clean-desk-0702
收藏Hugging Face2026-07-05 更新2026-07-05 收录
下载链接:
https://huggingface.co/datasets/TacVerse/clean-desk-0702
下载链接
链接失效反馈官方服务:
资源简介:
该数据集是一个使用LeRobot工具创建的机器人任务数据集,专门为双触觉抓取器(bi_taccap_gripper)设计。它包含10个完整任务序列(episodes),总计17582帧数据,帧率为30fps,数据以Parquet格式存储。数据特征包括:动作数据(20维浮点数,用于控制左右机械臂的工具中心点位置、旋转和夹爪位置)、观测状态数据(与动作相同维度)、以及来自左右触觉传感器(左右各两个,分辨率400x700)和左右腕部摄像头(分辨率480x640)的视频观测,所有视频均为彩色(3通道),编码为h264。此外,数据集还包含时间戳、帧索引、episode索引等元数据。该数据集适用于机器人控制或学习任务,但具体任务内容未在README中明确描述。
This dataset is a robot task dataset created using the LeRobot tool, specifically designed for a bi-tactile gripper (bi_taccap_gripper). It contains 10 complete task episodes, totaling 17,582 frames of data at a frame rate of 30fps, stored in Parquet format. The data features include: action data (20-dimensional floating-point numbers controlling the tool center point position, rotation, and gripper position of the left and right robotic arms), observation state data (with the same dimensions as the action), and video observations from left and right tactile sensors (two on each side, with a resolution of 400x700) and left and right wrist cameras (resolution 480x640). All videos are in color (3 channels) and encoded as h264. Additionally, the dataset includes metadata such as timestamps, frame indices, and episode indices. It is suitable for robot control or learning tasks, but the specific task content is not explicitly described in the README.
提供机构:
TacVerse创建时间:
2026-07-02
原始信息汇总
数据集概述
- 数据集名称:clean-desk-0702
- 许可证:Apache-2.0
- 任务类别:机器人学
- 标签:LeRobot
- 创建工具:LeRobot
数据集统计
| 指标 | 数值 |
|---|---|
| 总片段数 | 10 |
| 总帧数 | 17,582 |
| 总任务数 | 1 |
| 帧率 | 30 FPS |
| 数据文件大小 | 100 MB |
| 视频文件大小 | 200 MB |
| 分块大小 | 1,000 |
| 代码库版本 | v3.0 |
| 机器人类型 | bi_taccap_gripper |
| 训练集分割 | 0:10 |
数据集结构
数据路径
- 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
特征说明
| 特征名称 | 数据类型 | 形状 | 描述 |
|---|---|---|---|
| action | float32 | (20,) | 包含左右TCP位姿、旋转和夹爪位置 |
| observation.state | float32 | (20,) | 与action结构相同的状态观测 |
| observation.images.left_tactile_left | video | (400, 700, 3) | 左侧触觉左摄像头,H.264编码 |
| observation.images.left_tactile_right | video | (400, 700, 3) | 左侧触觉右摄像头,H.264编码 |
| observation.images.left_wrist | video | (480, 640, 3) | 左手腕摄像头,H.264编码 |
| observation.images.right_tactile_left | video | (400, 700, 3) | 右侧触觉左摄像头,H.264编码 |
| observation.images.right_tactile_right | video | (400, 700, 3) | 右侧触觉右摄像头,H.264编码 |
| observation.images.right_wrist | video | (480, 640, 3) | 右手腕摄像头,H.264编码 |
| timestamp | float32 | (1,) | 时间戳 |
| frame_index | int64 | (1,) | 帧索引 |
| episode_index | int64 | (1,) | 片段索引 |
| index | int64 | (1,) | 索引 |
| task_index | int64 | (1,) | 任务索引 |
Action / State 命名空间
共通20维向量命名如下(左右各10维):
- left_tcp.x ~ left_tcp.r6(6维TCP位姿)
- left_gripper.pos(夹爪位置)
- right_tcp.x ~ right_tcp.r6(6维TCP位姿)
- right_gripper.pos(夹爪位置)
引用
如需引用该数据集,请使用以下BibTeX格式:
bibtex @misc{xense-taccap-lerobot, author = {XenseRobotics Team}, title = {LeRobot-Xense: LeRobot with Xense Tactile Robotics Support}, howpublished = {url{https://github.com/Vertax42/xense-taccap-lerobot}}, year = {2026} }



