taccap-g1-cleandesk-0702
收藏Hugging Face2026-07-02 更新2026-07-02 收录
下载链接:
https://huggingface.co/datasets/TacVerse/taccap-g1-cleandesk-0702
下载链接
链接失效反馈官方服务:
资源简介:
该数据集由LeRobot创建,专注于机器人学任务,特别是涉及双TacCap夹爪(bi_taccap_gripper)的机器人操作。数据集包含10个episodes,总计17582帧,以30fps的帧率录制。数据内容包括动作(action)和状态观测(observation.state),均以浮点32位格式存储,涵盖左右夹爪的TCP(工具中心点)位置(x, y, z)和旋转参数(r1到r6),以及夹爪位置。观测部分还包括多视角图像数据:左右触觉摄像头图像(分辨率400x700,3通道)和左右腕部摄像头图像(分辨率480x640,3通道),这些图像以视频格式(h264编解码,yuv420p像素格式)存储。此外,数据集提供时间戳、帧索引、episode索引、任务索引等元数据。数据以parquet文件格式组织,视频文件以mp4格式存储,总大小约为300MB(数据100MB,视频200MB)。该数据集适用于机器人控制、感知和强化学习研究。
This dataset was created by LeRobot, focusing on robotics research tasks, particularly robotic manipulation employing the dual bi_taccap_gripper. It comprises 10 episodes, with a total of 17,582 frames recorded at a frame rate of 30 fps. The dataset contains two primary data types: action data and state observations (observation.state), both stored in 32-bit floating-point (float32) format. The state observations cover the TCP (Tool Center Point) positions (x, y, z) and rotation parameters (r1 to r6) of the left and right grippers, alongside the gripper positions. The observation module also includes multi-view image data: left and right tactile camera images (resolution 400×700, 3 channels) and left and right wrist camera images (resolution 480×640, 3 channels). These image sequences are stored as video files using the h264 codec and yuv420p pixel format. Additionally, the dataset provides supplementary metadata including timestamps, frame indices, episode indices, and task indices. The structured tabular data is organized in Parquet file format, while the video files are stored in MP4 format. The total size of the entire dataset is approximately 300 MB, with 100 MB corresponding to the tabular data and 200 MB to the video files. This dataset is suitable for research in robotic control, perception, and reinforcement learning.
提供机构:
TacVerse创建时间:
2026-07-02
原始信息汇总
数据集概述:TacVerse/taccap-g1-cleandesk-0702
该数据集用于机器人学领域的触觉操作任务,基于 LeRobot 框架创建,采用 Apache-2.0 许可证。
数据集基本信息
| 项目 | 内容 |
|---|---|
| 数据集名称 | taccap-g1-cleandesk-0702 |
| 任务类别 | 机器人学(robotics) |
| 许可证 | Apache-2.0 |
| 机器人类型 | bi_taccap_gripper |
| 数据集版本 | v3.0 |
| 总片段数 | 10 |
| 总帧数 | 17,582 |
| 总任务数 | 1 |
| 帧率 | 30 FPS |
| 数据文件总大小 | ~100 MB |
| 视频文件总大小 | ~200 MB |
| 数据拆分 | 全部10个片段用于训练(train: 0:10) |
数据特征结构
数据集包含以下四类特征:
1. 动作(action)与观测状态(observation.state)
- 数据类型:float32
- 维度:20维向量
- 包含内容:
- 左侧TCP(工具中心点)位姿:x, y, z, r1, r2, r3, r4, r5, r6
- 左侧夹爪位置:left_gripper.pos
- 右侧TCP位姿:x, y, z, r1, r2, r3, r4, r5, r6
- 右侧夹爪位置:right_gripper.pos
2. 视觉观测(observation.images)
数据集提供6路视频观测,均为 H.264 编码、30 FPS、无音频:
| 图像名称 | 分辨率 (高×宽) | 通道数 | 说明 |
|---|---|---|---|
| left_tactile_left | 400×700 | 3 (RGB) | 左侧触觉左摄像头 |
| left_tactile_right | 400×700 | 3 (RGB) | 左侧触觉右摄像头 |
| left_wrist | 480×640 | 3 (RGB) | 左侧腕部摄像头 |
| right_tactile_left | 400×700 | 3 (RGB) | 右侧触觉左摄像头 |
| right_tactile_right | 400×700 | 3 (RGB) | 右侧触觉右摄像头 |
| right_wrist | 480×640 | 3 (RGB) | 右侧腕部摄像头 |
3. 元数据字段
- 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:全局索引,int64,形状 [1]
- task_index:任务索引,int64,形状 [1]
数据文件存储结构
- 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 - 数据块大小:chunks_size = 1000
引用信息
如需引用该数据集,请使用以下 BibTeX 条目:
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} }



