so101-cube-put-20260706
收藏资源简介:
该数据集是使用LeRobot创建的机器人操作数据集,专门用于训练和评估机器人控制算法。数据集包含50个完整任务片段,总计20573帧数据,采用so_follower机器人平台。数据特征包括:6维关节位置的动作指令和观测状态;两个摄像头视角(顶部和腕部)的RGB视频流,分辨率为720x1280,帧率为30fps;以及时间戳、帧索引、片段索引等元数据。数据集以parquet格式存储,适用于机器人模仿学习和强化学习研究。
This dataset is a robotic manipulation dataset developed with LeRobot, specifically designed for training and evaluating robotic control algorithms. It contains 50 complete task episodes, totaling 20,573 frames collected using the so_follower robotic platform. The dataset features include 6-dimensional joint position action commands and observation states; RGB video streams from two camera perspectives (top and wrist-mounted) with a resolution of 720×1280 and a frame rate of 30 fps; as well as metadata such as timestamps, frame indices, and episode indices. Stored in Parquet format, this dataset is suitable for research on robotic imitation learning and reinforcement learning.
数据集概述
数据集名称: so101-cube-put-20260706
许可证: Apache-2.0
任务类别: 机器人技术(Robotics)
标签: LeRobot
数据集详情
- 机器人类型: so_follower
- 总片段数(Episodes): 50
- 总帧数(Frames): 20,573
- 总任务数(Tasks): 1
- 数据分块大小(Chunks Size): 1,000
- 数据文件大小: 100 MB
- 视频文件大小: 200 MB
- 帧率(FPS): 30
- 数据集划分: 仅包含训练集(train: 0:50)
数据组成
数据集包含以下两种模态:
-
动作(Action)与观测状态(Observation.State)
- 数据类型: float32
- 维度: 6
- 特征名称: shoulder_pan.pos, shoulder_lift.pos, elbow_flex.pos, wrist_flex.pos, wrist_roll.pos, gripper.pos
-
观测图像(Observation.Images)
- 两个摄像头视角:
- top(顶部视角): 分辨率 720×1280, 3通道, H.264编码, 30 FPS
- wrist(腕部视角): 分辨率 720×1280, 3通道, H.264编码, 30 FPS
- 两个摄像头视角:
元数据字段
- timestamp: float32, 形状 [1]
- frame_index: int64, 形状 [1]
- episode_index: int64, 形状 [1]
- index: int64, 形状 [1]
- task_index: int64, 形状 [1]
数据路径
- 数据文件路径格式:
data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet - 视频文件路径格式:
videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4
创建工具
该数据集使用 LeRobot 框架创建(GitHub 仓库)。



