klvarshaa/so101_barcode_scan_100ep_20260528_142840
收藏资源简介:
该数据集是一个用于机器人学习的演示数据集,专门针对条形码扫描任务。它包含100个episode,总计3211帧数据,由LeRobot工具创建。数据集记录了机器人(类型为so_follower)在执行任务过程中的动作和状态:动作数据包括6个关节(肩部平移、肩部升降、肘部弯曲、腕部弯曲、腕部旋转、夹爪)的位置;状态观测同样包含这6个关节的位置。此外,数据集提供了多视角视觉观测,包括顶部、侧面和腕部三个摄像头拍摄的视频,每帧图像分辨率为480x640,帧率为30fps,采用AV1编码。数据集还包含时间戳、帧索引、episode索引等元数据。数据以Parquet格式存储,总数据文件大小为100MB,视频文件大小为200MB。数据集适用于机器人控制、模仿学习等研究任务,采用Apache 2.0许可证。
This dataset is a demonstration dataset for robotics learning, specifically for barcode scanning tasks. It contains 100 episodes with a total of 3211 frames, created using the LeRobot tool. The dataset records the actions and states of a robot (type so_follower) during task execution: action data includes positions of 6 joints (shoulder pan, shoulder lift, elbow flex, wrist flex, wrist roll, gripper); state observations also include positions of these 6 joints. Additionally, the dataset provides multi-view visual observations from three cameras (top, side, and wrist), with each video frame at 480x640 resolution, 30 fps, and AV1 encoding. The dataset also includes metadata such as timestamps, frame indices, episode indices, etc. Data is stored in Parquet format, with a total data file size of 100MB and video file size of 200MB. The dataset is suitable for robotics control, imitation learning, and related research tasks, licensed under Apache 2.0.




