tray2tray_merged
收藏资源简介:
该数据集由LeRobot创建,专注于机器人学领域,特别是用于机器人控制和视觉任务。它包含动作数据(包括6个关节位置:肩部平移、肩部升降、肘部弯曲、腕部弯曲、腕部旋转和夹爪位置)和观察状态数据(同样包含这些关节位置)。此外,数据集提供了两个摄像头的视频观察数据,每个视频分辨率为480x640、30fps、RGB格式,用于捕捉机器人环境。数据集还包含元数据,如时间戳、帧索引、片段索引和任务索引。数据集总共有121个片段、76,307帧,覆盖2个不同任务。数据以parquet文件格式存储,视频以mp4格式存储,机器人类型为so_follower。该数据集适用于机器人学习、模仿学习或强化学习等应用场景。
This dataset, created by LeRobot, focuses on the field of robotics, particularly for robot control and vision tasks. It contains action data (including 6 joint positions: shoulder translation, shoulder elevation, elbow flexion, wrist flexion, wrist rotation, and gripper position) and observation state data (which also includes these joint positions). Additionally, the dataset provides video observation data from two cameras, with each video having a resolution of 480x640, 30fps, and RGB format for capturing the robot environment. The dataset also includes metadata such as timestamps, frame indices, segment indices, and task indices. In total, the dataset contains 121 segments and 76,307 frames, covering 2 distinct tasks. The structured data is stored in Parquet file format, while the videos are stored in MP4 format, and the robot type is so_follower. This dataset is suitable for applications such as robotic learning, imitation learning, and reinforcement learning.
数据集概述:EJ-DEMO/tray2tray_merged
基本信息
- 许可证:Apache-2.0
- 任务类别:机器人(Robotics)
- 标签:LeRobot
- 创建工具:使用 LeRobot 框架创建
数据集结构
- 代码版本:v3.0
- 帧率:30 FPS
- 总片段数:121 个
- 总帧数:76,307
- 总任务数:2 个
- 数据文件大小:100 MB
- 视频文件大小:200 MB
- 机器人类型:so_follower
特征与数据格式
- 动作(Action):
- 数据类型:float32
- 形状:6 维
- 命名维度:shoulder_pan.pos, shoulder_lift.pos, elbow_flex.pos, wrist_flex.pos, wrist_roll.pos, gripper.pos
- 观察状态(Observation State):
- 数据类型:float32
- 形状:6 维
- 命名维度:与动作相同
- 观察图像(Observation Images):
- 两个摄像头:observation.images.1 和 observation.images.2
- 图像尺寸:480×640 像素,3 通道
- 视频编码:AV1
- 帧率:30 FPS
- 非深度图
- 其他特征:timestamp, frame_index, episode_index, index, task_index
文件路径与拆分
- 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 - 数据集拆分:仅包含训练集(train),范围 0 到 121
数据可视化
- 可通过 LeRobot 数据集可视化工具在线查看:https://huggingface.co/spaces/lerobot/visualize_dataset?path=EJ-DEMO/tray2tray_merged



