遇见数据集

take-apart-and-reassemble-vacuum-cleaner-0703

收藏
Hugging Face2026-07-05 更新2026-07-05 收录
官方服务:

资源简介:

该数据集是一个机器人学数据集,使用LeRobot工具创建,专注于双机械手触觉感知与操作任务。它包含7个完整的情节(episodes),总计27286帧,帧率为30 fps,专门设计用于拆卸和重新组装吸尘器(基于数据集名称推断)。数据采用Apache 2.0许可证。数据集结构包括动作和观察特征:动作特征是一个20维浮点向量,涵盖左右机械手的工具中心点(TCP)位置(x, y, z)、旋转参数(r1至r6)以及夹爪位置;观察特征包括类似的状态向量和多个图像流,如左右触觉传感器的左右图像(分辨率400x700,RGB通道)和左右手腕摄像头图像(分辨率480x640,RGB通道),所有图像均以h264编码的视频格式存储。数据集还包含时间戳、帧索引、情节索引等元数据,用于序列跟踪。数据以parquet文件分块组织,总数据文件大小为100 MB,视频文件大小为200 MB,仅提供训练分割(索引0到7)。该数据集适用于机器人学习、触觉感知和双手机械手控制的研究与应用。

This dataset is a robotics dataset created using the LeRobot tool, focusing on dual-arm tactile perception and manipulation tasks. It contains 7 complete episodes, totaling 27,286 frames at a frame rate of 30 fps, specifically designed for the task of disassembling and reassembling a vacuum cleaner (inferred from the dataset name). The data is licensed under Apache 2.0. The dataset structure includes action and observation features: the action feature is a 20-dimensional floating-point vector covering the tool center point (TCP) positions (x, y, z), rotation parameters (r1 to r6), and gripper positions for both left and right manipulators; the observation features include similar state vectors and multiple image streams, such as left and right images from tactile sensors (resolution 400x700, RGB channels) and left and right wrist camera images (resolution 480x640, RGB channels), all stored in h264-encoded video format. The dataset also includes metadata like timestamps, frame indices, and episode indices for sequence tracking. Data is organized in chunked parquet files, with a total data file size of 100 MB and video file size of 200 MB, providing only a training split (indices 0 to 7). This dataset is suitable for research and applications in robotics learning, tactile perception, and dual-arm manipulator control.

提供机构:
TacVerse
创建时间:
2026-07-03
原始信息汇总

数据集概述

数据集名称:take-apart-and-reassemble-vacuum-cleaner-0703
数据集地址:https://huggingface.co/datasets/TacVerse/take-apart-and-reassemble-vacuum-cleaner-0703
许可证:Apache-2.0
任务类别:机器人(Robotics)
创建工具:LeRobot(https://github.com/huggingface/lerobot)


数据集结构

  • 机器人类型:bi_taccap_gripper
  • 总片段数(Episodes):7
  • 总帧数:27286
  • 总任务数:1
  • Chunk 大小:1000
  • 数据文件大小:100 MB
  • 视频文件大小:200 MB
  • 帧率(FPS):30
  • 数据集划分:训练集(train)包含全部 7 个片段(索引 0:7)

特征(Features)

动作(action)

  • 数据类型:float32
  • 形状:[20]
  • 包含字段(左右各10个):
    • 左臂left_tcp.x, left_tcp.y, left_tcp.z, left_tcp.r1...r6, left_gripper.pos
    • 右臂right_tcp.x, right_tcp.y, right_tcp.z, right_tcp.r1...r6, right_gripper.pos

观察状态(observation.state)

  • 数据类型:float32
  • 形状:[20]
  • 字段:与 action 完全一致(左右臂TCP位置、姿态及夹爪位置)

图像观测(observation.images)

共包含6个摄像头视角,均为视频格式(H.264,30FPS,无音频):

特征名 分辨率 通道数
left_tactile_left 400×700 3 (RGB)
left_tactile_right 400×700 3 (RGB)
left_wrist 480×640 3 (RGB)
right_tactile_left 400×700 3 (RGB)
right_tactile_right 400×700 3 (RGB)
right_wrist 480×640 3 (RGB)

其他特征

  • timestamp(float32):时间戳
  • frame_index(int64):帧索引
  • episode_index(int64):片段索引
  • index(int64):全局索引
  • task_index(int64):任务索引

引用

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} }

二维码
社区交流群
二维码
科研交流群
商业服务