libero_plus_object_lighting_all_sawyer
收藏资源简介:
这是一个基于Sawyer机器人平台的多视角视觉运动控制数据集,用于机器人任务研究。数据集包含888个episode和132080帧数据,覆盖10个不同任务,数据特征包括:机器人状态观测(如末端执行器状态、关节状态、夹爪状态)、多视角图像(包括图像、腕部图像、前视图图像、侧视图图像,均为256x256分辨率的RGB视频,帧率20fps)、动作控制指令(7维向量)、时间戳和索引信息。数据集使用LeRobot工具创建,适用于机器人学习研究,特别关注多摄像头和光照条件的变化。数据以Parquet格式存储,总大小约100MB。
This is a multi-view visual motor control dataset based on the Sawyer robot platform for robotic task research. The dataset contains 888 episodes and 132,080 frames of data, covering 10 distinct tasks. The data features include robotic state observations (such as end-effector state, joint state, and gripper state), multi-view images (including wrist-mounted images, front-view images, and side-view images, all 256×256 resolution RGB videos with a frame rate of 20 fps), action control commands (7-dimensional vectors), as well as timestamp and index information. Created using the LeRobot toolkit, this dataset is tailored for robot learning research, with a particular focus on variations in multi-camera configurations and lighting conditions. The data is stored in Parquet format, with a total size of approximately 100 MB.
数据集概述:libero_plus_object_lighting_all_sawyer
- 数据集名称:libero_plus_object_lighting_all_sawyer
- 许可证:Apache-2.0
- 任务类别:机器人学(Robotics)
- 标签:LeRobot、libero、libero-plus、sawyer、multicam、lighting
- 数据集格式:Parquet 文件,路径为
data/*/*.parquet - 创建工具:使用 LeRobot 创建
数据集结构
- 代码库版本:v3.0
- 帧率(FPS):20
- 总片段数:888
- 总帧数:132,080
- 总任务数:10
- 分块大小:1000
- 数据文件大小:100 MB
- 视频文件大小:约 1e-06 MB
- 机器人类型:Sawyer
- 数据划分:训练集(train)包含全部 888 个片段(索引 0 至 887)
特征(Features)
数据集包含以下特征,每个特征的 dtype、shape 及名称如下:
| 特征名称 | dtype | shape | 名称 |
|---|---|---|---|
observation.state |
float32 | [8] | x, y, z, axis_angle1, axis_angle2, axis_angle3, gripper, gripper |
observation.states.ee_state |
float32 | [6] | x, y, z, axis_angle1, axis_angle2, axis_angle3 |
observation.states.joint_state |
float32 | [7] | joint_0 至 joint_6 |
observation.states.gripper_state |
float32 | [2] | gripper, gripper |
observation.images.image |
video | [256, 256, 3] | 高度、宽度、RGB |
observation.images.wrist_image |
video | [256, 256, 3] | 高度、宽度、RGB |
observation.images.frontview_image |
video | [256, 256, 3] | 高度、宽度、RGB |
observation.images.sideview_image |
video | [256, 256, 3] | 高度、宽度、RGB |
action |
float32 | [7] | x, y, z, axis_angle1, axis_angle2, axis_angle3, gripper |
timestamp |
float32 | [1] | — |
frame_index |
int64 | [1] | — |
episode_index |
int64 | [1] | — |
index |
int64 | [1] | — |
task_index |
int64 | [1] | — |
视频详情
所有图像特征(image、wrist_image、frontview_image、sideview_image)均采用 AV1 视频编码,分辨率为 256x256,帧率 20 FPS,颜色格式为 yuv420p,无音频,深度图标识为 false。视频路径格式为 videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4。
引用
目前引用信息尚未提供(状态为 [More Information Needed])。



