RBO Dataset of Articulated Objects and Interactions

RBO数据集是由德国柏林工业大学的机器人与生物学实验室创建，包含14种常见于人类环境的关节对象及其交互的RGB-D视频序列和力矩数据。数据集总计包含358个交互序列，时长67分钟，涵盖多种实验条件。每个交互序列均标注有刚体部分的真值姿态和由动作捕捉系统获得的关节状态。数据集旨在帮助评估和开发基于视觉数据感知关节对象的算法，适用于机器人操作和交互研究。

The RBO Dataset was created by the Robotics and Biology Laboratory at Technische Universität Berlin, Germany. It contains 14 types of articulated objects commonly found in human environments, along with RGB-D video sequences and torque data corresponding to their interactions. In total, the dataset includes 358 interaction sequences with a total duration of 67 minutes, covering diverse experimental conditions. Each interaction sequence is annotated with ground-truth poses of its rigid-body components and joint states acquired via a motion capture system. This dataset is intended to facilitate the evaluation and development of algorithms for perceiving articulated objects from visual data, and is applicable to research in robotic manipulation and interaction.