tersoohaan/eval_pi0_light
收藏资源简介:
该数据集是通过LeRobot创建的机器人数据集,专门用于机器人技术研究。数据集包含一个名为so101_follower的机器人类型,总共有15个episodes和8332帧,帧率为30fps。数据特征包括动作(6个关节位置: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)、观测状态(与动作相同的6个关节位置)、前视图像(RGB格式,分辨率480x640,3通道)和前视深度图像(uint16格式,分辨率480x640),以及顶视图像(RGB格式,分辨率480x640,3通道)。此外,数据集还包含时间戳、帧索引、episode索引、索引和任务索引等元数据。数据以parquet文件格式存储,视频以mp4格式存储,总数据文件大小为100MB,视频文件大小为500MB。数据集仅包含训练分割(episodes 0到15),适用于机器人控制、模仿学习或强化学习等任务。
This dataset is a robotics dataset created using LeRobot, designed for robotics research. It features a robot type named so101_follower, with a total of 15 episodes and 8332 frames at 30fps. The data includes actions (6 joint positions: shoulder pan, shoulder lift, elbow flex, wrist flex, wrist roll, and gripper positions), observation states (same 6 joint positions as actions), front-view images (RGB format, 480x640 resolution, 3 channels) and front-view depth images (uint16 format, 480x640 resolution), as well as top-view images (RGB format, 480x640 resolution, 3 channels). Additionally, the dataset contains metadata such as timestamp, frame index, episode index, index, and task index. The data is stored in parquet file format, with videos in mp4 format; the total data file size is 100MB, and video file size is 500MB. The dataset includes only a training split (episodes 0 to 15) and is suitable for tasks like robot control, imitation learning, or reinforcement learning.




