remove_orange_duck_pruned
收藏资源简介:
该数据集由LeRobot项目创建,专注于机器人技术领域,旨在为机器人学习任务提供示范数据。它包含88个独立的任务片段,总计73,728帧数据。每个数据样本包括动作指令和观测信息:动作指令由6个浮点数组成,分别控制机器人的肩部平移、肩部升降、肘部弯曲、腕部弯曲、腕部旋转和夹爪位置;观测信息包括机器人状态(同样由6个关节位置浮点数表示)和来自顶部摄像头的视觉输入。视觉数据以视频形式存储,分辨率为1280x720,帧率为30fps,采用h264编码。数据集还包含时间戳、帧索引、片段索引等元数据字段,以Parquet文件格式组织,总大小约为300MB(其中数据文件100MB,视频文件200MB)。它适用于机器人模仿学习、强化学习或行为克隆等任务,特别针对so_follower型机器人平台,所有数据已划分为训练集。
This dataset was created by the LeRobot project, focusing on the robotics field and designed to provide demonstration data for robot learning tasks. It contains 88 independent task segments, totaling 73,728 frames. Each data sample includes action commands and observation information: action commands consist of 6 floating-point numbers controlling the robots shoulder translation, shoulder elevation, elbow flexion, wrist flexion, wrist rotation, and gripper position; observation information includes robot state (also represented by 6 joint position floating-point numbers) and visual input from a top-mounted camera. Visual data is stored in video format with a resolution of 1280x720, a frame rate of 30fps, and h264 encoding. The dataset also includes metadata fields such as timestamps, frame indices, and segment indices, organized in Parquet file format with a total size of approximately 300MB (100MB for data files and 200MB for video files). It is suitable for tasks like robot imitation learning, reinforcement learning, or behavior cloning, specifically targeting the so_follower robot platform, with all data already divided into a training set.
数据集概述:remove_orange_duck_pruned
- 数据集主页: https://huggingface.co/datasets/KuphDev/remove_orange_duck_pruned
- 许可协议: Apache-2.0
- 任务类型: 机器人学 (Robotics)
- 标签: LeRobot
- 创建工具: 该数据集使用 LeRobot 创建
数据集结构与规格
- 数据集版本: v3.0
- 帧率 (FPS): 30
- 机器人类型: so_follower (跟随机械臂)
- 总片段数 (Episodes): 128
- 总帧数 (Frames): 103,462
- 总任务数 (Tasks): 1
- 数据文件大小: 100 MB
- 视频文件大小: 200 MB
- 分块大小: 1000
数据集划分
- 训练集 (train): 片段索引 0 至 127 (共128个片段)
特征 (Features)
| 特征名称 | 数据类型 | 形状 | 说明 |
|---|---|---|---|
action |
float32 | (6,) | 动作指令,包含6个关节位置 |
observation.state |
float32 | (6,) | 观测状态,包含6个关节位置 |
observation.images.top |
视频 (h264) | (720, 1280, 3) | 顶部摄像头图像,分辨率720x1280,3通道彩色,帧率30 |
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 |
动作(Action)与观测状态(State)的命名含义
action 和 observation.state 均为6维向量,对应关节名称依次为:
- shoulder_pan.pos (肩关节旋转)
- shoulder_lift.pos (肩关节升降)
- elbow_flex.pos (肘关节弯曲)
- wrist_flex.pos (腕关节弯曲)
- wrist_roll.pos (腕关节旋转)
- gripper.pos (夹爪)
视频数据规格
- 视频编解码器: h264
- 像素格式: yuv420p
- 单帧尺寸: 720 (高) x 1280 (宽) x 3 (通道)
- 是否为深度图: 否
- 帧率: 30 FPS
- 无音频
数据文件路径
- 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
引用
目前暂无可用的 BibTeX 引用格式。




