HyeonseokE/SO101-cap_stack_RGBblock_on_bluedish_10fps_40epi
收藏资源简介:
该数据集是一个用于机器人学任务的数据集,使用LeRobot创建。数据集包含40个总剧集、34780个总帧和1个总任务,帧率为10fps。数据以Parquet文件格式存储,总数据文件大小为100MB,视频文件大小为200MB。数据集结构丰富,包括观测状态(如机器人关节位置)、动作命令、图像观测(来自顶部和左腕摄像头的RGB视频,分辨率为480x640,帧率10fps)、末端执行器位置(XYZRPY坐标)、二进制夹爪状态、技能信息(自然语言描述、验证问题、技能类型、进度和目标位置)、子任务信息(自然语言描述、对象名称和目标位置),以及时间戳、帧索引、剧集索引等元数据。数据集专为训练和评估机器人控制模型设计,适用于模仿学习、强化学习等场景。
This dataset is designed for robotics tasks and was created using LeRobot. It contains a total of 40 episodes, 34,780 frames, and 1 task, with a frame rate of 10fps. The data is stored in Parquet format, with a total data file size of 100MB and video file size of 200MB. The dataset features a rich structure, including observation states (such as robot joint positions), action commands, image observations (RGB videos from top and left wrist cameras, with a resolution of 480x640 and 10fps frame rate), end-effector positions (XYZRPY coordinates), binary gripper state, skill information (natural language descriptions, verification questions, skill types, progress, and goal positions), subtask information (natural language descriptions, object names, and target positions), as well as metadata like timestamps, frame indices, and episode indices. It is specifically designed for training and evaluating robot control models, suitable for scenarios such as imitation learning and reinforcement learning.




