board_insertion_ablation_dagger
收藏资源简介:
该数据集是一个专注于板插入任务的机器人操作数据集,使用LeRobot工具创建。它包含机器人状态观测(15个浮点数,表示左臂7个关节和左夹爪8个关节的位置)、动作(8个浮点数,控制左臂7个关节和左夹爪)、图像观测(头部摄像头:720x1280分辨率,左右手腕摄像头:240x424分辨率,均为30fps视频),以及时间戳、帧索引、episode索引等元数据。总共有55个训练episodes,53230帧数据,数据以Parquet格式存储,视频以MP4格式存储。该数据集适用于机器人学习和模仿学习任务。
This dataset is a robot manipulation dataset focused on board insertion tasks, created using the LeRobot tool. It includes robot state observations (15 floating-point numbers representing the positions of 7 joints in the left arm and 8 joints in the left gripper), actions (8 floating-point numbers controlling the 7 joints of the left arm and the left gripper), image observations (head camera: 720x1280 resolution, left and right wrist cameras: 240x424 resolution, all at 30fps video), along with metadata such as timestamps, frame indices, and episode indices. There are a total of 55 training episodes and 53,230 frames of data, stored in Parquet format for data and MP4 format for videos. The dataset is suitable for robot learning and imitation learning tasks.
数据集概述
数据集名称:board_insertion_ablation_dagger
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 框架创建
数据集结构
- 总集数:55
- 总帧数:53,230
- 总任务数:1
- 帧率:30 FPS
- 数据文件大小:100 MB
- 视频文件大小:200 MB
- 数据路径格式:
data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet - 视频路径格式:
videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4 - 划分:训练集包含全部 55 集
特征说明
| 特征名称 | 数据类型 | 形状 | 描述 |
|---|---|---|---|
observation.state |
float32 | (15,) | 机械臂和夹爪的关节状态(左臂7个关节 + 左夹爪8个参数) |
action |
float32 | (8,) | 动作指令(左臂7个关节 + 左夹爪) |
observation.images.head |
video | (3, 720, 1280) | 头部摄像头视频(AV1编码,30FPS,720p) |
observation.images.left_wrist |
video | (3, 240, 424) | 左腕部摄像头视频(AV1编码,30FPS,240p) |
observation.images.right_wrist |
video | (3, 240, 424) | 右腕部摄像头视频(AV1编码,30FPS,240p) |
timestamp |
float32 | (1,) | 时间戳 |
frame_index |
int64 | (1,) | 帧索引 |
episode_index |
int64 | (1,) | 集索引 |
index |
int64 | (1,) | 全局索引 |
task_index |
int64 | (1,) | 任务索引 |
intervention |
bool | (1,) | 干预标记 |
任务与场景
- 数据集专注于板插入任务,使用左侧机械臂和夹爪进行操作。
- 包含来自头部、左腕部和右腕部的多视角视频观测。



