HyeonseokE/SO101-cap_close_drawer_10fps_60epi
收藏资源简介:
该数据集是一个用于机器人学习的数据集,专门针对机械臂操作任务。它包含60个训练片段,总计14061帧,数据以.parquet格式存储,视频以.mp4格式存储,帧率为10fps。数据集提供了丰富的多模态特征,包括机器人关节状态观测(如肩部平移、肩部提升、肘部弯曲、腕部弯曲、腕部旋转和夹爪位置)、动作指令、图像观测(来自顶部和左腕摄像头的RGB视频,分辨率为480x640)、末端执行器位置(XYZ坐标和欧拉角)、夹爪二进制状态、技能信息(如自然语言描述、验证问题、类型、进度和目标位置)、子任务信息(如自然语言描述、对象名称和目标位置)以及时间戳和索引数据。这些数据适用于机器人控制、模仿学习或强化学习任务,特别是涉及抽屉关闭等操作场景。数据集基于LeRobot平台创建,采用Apache 2.0许可证。
This dataset is designed for robotics learning, specifically for robotic arm manipulation tasks. It contains 60 training episodes with a total of 14,061 frames, stored in .parquet format for data and .mp4 format for videos, at a frame rate of 10fps. The dataset offers rich multimodal features, including robot joint state observations (e.g., shoulder pan, shoulder lift, elbow flex, wrist flex, wrist roll, and gripper positions), action commands, image observations (RGB videos from top and left wrist cameras with 480x640 resolution), end-effector positions (XYZ coordinates and Euler angles), gripper binary states, skill information (e.g., natural language descriptions, verification questions, types, progress, and goal positions), subtask information (e.g., natural language descriptions, object names, and target positions), as well as timestamps and index data. It is suitable for robotics control, imitation learning, or reinforcement learning tasks, particularly in scenarios like drawer closing operations. The dataset was created using the LeRobot platform and is licensed under Apache 2.0.




