train_baseline_20260722_200824
收藏资源简介:
该数据集是一个机器人控制数据集,基于LeRobot框架构建,专注于收集机器人动作和观测数据。它包含5个总episodes和2295个帧,帧率为30fps。数据特征包括动作(由6个关节位置组成:肩部平移、肩部提升、肘部弯曲、腕部弯曲、腕部旋转和夹爪位置)和观测状态(同样为6个关节位置)。此外,数据集提供来自顶部和前置摄像头的视频观测,图像分辨率为480x640,视频编码为av1。元数据包括时间戳、帧索引、episode索引、任务索引等辅助信息。机器人类型为“so_follower”,数据集以parquet文件格式存储,视频文件为mp4格式,总数据大小约为100MB,视频文件大小约为200MB。该数据集适用于机器人学习任务,如模仿学习或强化学习,旨在支持机器人控制算法的开发和评估。
This dataset is a robot control dataset built using the LeRobot framework, focusing on the collection of robot actions and observations. It contains 5 total episodes and 2295 frames with a frame rate of 30fps. The data features include actions (composed of 6 joint positions: shoulder translation, shoulder elevation, elbow flexion, wrist flexion, wrist rotation, and gripper position) and observation states (also 6 joint positions). Additionally, the dataset provides video observations from top and front cameras, with an image resolution of 480x640 and video encoding in av1. Metadata includes auxiliary information such as timestamps, frame indices, episode indices, and task indices. The robot type is so_follower, and the dataset is stored in parquet file format, with video files in mp4 format. The total data size is approximately 100MB, and the video file size is about 200MB. This dataset is suitable for robot learning tasks, such as imitation learning or reinforcement learning, aiming to support the development and evaluation of robot control algorithms.
数据集概述
- 名称:
train_baseline_20260722_200824 - 许可证: Apache-2.0
- 任务类别: 机器人学 (robotics)
- 标签: LeRobot
- 创建工具: LeRobot
- 数据集规模:
- 总片段数 (total_episodes): 5
- 总帧数 (total_frames): 2295
- 总任务数 (total_tasks): 1
- 数据文件大小: 100 MB
- 视频文件大小: 200 MB
- 机器人类型: so_follower
- 帧率 (fps): 30
- 划分 (splits): 训练集 (train): 片段 0 至 4
数据集结构
数据集包含以下特征字段:
action(float32, 形状 [6]): 机器人动作指令,包含 6 个关节的位置:shoulder_pan.posshoulder_lift.poselbow_flex.poswrist_flex.poswrist_roll.posgripper.pos
observation.state(float32, 形状 [6]): 机器人观测状态,与action相同的关节位置信息。observation.images.top(视频数据, 形状 [480, 640, 3]): 顶部摄像头图像,分辨率 480x640,RGB 3 通道,AV1 编码,30 fps。observation.images.front(视频数据, 形状 [480, 640, 3]): 前部摄像头图像,分辨率 480x640,RGB 3 通道,AV1 编码,30 fps。timestamp(float32, 形状 [1]): 时间戳。frame_index(int64, 形状 [1]): 帧索引。episode_index(int64, 形状 [1]): 片段索引。index(int64, 形状 [1]): 全局索引。task_index(int64, 形状 [1]): 任务索引。
数据文件与视频文件路径
- 数据文件:
data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet - 视频文件:
videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4



