pick_cubo_2
收藏资源简介:
该数据集是一个机器人技术数据集,专门用于pick_cubo_2任务(从可视化链接推断),旨在支持机器人学习和控制研究。数据集使用LeRobot框架创建,包含40个总集数和21250个总帧数,以25帧每秒的速率采集。数据特征包括:动作数据(9维浮点数,表示机器人关节位置和速度)、观察状态(9维浮点数,与动作数据相同)、前视图像和腕部图像(均为视频格式,分辨率480x640,3通道,使用av1编解码器,无音频)。此外,数据集还包含时间戳、帧索引、集索引、索引和任务索引等元数据。数据以Parquet文件格式存储,视频以MP4格式存储,总数据文件大小为100MB,视频文件大小为200MB。机器人类型为lekiwi_client,数据集仅包含训练集(覆盖所有40个集)。该数据集适用于机器人视觉-动作映射、强化学习或模仿学习等应用场景。
This is a robotics dataset specifically designed for the pick_cubo_2 task (inferred from the visualization link), aiming to support robotics learning and control research. The dataset was developed using the LeRobot framework, with a total of 40 episodes and 21250 total frames collected at a rate of 25 frames per second. The dataset includes the following features: 9-dimensional floating-point action vectors representing the robot's joint positions and velocities; 9-dimensional observation floating-point vectors identical to the action data; front-view camera images and wrist-worn camera images, both in video format with a resolution of 480×640, 3 color channels, encoded using the av1 codec and containing no audio. Additionally, the dataset contains metadata such as timestamps, frame indices, episode indices, sample indices and task indices. The structured data is stored in Parquet file format, while the videos are stored in MP4 format. The total size of the data files is 100 MB, and the total size of the video files is 200 MB. The robot platform used is lekiwi_client, and the dataset only contains the training split covering all 40 episodes. This dataset is applicable to scenarios such as robotic vision-action mapping, reinforcement learning and imitation learning.
数据集概述:pick_cubo_2
- 许可证:Apache-2.0
- 任务类别:机器人技术
- 标签:LeRobot
- 创建工具:LeRobot(https://github.com/huggingface/lerobot)
数据集结构
- 代码库版本:v3.0
- 帧率(FPS):25
- 机器人类型:lekiwi_client
- 总演示片段数:40
- 总帧数:21,250
- 任务数量:1
- 数据文件大小:约100 MB
- 视频文件大小:约200 MB
特征
-
动作(action):
- 数据类型:float32
- 形状:[9]
- 命名:arm_shoulder_pan.pos, arm_shoulder_lift.pos, arm_elbow_flex.pos, arm_wrist_flex.pos, arm_wrist_roll.pos, arm_gripper.pos, x.vel, y.vel, theta.vel
-
观察状态(observation.state):
- 数据类型:float32
- 形状:[9]
- 命名:arm_shoulder_pan.pos, arm_shoulder_lift.pos, arm_elbow_flex.pos, arm_wrist_flex.pos, arm_wrist_roll.pos, arm_gripper.pos, x.vel, y.vel, theta.vel
-
观察图像 - 前置摄像头(observation.images.front):
- 数据类型:视频
- 形状:[480, 640, 3](高度 × 宽度 × 通道数)
- 编码:AV1,像素格式 yuv420p,帧率 25,无音频
-
观察图像 - 腕部摄像头(observation.images.wrist):
- 数据类型:视频
- 形状:[480, 640, 3]
- 编码:AV1,像素格式 yuv420p,帧率 25,无音频
-
时间戳(timestamp):float32,形状 [1]
-
帧索引(frame_index):int64,形状 [1]
-
演示片段索引(episode_index):int64,形状 [1]
-
索引(index):int64,形状 [1]
-
任务索引(task_index):int64,形状 [1]
数据集划分
- 训练集:0:40(所有40个演示片段)
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
可视化
- 可通过以下链接可视化数据集:https://huggingface.co/spaces/lerobot/visualize_dataset?path=RobotSo101/pick_cubo_2



