so101_lego_20260613_20260613_131100
收藏资源简介:
该数据集由LeRobot平台创建,专为机器人技术任务设计,包含机器人操作任务的演示数据,涵盖状态观测、动作执行及视觉感知。数据集规模为157个任务片段(episodes)和95234个时间步(frames),以Parquet文件格式存储,并包含MP4格式的视频文件。核心特征包括:动作空间(action)为6维浮点数组,对应肩部平移、肩部抬升、肘部弯曲、腕部弯曲、腕部旋转和夹爪位置;状态观测(observation.state)为6维浮点数组,记录机器人各关节实时位置;视觉观测(observation.images.front)提供前向摄像头视频流,分辨率为480x640,3通道彩色图像,编码格式为AV1,帧率为30fps。此外,数据集包含时间戳、帧索引、片段索引、全局索引和任务索引等元数据字段,数据按块(chunks)组织,每块包含1000个frames。适用于机器人模仿学习、强化学习、行为克隆及视觉-动作策略学习等研究与应用场景。
This dataset is created by the LeRobot platform and designed for robotics tasks, containing demonstration data for robot manipulation tasks that encompass state observations, action execution, and visual perception. The dataset scale includes 157 complete task episodes and 95,234 timesteps (frames), stored in Parquet file format along with MP4 video files. Core features include: action space (action) as a 6-dimensional floating-point array corresponding to shoulder translation, shoulder elevation, elbow flexion, wrist flexion, wrist rotation, and gripper position; state observation (observation.state) as a 6-dimensional floating-point array recording real-time joint positions of the robot; visual observation (observation.images.front) providing a front-facing camera video stream with a resolution of 480x640, 3-channel color images, AV1 encoding format, and a frame rate of 30fps. Additionally, the dataset includes metadata fields such as timestamps, frame indices, episode indices, global indices, and task indices. Data is organized into chunks, with each chunk containing 1,000 frames. It is suitable for research and application scenarios like robot imitation learning, reinforcement learning, behavior cloning, and vision-action policy learning.
数据集概述
- 数据集名称:
emboss369/so101_lego_20260613_20260613_131100 - 许可证: Apache-2.0
- 任务类别: 机器人学 (Robotics)
- 标签: LeRobot
数据集来源与创建
- 该数据集使用 LeRobot 框架创建。
- 机器人类型:
so_follower
数据集规模
- 总集数 (Episodes): 157
- 总帧数 (Frames): 95,234
- 总任务数 (Tasks): 1
- 帧率 (FPS): 30
- 数据文件大小: 100 MB
- 视频文件大小: 200 MB
数据划分
- 训练集 (Train): 所有 157 集 (索引 0 到 156)
数据结构
数据集包含以下特征字段:
| 字段名 | 数据类型 | 形状 | 说明 |
|---|---|---|---|
action |
float32 | (6,) | 机器人动作,包含6个关节位置 (肩部回转、肩部升降、肘部弯曲、腕部弯曲、腕部旋转、夹爪) |
observation.state |
float32 | (6,) | 机器人观测状态,与动作维度相同 |
observation.images.front |
video | (480, 640, 3) | 前置摄像头视频,编码格式为 AV1,像素格式 yuv420p,帧率30 |
timestamp |
float32 | (1,) | 时间戳 |
frame_index |
int64 | (1,) | 帧索引 |
episode_index |
int64 | (1,) | 集索引 |
index |
int64 | (1,) | 整体索引 |
task_index |
int64 | (1,) | 任务索引 |
数据存储格式:
- 表格数据存储为 Parquet 文件,路径格式:
data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet - 视频文件存储为 MP4,路径格式:
videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4
代码库版本: v3.0
引用
暂无提供 BibTeX 引用信息。




