遇见数据集

taccap-g1-medicine-bottle-label-orientation-and-batch-number-verification-0715

收藏
Hugging Face2026-07-15 更新2026-07-15 收录
官方服务:

资源简介:

该数据集是一个机器人任务数据集,使用LeRobot创建,专门用于双夹持器机器人(bi_taccap_gripper)执行操作任务,具体任务为药品瓶标签方向和批次号验证。数据集包含17个完整的情节(episodes),总计82766帧数据,帧率为30 fps。数据特征包括动作(action)和观测(observation),其中动作涉及左右夹持器的TCP(工具中心点)位置、旋转和夹持器位置,共20个浮点数值;观测包括状态观测(与动作类似的结构)和图像观测,图像观测来自多个视角:左触觉左传感器、左触觉右传感器、左腕部摄像头、右触觉左传感器、右触觉右传感器、右腕部摄像头,这些图像以视频形式存储,分辨率分别为400x700(触觉传感器)和480x640(腕部摄像头),均为彩色图像(3通道)。此外,数据集还包含时间戳、帧索引、情节索引等元数据。所有数据以Parquet格式存储,视频以MP4格式存储,适用于机器人学习、触觉感知和操作任务的研究。

This dataset is a robot task dataset created using LeRobot, specifically designed for a bi-tacap gripper robot to perform operational tasks, with the specific task being verification of medicine bottle label orientation and batch numbers. It contains 17 complete episodes, totaling 82,766 frames at a frame rate of 30 fps. The data features include actions and observations: actions involve the TCP (Tool Center Point) positions, rotations, and gripper positions for both left and right grippers, consisting of 20 floating-point values; observations include state observations (structured similarly to actions) and image observations from multiple perspectives: left tactile left sensor, left tactile right sensor, left wrist camera, right tactile left sensor, right tactile right sensor, and right wrist camera. These images are stored as videos with resolutions of 400x700 (tactile sensors) and 480x640 (wrist cameras), all in color (3 channels). Additionally, the dataset includes metadata such as timestamps, frame indices, and episode indices. All data is stored in Parquet format, with videos in MP4 format, making it suitable for research in robot learning, tactile perception, and manipulation tasks.

提供机构:
TacVerse
创建时间:
2026-07-15
原始信息汇总

数据集概述

  • 数据集名称: taccap-g1-medicine-bottle-label-orientation-and-batch-number-verification-0715
  • 许可证: Apache-2.0
  • 任务类型: 机器人技术
  • 标签: LeRobot
  • 创建工具: LeRobot

数据集结构

  • 代码库版本: v3.0
  • 机器人类型: bi_taccap_gripper
  • 总片段数: 17
  • 总帧数: 82766
  • 总任务数: 1
  • 数据块大小: 1000
  • 数据文件大小: 100 MB
  • 视频文件大小: 200 MB
  • 帧率: 30 FPS
  • 数据集划分: 训练集包含全部 17 个片段

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

特征描述

动作(Action)

  • 维度: 20
  • 数据类型: float32
  • 组件:
    • 左机械臂末端位姿 (left_tcp.x, left_tcp.y, left_tcp.z, left_tcp.r1~r6)
    • 左夹爪位置 (left_gripper.pos)
    • 右机械臂末端位姿 (right_tcp.x, right_tcp.y, right_tcp.z, right_tcp.r1~r6)
    • 右夹爪位置 (right_gripper.pos)

观测状态(Observation.State)

  • 与动作特征结构完全相同

观测图像(Observation.Images)

特征名称 图像尺寸 编码格式 帧率
left_tactile_left 400×700×3 H.264 30 FPS
left_tactile_right 400×700×3 H.264 30 FPS
left_wrist 480×640×3 H.264 30 FPS
right_tactile_left 400×700×3 H.264 30 FPS
right_tactile_right 400×700×3 H.264 30 FPS
right_wrist 480×640×3 H.264 30 FPS

其他特征

  • timestamp: float32, 形状 [1]
  • frame_index: int64, 形状 [1]
  • episode_index: int64, 形状 [1]
  • index: int64, 形状 [1]
  • task_index: int64, 形状 [1]

引用信息

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