training_xylophone

该数据集是通过LeRobot创建的机器人控制数据集，专注于so100_follower机器人类型。数据集包含120个训练集，总帧数72108，涉及8个不同任务。数据以Parquet文件格式存储，并包含视频文件（总大小约200MB）。特征包括动作（6个浮点数，对应机器人关节如肩部、肘部、腕部和夹爪控制）、观测状态（同样6个浮点数，表示机器人状态）、两个摄像头图像观测（顶部摄像头分辨率为720x1280，侧边摄像头分辨率为480x640，均为30fps视频），以及时间戳、帧索引、集索引、索引和任务索引等元数据。数据集主要用于机器人学习和控制任务，支持训练和评估算法。This dataset is a robot control dataset created via LeRobot, focusing on the so100_follower robot type. It contains 120 training episodes, with a total of 72,108 frames, covering 8 different tasks. The data is stored in Parquet file format and includes video files (total size approximately 200MB). Features include actions (6 floating-point numbers corresponding to robot joints such as shoulder, elbow, wrist, and gripper control), observation states (also 6 floating-point numbers representing robot states), two camera image observations (top camera with a resolution of 720x1280 and side camera with a resolution of 480x640, both at 30fps video), as well as metadata such as timestamps, frame indices, episode indices, indices, and task indices. The dataset is primarily used for robot learning and control tasks, supporting algorithm training and evaluation.

原始信息汇总

数据集概述

• 数据集名称: training_xylophone
• 任务类别: 机器人 (robotics)
• 许可证: Apache-2.0
• 创建工具: 使用 LeRobot 创建

数据集结构

• 代码库版本: v3.0
• 机器人类型: so100_follower
• 总片段数: 120
• 总帧数: 72,108
• 总任务数: 8
• 块大小: 1,000
• 数据文件大小: 100 MB
• 视频文件大小: 200 MB
• 帧率 (FPS): 30
• 数据分割: 训练集（0 至 120 片段）

特征描述

特征名	数据类型	形状	说明
action	float32	[6]	六维动作向量：shoulder_pan, shoulder_lift, elbow_flex, wrist_flex, wrist_roll, gripper
observation.state	float32	[6]	六维状态观测，命名与动作相同
observation.images.top	视频	[720, 1280, 3]	顶部摄像机画面，AV1 编码，30 FPS，分辨率 1280x720
observation.images.side	视频	[480, 640, 3]	侧面摄像机画面，AV1 编码，30 FPS，分辨率 640x480
timestamp	float32	[1]	时间戳
frame_index	int64	[1]	帧索引
episode_index	int64	[1]	片段索引
index	int64	[1]	全局索引
task_index	int64	[1]	任务索引

数据文件路径

• 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