FORCE
收藏资源简介:
FORCE数据集是由图宾根大学与图宾根人工智能中心等机构合作创建的,专注于捕捉和模拟人类与物体交互中的物理属性。该数据集包含450个动作序列,总计192,000帧,涵盖了推、拉和携带等多种交互动作,每帧都提供了高质量的人体和物体姿态。数据集通过结合4Kinect RGB-D相机和17个惯性测量单元(IMUs)进行数据采集,确保了动作捕捉的精确性。FORCE数据集不仅支持多样化的动作风格和不同级别的阻力模拟,还通过新颖的直观物理编码方法,增强了模型的物理交互真实感。该数据集的应用领域广泛,包括增强现实/虚拟现实、游戏开发以及人机交互等,旨在解决现有模型在处理复杂物理交互时的局限性。
The FORCE dataset was collaboratively developed by the University of Tübingen, Tübingen AI Center, and other institutions, focusing on capturing and simulating physical properties during human-object interactions. This dataset includes 450 motion sequences, totaling 192,000 frames, covering a variety of interactive actions such as pushing, pulling, and carrying, with high-quality human and object poses provided for every frame. Data collection was conducted using a combination of 4 Kinect RGB-D cameras and 17 inertial measurement units (IMUs), which ensures the precision of motion capture. The FORCE dataset not only supports diverse motion styles and resistance simulations at different levels, but also enhances the realism of physical interactions for models through a novel and intuitive physical encoding method. This dataset has broad application scenarios including augmented reality/virtual reality, game development, human-computer interaction, and others, aiming to address the limitations of existing models when dealing with complex physical interactions.

- 1FORCE: Dataset and Method for Intuitive Physics Guided Human-object Interaction图宾根大学与图宾根人工智能中心 · 2024年



