Waymo-3DSkelMo
收藏资源简介:
Waymo-3DSkelMo 是一个大规模、高质量的 3D 骨骼运动数据集,用于自动驾驶中的行人交互建模。该数据集从 Waymo 感知数据集中提取,包含超过 14,000 秒的 3D 骨骼运动序列,涵盖超过 800 个真实驾驶场景,每个场景平均有 27 个智能体进行交互。数据集采用 LiDAR 和运动先验技术,克服了现有数据集的遮挡和时间连续性问题,提供了丰富多样的多智能体交互模式。
Waymo-3DSkelMo is a large-scale, high-quality 3D skeletal motion dataset designed for pedestrian interaction modeling in autonomous driving. Extracted from the Waymo Perception Dataset, it contains over 14,000 seconds of 3D skeletal motion sequences across more than 800 real-world driving scenarios, with an average of 27 interacting agents per scenario. The dataset utilizes LiDAR and motion prior technologies to address the occlusion and temporal continuity limitations of existing datasets, and provides a rich spectrum of multi-agent interaction patterns.




