taccap-g1-arrange-food-packaging-bags
收藏资源简介:
TaccAP-G1 Arrange Food Packaging Bags 数据集是一个用于机器人操作研究的数据集,由LeRobot框架创建,许可证为Apache-2.0。该数据集聚焦于双臂机器人(bi_taccap_gripper)执行整理食品包装袋的任务。数据集包含388个演示片段(episodes),总计495174帧(frames),采样率为30帧/秒。所有数据划分为训练集(0:388个片段)。数据特征包括:动作(action)和观测状态(observation.state),各包含20维的左右机械臂TCP位姿(位置和旋转)及夹爪位置;多个视觉观测模态:左侧触觉图像(左/右两个传感器,400x700)、右侧触觉图像(左/右两个传感器,400x700)、左腕部摄像头图像(480x640)、右腕部摄像头图像(480x640),均为视频格式(H.264编码,30fps);此外还有时间戳、帧索引、片段索引、索引和任务索引等辅助字段。数据以parquet文件和MP4视频文件存储。该数据集适用于机器人模仿学习、双臂操作策略学习、触觉感知与视觉融合等任务。
The TaccAP-G1 Arrange Food Packaging Bags dataset is a dataset for robot manipulation research, created by the LeRobot framework and licensed under Apache-2.0. It focuses on a dual-arm robot (bi_taccap_gripper) performing the task of arranging food packaging bags. The dataset contains 388 episodes, totaling 495,174 frames at a sampling rate of 30 fps. All data is split into a training set (episodes 0 to 388). Data features include: action and observation state, each with 20 dimensions representing left and right arm TCP poses (position and rotation) and gripper positions; multiple visual observation modalities: left tactile images (left/right sensors, 400x700), right tactile images (left/right sensors, 400x700), left wrist camera images (480x640), right wrist camera images (480x640), all in video format (H.264 encoding, 30fps); additionally, auxiliary fields such as timestamps, frame indices, episode indices, index, and task index. Data is stored as parquet files and MP4 video files. The dataset is suitable for robot imitation learning, dual-arm manipulation policy learning, tactile perception and visual fusion, among other tasks.
数据集概述:TacVerse/taccap-g1-arrange-food-packaging-bags
基本信息
- 任务类别: 机器人学(Robotics)
- 许可证: Apache-2.0
- 创建工具: 使用 LeRobot(https://github.com/huggingface/lerobot)创建
- 数据集可视化: https://huggingface.co/spaces/lerobot/visualize_dataset?path=TacVerse/taccap-g1-arrange-food-packaging-bags
数据集规模
| 项目 | 数值 |
|---|---|
| 总集数(Episodes) | 388 |
| 总帧数(Frames) | 495,174 |
| 总任务数 | 1 |
| 分块大小(Chunks Size) | 1,000 |
| 数据文件大小 | 100 MB |
| 视频文件大小 | 200 MB |
| 帧率(FPS) | 30 |
| 训练集划分 | 训练集(全部388集) |
机器人配置
- 机器人类型: bi_taccap_gripper
- 代码库版本: v3.0
数据特征
动作特征(Action)
- 数据类型: float32
- 维度: 20维
- 包含字段:
- 左/右TCP坐标(x, y, z)及旋转参数(r1-r6)
- 左/右夹爪位置
状态观测特征(Observation State)
- 数据类型: float32
- 维度: 20维
- 字段与动作特征相同
图像观测特征(Observation Images)
共6个摄像头视角,均为视频数据:
| 视角 | 分辨率 | 编码格式 |
|---|---|---|
| left_tactile_left | 400×700×3 | H.264 (yuv420p) |
| left_tactile_right | 400×700×3 | H.264 (yuv420p) |
| left_wrist | 480×640×3 | H.264 (yuv420p) |
| right_tactile_left | 400×700×3 | H.264 (yuv420p) |
| right_tactile_right | 400×700×3 | H.264 (yuv420p) |
| right_wrist | 480×640×3 | H.264 (yuv420p) |
- 帧率:30 FPS,无音频
其他特征
- timestamp: float32
- frame_index: int64
- episode_index: int64
- index: int64
- task_index: int64
数据存储结构
- 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
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} }




