23122025-foldclo-01_lerobot_format
收藏资源简介:
该数据集使用LeRobot工具创建,专注于机器人学领域,具体针对YAM机器人。数据集包含9个episodes,总计16772帧,帧率为30fps,数据以parquet格式存储。特征包括来自三个视角(顶部、左侧、右侧)的图像观测,每个图像分辨率为360x640x3;状态观测为14维浮点向量;动作为14维浮点向量。此外,还包含时间戳、帧索引、episode索引等元数据。数据集适用于机器人控制、模仿学习等任务。
This dataset is created using the LeRobot toolkit, focusing on the field of robotics, specifically targeting the YAM robot. The dataset consists of 9 episodes, totaling 16772 frames with a frame rate of 30fps, and the data is stored in Parquet format. Features include image observations from three perspectives: top, left, and right, with each image having a resolution of 360×640×3; the state observation is a 14-dimensional floating-point vector; and the action is also a 14-dimensional floating-point vector. In addition, it also includes metadata such as timestamps, frame indices, and episode indices. This dataset is applicable to tasks such as robot control and imitation learning.
数据集概述
- 数据集名称:
leokswang/23122025-foldclo-01_lerobot_format - 许可证: Apache-2.0
- 任务类别: 机器人学(Robotics)
- 标签: LeRobot, yam, lerobot, allenai_23122025-foldclo-01
数据集结构
- 机器人类型: yam
- 总片段数: 9
- 总帧数: 16772
- 总任务数: 1
- 视频数量: 0
- 帧率(FPS): 30
- 数据集划分: 训练集(片段0至8)
数据特征
- 观测图像:
observation.images.top: 分辨率 360x640,RGB三通道observation.images.left: 分辨率 360x640,RGB三通道observation.images.right: 分辨率 360x640,RGB三通道
- 观测状态:
observation.state,14维浮点数向量 - 动作:
action,14维浮点数向量 - 其他字段:
timestamp(时间戳)、frame_index(帧索引)、episode_index(片段索引)、index(全局索引)、task_index(任务索引)
数据格式
- 数据以 Parquet 格式存储,路径为
data/*/*.parquet - 视频路径(当前无视频):
videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4
创建工具
- 使用 LeRobot 框架创建



