SP_Referential_Disambiguation_200
收藏资源简介:
该数据集是一个用于机器人学习任务的多模态时序数据集,使用LeRobot框架创建。它包含180个episodes(总计88,939帧),涵盖4个不同任务。每个数据样本包括机器人的动作指令和来自多个传感器的观测信息:动作部分是一个6维浮点向量,对应机器人肩部平移、肩部抬升、肘部弯曲、腕部弯曲、腕部旋转和夹爪的位置控制;观测部分包括一个同维度的浮点状态观测向量(反映关节位置),以及来自五个固定视角(腕部、中部、上方、右侧、左侧)的RGB视频流,视频分辨率为640x480,帧率为30 fps,采用AV1编码。此外,每个样本还附带有时间戳、帧索引、episode索引、任务索引等元数据。数据集采用Apache-2.0许可证,所有数据均标记为训练集,适用于机器人模仿学习、视觉运动策略学习、多视角感知与控制等研究领域。
This dataset is a multimodal time-series dataset for robot learning tasks, created using the LeRobot framework. It contains 180 episodes (totaling 88,939 frames) covering 4 different tasks. Each data sample includes robot action commands and observations from multiple sensors: the action part is a 6-dimensional floating-point vector corresponding to position control for robot shoulder translation, shoulder lift, elbow bend, wrist bend, wrist rotation, and gripper; the observation part includes a floating-point state observation vector of the same dimension (reflecting joint positions), and RGB video streams from five fixed perspectives (wrist, middle, top, right, left), with each video having a resolution of 640x480, a frame rate of 30 fps, and AV1 encoding. Additionally, each sample comes with metadata such as timestamps, frame indices, episode indices, and task indices. The dataset uses the Apache-2.0 license, all data is labeled as training set, and it is suitable for research areas like robot imitation learning, visual-motor policy learning, and multi-view perception and control.
数据集概况
- 名称: SP_Referential_Disambiguation_200
- 许可证: Apache-2.0
- 任务类别: 机器人学 (Robotics)
- 标签: LeRobot
- 创建工具: 基于 LeRobot 创建
数据集结构
- 数据格式: Parquet 文件,位于
data/*/*.parquet - 视频编码: AV1,分辨率 480×640,帧率 30 FPS
- 帧率 (FPS): 30
特征 (Features)
| 特征名称 | 数据类型 | 形状 | 描述 |
|---|---|---|---|
action |
float32 | [6] | 机器人关节动作:shoulder_pan.pos, shoulder_lift.pos, elbow_flex.pos, wrist_flex.pos, wrist_roll.pos, gripper.pos |
observation.state |
float32 | [6] | 机器人关节状态(同上) |
observation.images.wrist |
video | [480, 640, 3] | 腕部摄像头图像 |
observation.images.middle |
video | [480, 640, 3] | 中间摄像头图像 |
observation.images.above |
video | [480, 640, 3] | 上方摄像头图像 |
observation.images.right |
video | [480, 640, 3] | 右侧摄像头图像 |
observation.images.left |
video | [480, 640, 3] | 左侧摄像头图像 |
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 |
数据集规模
- 总回合数: 180
- 总帧数: 88,939
- 总任务数: 4
- 数据文件大小: 约 100 MB
- 视频文件大小: 约 200 MB
划分
- 训练集: 0 至 179 回合(共 180 回合)
机器人类型
- 机器人型号: so_follower
引用
- 论文: 暂无信息
- 引用格式: 暂缺 BibTeX 条目




