five

chaoqi-liu/mshab_settable_clean

收藏
Hugging Face2026-04-27 更新2026-05-03 收录
下载链接:
https://hf-mirror.com/datasets/chaoqi-liu/mshab_settable_clean
下载链接
链接失效反馈
官方服务:
资源简介:
mshab_settable_clean数据集是已发布的MS-HAB SetTable子任务的LeRobot格式转换版本。该数据集专注于SetTable子任务,包含两个任务字符串:set table: pick object(设置桌子:拾取物体)和set table: place object(设置桌子:放置物体)。数据集总共有4000个episodes,每个任务各2000个episodes,总帧数为151547帧,帧率为20fps。数据以parquet文件格式存储,并包含RGB视频(来自fetch_head和fetch_hand摄像头)。观测数据包括42维扁平状态、12维关节位置和速度、7维TCP和物体姿态、3维目标位置、1维抓取状态,以及128x128x3的RGB图像。动作数据为13维浮点控制向量,对应机器人关节和基座控制。数据集还保留了原始HDF5源的相机矩阵,并提供了详细的元数据和转换报告。

mshab_settable_clean is a LeRobot-format conversion of the released MS-HAB SetTable subtasks. It focuses on the SetTable subtasks, with task strings: set table: pick object and set table: place object. The dataset contains 4000 episodes in total, with 2000 episodes per task, 151547 total frames, and a frame rate of 20 fps. Data is stored in parquet format and includes RGB videos (from fetch_head and fetch_hand cameras). Observations include a 42-D flat state, 12-D joint positions and velocities, 7-D TCP and object poses, 3-D goal positions, 1-D grasp status, and 128x128x3 RGB images. Actions are 13-D float32 control vectors for robot joints and base. The dataset preserves camera matrices from the original HDF5 source and provides detailed metadata and conversion reports.
提供机构:
chaoqi-liu
5,000+
优质数据集
54 个
任务类型
进入经典数据集
二维码
社区交流群

面向社区/商业的数据集话题

二维码
科研交流群

面向高校/科研机构的开源数据集话题

数据驱动未来

携手共赢发展

商业合作