taccap-g1-arrange-pens-sticky-notes-0721
收藏资源简介:
该数据集名为TacVerse/taccap-g1-arrange-pens-sticky-notes-0721,是一个机器人学数据集,专门用于整理笔和便签纸的任务。数据集使用LeRobot框架创建,包含双TacCap夹持器机器人(bi_taccap_gripper)收集的数据。总共有34个episodes,共计173,635帧,帧率为30fps。数据以parquet文件和视频文件形式存储,总数据大小约为100MB,视频文件大小约为200MB。数据集仅包含训练集,覆盖所有episodes。特征包括:动作数据(20维浮点数组,表示左右夹持器的TCP位置、旋转和夹持器位置)、观测状态(与动作相同的20维数据)、多视角图像观测(包括左腕部摄像头、右腕部摄像头以及左右触觉传感器的左右图像,分辨率分别为480x640和400x700,均为RGB视频),以及时间戳、帧索引、episode索引等元数据。该数据集适用于机器人控制、触觉感知和多模态学习的研究。
This dataset, named TacVerse/taccap-g1-arrange-pens-sticky-notes-0721, is a robotics dataset specifically designed for the task of arranging pens and sticky notes. It was created using the LeRobot framework and contains data collected by a dual TacCap gripper robot (bi_taccap_gripper). A total of 34 episodes are included, amounting to 173,635 frames with a frame rate of 30 fps. The dataset is stored in parquet files and video files, with a total data size of approximately 100 MB, and the video files alone totaling around 200 MB. This dataset only provides the training split, covering all episodes. Key features of the dataset include: action data (a 20-dimensional floating-point array representing the TCP position, rotation, and gripper position for both the left and right grippers), observation states (20-dimensional data identical to the action data), multi-view image observations (including images from the left wrist camera, right wrist camera, and left/right images of the left and right tactile sensors, with resolutions of 480x640 and 400x700 respectively, both in RGB video format), as well as metadata such as timestamps, frame indices, and episode indices. This dataset is suitable for research in robot control, tactile perception, and multimodal learning.
数据集:TacVerse/taccap-g1-arrange-pens-sticky-notes-0721
基本概览
- 任务类型:机器人学(Robotics)
- 许可协议:Apache-2.0
- 创建工具:LeRobot (https://github.com/huggingface/lerobot)
- 机器人类型:bi_taccap_gripper(双夹爪触觉机器人)
数据集规模
- 总片段数:34
- 总帧数:173,635
- 总任务数:1
- 帧率:30 FPS
- 数据集文件大小:100 MB
- 视频文件大小:200 MB
- 数据块大小:1000
分割
- 仅包含训练集,覆盖所有34个片段(0:34)
特征结构
动作(action)与观测状态(observation.state)
- 数据类型:float32
- 维度:20
- 包含:左右TCP位置(x, y, z, r1-r6)和夹爪位置(left_gripper.pos, right_gripper.pos)
图像观测(observation.images)
- left_tactile_left / left_tactile_right / right_tactile_left / right_tactile_right:400×700像素,3通道,H.264编码,30 FPS
- left_wrist / right_wrist:480×640像素,3通道,H.264编码,30 FPS
其他特征
- timestamp(时间戳):float32,形状[1]
- frame_index(帧索引):int64,形状[1]
- episode_index(片段索引):int64,形状[1]
- index(索引):int64,形状[1]
- task_index(任务索引):int64,形状[1]
引用
@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} }



