SortPaper_20260529_145930
收藏资源简介:
该数据集是一个用于机器人学习的多模态数据集,基于LeRobot框架构建,专注于so_follower型机器人的操作任务。它包含60个完整的操作序列(episodes),总计32,472帧数据,采集频率为30帧/秒,数据规模约为300MB,其中包含100MB的结构化数据文件(Parquet格式)和200MB的视频文件。数据集的核心内容包括:动作(一个6维浮点向量,表示机器人六个关节的目标位置,如shoulder_pan.pos等)、观察(状态为6维浮点向量表示相同关节的当前观测位置,以及来自两个视角的同步视频流:前视和顶视摄像头,分辨率640x480像素,RGB格式,编码为AV1,帧率30fps)和元数据与索引(如时间戳、帧索引、episode索引等)。该数据集适用于机器人模仿学习、强化学习或行为克隆等任务,提供动作与多视角视觉感知及本体状态的配对数据,所有数据均被划分为训练集使用。
This dataset is a multimodal dataset for robot learning, created using the LeRobot framework, focusing on robotic manipulation tasks specifically for so_follower type robots. It contains 60 complete operation sequences (episodes), totaling 32,472 frames of data collected at a frequency of 30 frames per second. The dataset size is approximately 300MB, including 100MB of structured data files and 200MB of video files, stored in Parquet format. The core content includes: action (a 6-dimensional floating-point vector representing target positions for six robot joints, such as shoulder_pan.pos, shoulder_lift.pos, elbow_flex.pos, wrist_flex.pos, wrist_roll.pos, gripper.pos), observation (state as a 6-dimensional floating-point vector for the current observed positions of the same six joints, and synchronized video streams from two perspectives: front and top cameras, with a resolution of 640x480 pixels, RGB format, AV1 encoding, and a frame rate of 30fps), and metadata and indexing (such as timestamp, frame_index, episode_index, global index, and task_index, with this dataset containing only one task). This dataset is suitable for tasks like robot imitation learning, reinforcement learning, or behavior cloning, providing paired data of robot joint control commands (actions) with corresponding multi-view visual perception and proprioceptive states (observations). All data is divided for training set use.
数据集概述:SortPaper_20260529_145930
- 许可证: Apache-2.0
- 任务类别: 机器人学(robotics)
- 标签: LeRobot
- 创建工具: 使用 LeRobot 创建
数据集结构
数据集包含以下特征字段:
| 字段名称 | 数据类型 | 形状 | 说明 |
|---|---|---|---|
action |
float32 | (6,) | 机器人动作指令,包含6个关节(shoulder_pan、shoulder_lift、elbow_flex、wrist_flex、wrist_roll、gripper)的位置 |
observation.state |
float32 | (6,) | 机器人观测状态,包含与动作相同的6个关节位置 |
observation.images.front |
video | (480, 640, 3) | 前置摄像头视频流,分辨率480x640,AV1编码,30 FPS |
observation.images.top |
video | (480, 640, 3) | 顶部摄像头视频流,分辨率480x640,AV1编码,30 FPS |
timestamp |
float32 | (1,) | 时间戳 |
frame_index |
int64 | (1,) | 帧索引 |
episode_index |
int64 | (1,) | 回合索引 |
index |
int64 | (1,) | 全局索引 |
task_index |
int64 | (1,) | 任务索引 |
数据集统计
- 总回合数: 60
- 总帧数: 32,472
- 总任务数: 1
- 数据文件大小: 约100 MB
- 视频文件大小: 约200 MB
- 帧率: 30 FPS
- 机器人类型: so_follower
数据划分
- 训练集: 全部60个回合(索引0-59)
数据格式
- 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 - 代码库版本: v3.0
- 数据块大小: 1000帧/块
引用信息
暂无提供引用信息。




