SLOPER4D
收藏资源简介:
SLOPER4D是一个大规模场景感知数据集,用于城市环境中全球4D人体姿态估计研究。该数据集由厦门大学、上海科技大学和马克斯·普朗克智能系统研究所合作创建,包含15个运动序列,覆盖超过2000平方米至13000平方米的区域,总距离超过8公里。数据集内容丰富,包括超过10万帧LiDAR数据、30万视频帧和50万基于IMU的运动帧。创建过程中,使用头戴式LiDAR和相机捕捉12名受试者在10个多样城市场景中的活动,并提供帧级2D关键点、3D姿态参数及全球平移标注。SLOPER4D旨在解决大规模动态场景中的人体姿态估计问题,特别适用于增强/虚拟现实、自动驾驶和智能城市等领域的研究。
SLOPER4D is a large-scale scene-aware dataset developed for research on global 4D human pose estimation in urban environments. It was collaboratively constructed by Xiamen University, ShanghaiTech University, and the Max Planck Institute for Intelligent Systems. The dataset contains 15 motion sequences, covering areas ranging from over 2,000 to 13,000 square meters, with a total traversed distance exceeding 8 kilometers. It boasts a rich array of data, including over 100,000 LiDAR frames, 300,000 video frames, and 500,000 IMU-based motion frames. During the data collection process, head-mounted LiDAR sensors and cameras were used to capture the activities of 12 human subjects across 10 diverse urban scenarios, and frame-level 2D keypoints, 3D pose parameters, as well as global translation annotations are provided. SLOPER4D aims to address the challenge of human pose estimation in large-scale dynamic scenes, and is particularly suitable for research in fields such as augmented/virtual reality, autonomous driving, and smart cities.

- 1SLOPER4D: A Scene-Aware Dataset for Global 4D Human Pose Estimation in Urban Environments厦门大学, 中国 上海科技大学, 中国 马克斯·普朗克智能系统研究所, 德国 · 2023年



