MHAD stands for Multimodal Human Action Dataset. This dataset comprises videos of 27 distinct actions performed by 8 subjects. Each subject repeated each action 4 times, yielding a total of 861 action sequences, with 3 sequences removed due to data corruption. The dataset includes four time-synchronized data modalities: RGB video, depth video, skeletal joint positions, and inertial signals collected from both a Kinect camera and wearable inertial sensors.

搜集汇总

背景与挑战

背景概述

UTD-MHAD是一个多模态人体动作识别数据集，包含8名受试者执行27个动作的RGB视频、深度视频、骨骼位置和惯性信号数据，总大小1.17GB。该数据集适用于多模态融合研究和人类行为识别领域。