so101_stacking_big_rings
收藏Hugging Face2026-06-04 更新2026-06-02 收录
下载链接:
https://huggingface.co/datasets/lorenzouttini/so101_stacking_big_rings
下载链接
链接失效反馈官方服务:
资源简介:
该数据集是一个机器人操作数据集,基于LeRobot框架构建,专门针对so_follower机器人类型。它包含100个完整的情节(episodes),总计24225帧数据,用于单一任务(推测为堆叠大环任务)。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,帧率为30fps。数据集的特征包括:动作空间(6维浮点数,对应机器人肩部平移、肩部提升、肘部弯曲、腕部弯曲、腕部旋转和夹爪位置);状态观测(与动作空间相同的6维关节位置);图像观测(前视摄像头和腕部摄像头的视频数据,分辨率均为480x640,3通道RGB,30fps,使用av1编码);以及元数据如时间戳、帧索引、情节索引和任务索引。数据集仅包含训练集,所有数据用于训练。
This dataset is a robotic manipulation dataset built on the LeRobot framework, specifically designed for the so_follower robot type. It contains 100 full episodes, totaling 24225 frames of data for a single task (presumably the large ring stacking task). The data is stored in Parquet format, with a total data file size of 100 MB and a video file size of 200 MB, at a frame rate of 30 fps. The dataset includes the following features: action space (6-dimensional floating-point values, corresponding to robot shoulder translation, shoulder elevation, elbow flexion, wrist flexion, wrist rotation, and gripper position); state observations (6-dimensional joint positions matching the action space); image observations (video data from a front-facing camera and a wrist camera, both with a resolution of 480×640, 3-channel RGB, 30 fps, encoded using AV1); as well as metadata such as timestamps, frame indices, episode indices, and task indices. The dataset only includes a training split, with all data used for training.
提供机构:
lorenzouttini创建时间:
2026-06-02
原始信息汇总
数据集概述
- 数据集名称:so101_stacking_big_rings
- 许可协议:Apache-2.0
- 任务类别:机器人学 (Robotics)
- 创建工具:基于 LeRobot 框架创建
数据集结构
| 属性 | 详情 |
|---|---|
| codebase_version | v3.0 |
| robot_type | so_follower |
| 总片段数 (total_episodes) | 100 |
| 总帧数 (total_frames) | 24225 |
| 总任务数 (total_tasks) | 1 |
| 分块大小 (chunks_size) | 1000 |
| 数据文件大小 | 100 MB |
| 视频文件大小 | 200 MB |
| 帧率 (fps) | 30 |
| 训练/测试拆分 | 训练集:片段0至99(共100个) |
数据特征
动作与状态特征
- action:6维浮点数组,对应6个自由度的位置信息(肩部旋转、肩部升降、肘部弯曲、腕部弯曲、腕部旋转、夹爪位置)
- observation.state:与动作特征相同的6维状态信息
图像观测
- observation.images.front:前置摄像头视频,分辨率480×640,3通道(RGB),编码格式AV1,帧率30fps
- observation.images.wrist:腕部摄像头视频,分辨率480×640,3通道(RGB),编码格式AV1,帧率30fps
其他特征
- timestamp:时间戳(float32)
- frame_index:帧索引(int64)
- episode_index:片段索引(int64)
- index:全局索引(int64)
- task_index:任务索引(int64)
数据存储路径
- 数据文件:
data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet - 视频文件:
videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4



