Task-one-2nd-onebag
收藏资源简介:
该数据集使用LeRobot工具创建,是一个机器人操作数据集,适用于机器人学习任务。数据集包含一个名为so100_wowskin的机器人平台收集的演示数据,数据规模为100个完整episodes,总计18,161帧,对应1个任务。数据以分块形式存储,包含约100MB的parquet格式数据文件和200MB的MP4视频文件。数据集提供30FPS的多模态观测和动作记录。主要特征包括:1) 动作空间:6维浮点数组,控制机器人的肩部平移、肩部升降、肘部弯曲、腕部弯曲、腕部旋转和夹爪位置;2) 观测状态:6维浮点数组,记录上述关节的位置状态;3) 观测wowskin:15维浮点数组;4) 图像观测:包含基座摄像头和腕部摄像头两个视角的RGB视频流,分辨率均为480x640,3通道,编码格式为AV1;5) 元数据:时间戳、帧索引、episode索引、数据索引和任务索引。数据集仅包含训练集,所有episodes均用于训练。
This dataset was created using the LeRobot toolkit and is a robotic manipulation dataset tailored for robot learning tasks. It contains demonstration data collected by the so100_wowskin robotic platform, with a scale of 100 full episodes totaling 18,161 frames, corresponding to one single task. The dataset is stored in chunked format, including approximately 100 MB of parquet-formatted data files and 200 MB of MP4 video files. It provides 30 FPS multimodal observation and action recordings. The main features are as follows: 1. Action space: A 6-dimensional floating-point array that controls the robot's shoulder translation, shoulder elevation, elbow flexion, wrist flexion, wrist rotation, and gripper position; 2. Observation state: A 6-dimensional floating-point array that records the position states of the aforementioned joints; 3. Wowskin observation: A 15-dimensional floating-point array; 4. Image observations: RGB video streams from two perspectives, namely the base camera and the wrist camera, both with a resolution of 480×640 and 3 color channels, encoded in the AV1 format; 5. Metadata: Timestamps, frame indices, episode indices, data indices, and task indices. The dataset only contains a training split, and all episodes are utilized for training.
数据集概述
- 名称: Task-one-2nd-onebag
- 许可证: Apache-2.0
- 任务类别: 机器人学 (Robotics)
- 标签: LeRobot
数据集来源与用途
该数据集使用 LeRobot 创建,旨在为机器人学习研究提供训练数据。
数据集配置
- 配置名称: default
- 数据文件:
data/*/*.parquet
数据集结构
元数据摘要(来自 meta/info.json)
| 属性 | 值 |
|---|---|
| 代码库版本 | v3.0 |
| 机器人类型 | so100_wowskin |
| 总片段数 (Episodes) | 100 |
| 总帧数 (Frames) | 18,161 |
| 总任务数 | 1 |
| 块大小 | 1000 |
| 数据文件总大小 | 100 MB |
| 视频文件总大小 | 200 MB |
| 帧率 (FPS) | 30 |
| 数据路径 | data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet |
| 视频路径 | videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4 |
数据划分
- 训练集: 片段索引 0 到 99(共 100 个片段)
特征字段
| 字段名 | 数据类型 | 形状 | 说明 |
|---|---|---|---|
action |
float32 | [6] | 动作:包括 6 个关节位置(shoulder_pan, shoulder_lift, elbow_flex, wrist_flex, wrist_roll, gripper) |
observation.state |
float32 | [6] | 观察状态:6 个关节位置,与 action 命名相同 |
observation.wowskin |
float32 | [15] | 触觉传感数据(wowskin),维度为 15 |
observation.images.base |
video (AV1) | [480, 640, 3] | 基础摄像头图像:高度 480,宽度 640,3 通道,30 FPS |
observation.images.wrist |
video (AV1) | [480, 640, 3] | 腕部摄像头图像:高度 480,宽度 640,3 通道,30 FPS |
timestamp |
float32 | [1] | 时间戳 |
frame_index |
int64 | [1] | 帧索引 |
episode_index |
int64 | [1] | 片段索引 |
index |
int64 | [1] | 全局索引 |
task_index |
int64 | [1] | 任务索引 |




