task10-move-to-beaker-storage-shelf_0804
收藏资源简介:
该数据集使用LeRobot工具创建,面向机器人操作任务,包含490个演示片段(episode),总计96,952帧。数据由mobileai_robot机器人采集,执行的任务为“移动至烧杯储存架”(task10-move-to-beaker-storage-shelf_0804)。每个样本包含16维的动作(action)和状态观测(observation.state),分别对应左右臂的6个关节位置、一个附加关节位置、线速度和角速度。同时提供三个摄像头视角的RGB视频:高位摄像头(cam_high)、左腕摄像头(cam_left_wrist)和右腕摄像头(cam_right_wrist),分辨率均为480×640像素,H.264编码,帧率30fps。此外还包括时间戳、帧索引、episode索引等辅助字段。数据以parquet格式存储,视频以mp4格式保存,所有数据均用于训练(无独立验证集)。该数据集适用于机器人模仿学习、示教学习、多模态感知与控制等任务。
This dataset is created using the LeRobot tool, targeting robot manipulation tasks. It contains 490 demonstration episodes with a total of 96,952 frames. The data was collected by the mobileai_robot robot performing the task move to beaker storage shelf (task10-move-to-beaker-storage-shelf_0804). Each sample includes 16-dimensional action and observation.state, corresponding to the 6 joint positions of each arm, an additional joint position, linear velocity, and angular velocity. Three camera views provide RGB videos: a high camera (cam_high), a left wrist camera (cam_left_wrist), and a right wrist camera (cam_right_wrist), all with a resolution of 480×640 pixels, H.264 encoding, and 30fps frame rate. Additional fields include timestamps, frame indices, and episode indices. The data is stored in parquet format, and videos are saved as mp4 files. All data is used for training (no separate validation set). This dataset is suitable for tasks such as robot imitation learning, teaching learning, and multimodal perception and control.
数据集详情总结
基本信息
- 数据集名称:task10-move-to-beaker-storage-shelf_0804
- 创建工具:LeRobot(https://github.com/huggingface/lerobot)
- 许可证:Apache-2.0
- 任务类别:机器人(Robotics)
- 标签:LeRobot
数据集规模
- 总片段数:490
- 总帧数:96,952
- 任务数:1
- 数据文件大小:100 MB
- 视频文件大小:200 MB
- 帧率:30 FPS
数据集结构
- 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 - 数据集划分:训练集包含全部 490 个片段(0:490),无验证集和测试集
- 数据格式:Parquet 文件,存储于
data/*/*.parquet
机器人类型与动作空间
- 机器人类型:mobileai_robot
- 动作空间(Action):16 维,包含:
- 左臂 7 个关节位置(left_joint_0 至 left_joint_5 及 left_left_carriage_joint)
- 右臂 7 个关节位置(right_joint_0 至 right_joint_5 及 right_left_carriage_joint)
- 机器人的线速度(x.vel)和角速度(theta.vel)
- 观测状态空间(observation.state):与动作空间相同,共 16 维
观测图像
数据集包含 3 个摄像头视角的图像观测,均为 H.264 编码视频,分辨率为 480×640,30 FPS:
- cam_high:高位摄像头
- cam_left_wrist:左手腕摄像头
- cam_right_wrist:右手腕摄像头
其他特征
- timestamp:时间戳,float32 类型
- frame_index:帧索引,int64 类型
- episode_index:片段索引,int64 类型
- index:全局索引,int64 类型
- task_index:任务索引,int64 类型
- chunks_size:1000(每个 chunk 包含的帧数)
待补充信息
- 主页:未提供
- 论文:未提供
- 引用:未提供 BibTeX 引用格式




