360◦-Motion Dataset
收藏资源简介:
360◦-Motion Dataset是由香港中文大学和快手科技合作创建的一个用于视频生成的高质量3D运动数据集。该数据集通过收集3D人类和动物资产,并使用GPT生成轨迹模板,结合12个均匀分布的相机在多样化的3D场景中捕捉运动。数据集的创建过程包括资产的统一缩放、轨迹生成、场景布置和运动捕捉,旨在解决视频生成中多实体3D运动控制的挑战。该数据集主要应用于虚拟电影制作、游戏中的3D虚拟场景探索以及实体AI系统的世界模型构建。
360°-Motion Dataset is a high-quality 3D motion dataset for video generation, jointly created by The Chinese University of Hong Kong and Kuaishou Technology. This dataset is constructed by collecting 3D human and animal assets, generating trajectory templates with GPT, and capturing motions in diverse 3D scenes using 12 uniformly distributed cameras. The dataset creation workflow encompasses unified asset scaling, trajectory generation, scene arrangement, and motion capture, with the goal of addressing the challenge of multi-entity 3D motion control in video generation. This dataset is primarily applied in virtual filmmaking, 3D virtual scene exploration in video games, and world model construction for embodied AI systems.




