rollout_stackblocks_iter2_correction
收藏资源简介:
该数据集使用LeRobot工具创建,是一个机器人操作任务相关的数据集,遵循Apache-2.0许可证。数据集包含25个训练episodes,总计20,838帧数据,涉及1个任务。数据以分块Parquet文件格式存储,并包含对应的视频文件。数据集的核心特征包括:机器人动作指令(action,6维浮点数组,对应肩部平移、肩部升降、肘部弯曲、腕部弯曲、腕部旋转和夹爪位置)、观测状态(observation.state,6维浮点数组,与动作指令维度一致)、来自顶部和腕部摄像头的图像观测(observation.images.top和observation.images.wrist,均为480x640分辨率、3通道的30fps AV1编码视频)、干预信号(intervention,布尔值)、下一时刻的成功标志(next.success,布尔值)、时间戳(timestamp)、帧索引(frame_index)、episode索引(episode_index)、全局索引(index)和任务索引(task_index)。视频数据总大小约200MB,结构化数据总大小约100MB。数据记录来自‘so_follower’型机器人,并已预定义为训练集分割。
This dataset is created using the LeRobot tool and is related to robot manipulation tasks, following the Apache-2.0 license. It contains 25 training episodes, totaling 20,838 frames of data, involving 1 task. The data is stored in chunked Parquet file format and includes corresponding video files. The core features of the dataset include: robot action commands (action, a 6-dimensional floating-point array corresponding to shoulder translation, shoulder elevation, elbow flexion, wrist flexion, wrist rotation, and gripper position), observation states (observation.state, a 6-dimensional floating-point array consistent with the action command dimensions), image observations from top and wrist cameras (observation.images.top and observation.images.wrist, both 480x640 resolution, 3-channel, 30fps AV1 encoded videos), intervention signals (intervention, boolean), next-step success flags (next.success, boolean), timestamps (timestamp), frame indices (frame_index), episode indices (episode_index), global indices (index), and task indices (task_index). The total video data size is approximately 200MB, and the structured data size is approximately 100MB. The data is recorded from a so_follower type robot and is pre-defined as a training set split.
数据集概述
- 数据集名称:rollout_stackblocks_iter2_correction
- 来源地址:https://huggingface.co/datasets/lilkm/rollout_stackblocks_iter2_correction
- 许可证:Apache-2.0
- 任务类别:机器人学 (Robotics)
- 创建工具:使用 LeRobot 创建
数据集结构
| 属性 | 描述 |
|---|---|
| 总片段数 (total_episodes) | 25 |
| 总帧数 (total_frames) | 20,838 |
| 总任务数 (total_tasks) | 1 |
| 帧率 (fps) | 30 |
| 数据文件大小 | 100 MB |
| 视频文件大小 | 200 MB |
| 机器人类型 | so_follower |
| 数据划分 | 训练集 (train):0-25 片段 |
数据特征 (Features)
- 动作 (action):6维浮点数组,包含各关节位置 (shoulder_pan, shoulder_lift, elbow_flex, wrist_flex, wrist_roll, gripper)
- 观测状态 (observation.state):6维浮点数组,与动作相同的关节位置
- 观测图像 (observation.images):
- top:顶部摄像头图像,分辨率 480×640×3,视频编码 av1,帧率 30fps
- wrist:腕部摄像头图像,分辨率 480×640×3,视频编码 av1,帧率 30fps
- 干预标志 (intervention):布尔值,表示是否有干预
- 下一帧成功标志 (next.success):布尔值,表示下一帧是否成功
- 时间戳 (timestamp):浮点数
- 帧索引 (frame_index):整数
- 片段索引 (episode_index):整数
- 索引 (index):整数
- 任务索引 (task_index):整数
文件路径
- 数据文件路径:
data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet - 视频文件路径:
videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4
引用信息
当前无专门的引用信息,如需引用可使用数据集首页链接。




