HuMMan
收藏资源简介:
HuMMan是由上海人工智能实验室创建的大规模多模态4D人体数据集,包含1000个人体对象,40万个序列和6000万帧。该数据集提供多种数据格式和标注,包括彩色图像、点云、关键点、SMPL参数和纹理网格,旨在支持动作识别、姿态估计、参数化人体重建和纹理网格重建等多项任务。数据集还包含一个由500个动作组成的动作集,覆盖基本运动,并支持跨设备域的研究。HuMMan数据集的创建过程涉及使用10个同步的RGB-D相机捕捉视频和深度序列,并通过后处理工具链将原始数据转换为多种格式。该数据集的应用领域广泛,旨在推动人体感知和建模技术的发展。
HuMMan is a large-scale multimodal 4D human dataset created by the Shanghai AI Laboratory. It includes 1000 human subjects, 400,000 sequences and 60 million frames. This dataset provides various data formats and annotations, including color images, point clouds, keypoints, SMPL parameters and textured meshes, aiming to support multiple tasks such as action recognition, pose estimation, parametric human reconstruction and textured mesh reconstruction. The dataset also contains an action set composed of 500 actions covering basic movements, and supports cross-device domain research. The creation of HuMMan dataset involved using 10 synchronized RGB-D cameras to capture video and depth sequences, and converted raw data into multiple formats via a post-processing toolchain. The dataset has a wide range of application fields and aims to promote the development of human perception and modeling technologies.




