BEHAVE
收藏资源简介:
BEHAVE数据集是由德国图宾根大学和马克斯普朗克信息学研究所共同创建的,旨在捕捉和分析人类与物体在自然环境中的交互行为。该数据集包含约15200帧的多视角RGBD图像,记录了8名参与者与20种常见物体在5个不同地点的交互情况。数据集不仅捕捉了人类和物体的3D模型,还包括了交互过程中的接触点标注。BEHAVE数据集的创建过程涉及使用便携式多摄像头设置,通过神经网络模型预测人类和物体之间的对应关系,从而实现对交互行为的精确跟踪。该数据集的应用领域广泛,包括游戏开发、虚拟与混合现实技术、人类行为分析以及人机协作等,旨在解决复杂环境下的交互行为建模问题。
The BEHAVE dataset was co-created by the University of Tübingen in Germany and the Max Planck Institute for Informatics, aiming to capture and analyze human-object interactions in natural environments. It contains approximately 15,200 frames of multi-view RGBD images, recording interactions between 8 participants and 20 common objects across 5 different locations. The dataset not only captures 3D models of humans and objects but also includes contact point annotations during interactions. The creation of the BEHAVE dataset involves using a portable multi-camera setup and neural network models to predict correspondences between humans and objects, thereby achieving precise tracking of interactive behaviors. The BEHAVE dataset has a wide range of application scenarios, including game development, virtual and mixed reality technologies, human behavior analysis, human-computer collaboration and more, aiming to address the problem of interactive behavior modeling in complex environments.

- 1BEHAVE: Dataset and Method for Tracking Human Object Interactions德国图宾根大学 · 2022年



