reset_top_slider_sweep
收藏资源简介:
这是一个由LeRobot创建的机器人数据集,专门用于机器人控制与学习任务。数据集包含1个完整的情节(episode),总计1118帧数据,帧率为30fps。数据以parquet格式存储,并附带相应的视频文件。核心特征包括:动作数据(12维,表示左右机械臂的关节位置,如肩部平移、肩部提升、肘部弯曲、腕部弯曲、腕部滚动和夹爪位置);观测状态数据(27维,包括相同的关节位置以及busybox设备的传感器数据,如开关状态、导线连接、滑块位置、按钮状态和旋钮值);以及来自三个摄像头的图像观测:左腕摄像头(720x1280分辨率)、右腕摄像头(720x1280分辨率)和顶部摄像头(576x1024分辨率),所有视频均以AV1编码。此外,数据集还包含时间戳、帧索引、情节索引等元数据。适用于机器人模仿学习、强化学习或计算机视觉任务的研究。
This is a robot dataset created by LeRobot, specifically designed for robot control and learning tasks. The dataset contains 1 complete episode, totaling 1118 frames of data at a frame rate of 30fps. Data is stored in parquet format and accompanied by corresponding video files. Core features include: action data (12-dimensional, representing the joint positions of the left and right robotic arms, such as shoulder pan, shoulder lift, elbow bend, wrist bend, wrist roll, and gripper position); observation state data (27-dimensional, including the same joint positions and sensor data from a busybox device, such as switch states, wire connections, slider positions, button states, and knob values); and image observations from three cameras: left wrist camera (720x1280 resolution), right wrist camera (720x1280 resolution), and top camera (576x1024 resolution), all videos encoded in AV1. Additionally, the dataset includes metadata such as timestamps, frame indices, and episode indices. It is suitable for research in robot imitation learning, reinforcement learning, or computer vision tasks.
数据集概述
数据集名称: jamesb123/reset_top_slider_sweep
许可证: Apache-2.0
任务类别: 机器人学 (Robotics)
创建工具: 基于 LeRobot 构建
数据集结构与规模
- 机器人类型: so-101
- 总片段数 (Episodes): 1
- 总帧数 (Frames): 1,118
- 总任务数 (Tasks): 1
- 帧率 (FPS): 30
- 数据文件大小: 100 MB
- 视频文件大小: 200 MB
- 数据格式: Parquet 文件 (
data/*/*.parquet) - 视频格式: MP4 文件 (
videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4) - 数据分块: 每块 1000 帧 (chunks_size: 1000)
- 训练/测试划分: 全部数据用于训练 (
train: "0:1")
数据特征 (Features)
| 特征名 | 数据类型 | 形状 (Shape) | 说明 |
|---|---|---|---|
action |
float32 | [12] | 左右双臂各6个关节位置(肩、肘、腕、夹爪) |
observation.state |
float32 | [27] | 包括12个关节位置 + 15个操作台状态(开关、滑块、按钮、旋钮等) |
observation.images.left_wrist |
video | [720, 1280, 3] | 左腕摄像头视频(帧率30fps,AV1编码) |
observation.images.right_wrist |
video | [720, 1280, 3] | 右腕摄像头视频(帧率30fps,AV1编码) |
observation.images.top |
video | [576, 1024, 3] | 顶部摄像头视频(帧率30fps,AV1编码) |
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 |
操作台状态细节 (State 特征补充)
observation.state 包含以下15个额外传感器值:
- 开关 (Switch):
bottom_switch,top_switch - 导线 (Wire):
blue_wire,green_wire,red_wire,yellow_wire - 滑块 (Slider):
bottom_slider,bottom_slider_raw,top_slider,top_slider_raw - 按钮 (Button):
green_button,red_button,white_button,yellow_button - 旋钮 (Knob):
knob
可视化
可通过以下空间直接预览数据集:
- 可视化链接: https://huggingface.co/spaces/lerobot/visualize_dataset?path=jamesb123/reset_top_slider_sweep
引用
如需引用,请参考 BibTeX 格式(当前论文信息待补充)。



