Human3.6m
收藏资源简介:
Human3.6m是由奥格斯堡大学等机构采用的经典人体姿态数据集,包含7名演员执行15类动作的3D骨骼序列。该数据集以50Hz频率采集,包含32个关节点,经预处理后常采用22个动态关节进行模型训练,每段序列采样256个随机样本。数据通过专业运动捕捉系统在受控环境中录制,经指数映射转换为三维坐标。该资源主要应用于人体姿态预测研究,通过分析历史关节序列预测未来运动轨迹,为自动驾驶、人机交互等场景提供关键运动建模支持。
Human3.6m is a classic human pose dataset adopted by institutions such as the University of Augsburg. It contains 3D skeletal motion sequences of 7 actors performing 15 distinct action categories. Captured at a sampling rate of 50 Hz, the dataset includes 32 joint points. After preprocessing, 22 dynamic joints are commonly used for model training, and 256 random samples are extracted from each sequence. The data was recorded in a controlled environment using professional motion capture systems, and converted to 3D coordinates via exponential mapping. This resource is primarily applied in human pose prediction research, where future motion trajectories are predicted by analyzing historical joint sequences, providing critical support for motion modeling in scenarios such as autonomous driving and human-computer interaction.




