sanjar-normuradov-agile-robots/austin-sirius

该数据集是一个用于机器人学习的数据集，包含559个episodes和279,939帧数据。数据集包含机器人状态（如末端执行器位姿、关节状态、夹爪宽度）、动作（如位置、姿态、夹爪控制）以及图像数据（如普通图像和腕部图像）。数据预处理包括平滑零状态样本、过滤前10帧以及夹爪值的映射。数据集结构通过meta/info.json文件详细描述，包括数据类型、形状、名称等。数据集适用于机器人学习和人机协作研究。

This dataset is designed for robot learning, containing 559 episodes and 279,939 frames. It includes robot states (e.g., end-effector pose, joint states, gripper width), actions (e.g., position, orientation, gripper control), and image data (e.g., regular images and wrist images). Pre-processing steps include smoothing zero-state samples, filtering the first 10 frames, and mapping gripper values. The dataset structure is detailed in meta/info.json, including data types, shapes, and names. It is suitable for robot learning and human-in-the-loop research.