test_lerobot_vis

该数据集采用LeRobot `v3.0`格式，包含16个OmniViTac机器人操作的片段，总计3278帧，帧率为15.0 FPS。数据集为多模态，包含视频、表格和时间序列数据。数据结构包括元信息、数据文件和视频文件。特征部分详细描述了各种标量值、状态/动作向量、深度测量、视频和源时间戳。

原始信息汇总

OmniViTac LeRobot v3 (Test Run) 数据集概述

数据集基本信息

• 许可证: Apache-2.0
• 任务类别: 机器人学
• 名称: OmniViTac LeRobot v3 (Test Run)
• 语言: 英语
• 标签: LeRobot, 机器人学, 多模态, 视频, 表格, 时间序列
• 格式版本: LeRobot v3.0

数据集摘要

• 机器人类型: OmniViTac
• 总情节数: 16
• 总帧数: 3278
• 帧率: 15.0 FPS
• 数据分割: train: 0:16
• 数据文件路径模板: data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet
• 视频文件路径模板: videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4

数据集结构

文件目录结构

meta/ info.json stats.json tasks.parquet episodes/chunk-000/file-000.parquet data/ chunk-000/file-000.parquet videos/ observation.images.camera1/chunk-000/file-000.mp4 observation.images.camera2/chunk-000/file-000.mp4 observation.images.tactile1_raw/chunk-000/file-000.mp4 observation.images.tactile2_raw/chunk-000/file-000.mp4 observation.images.tactile1_diff/chunk-000/file-000.mp4 observation.images.tactile2_diff/chunk-000/file-000.mp4

元数据信息 (meta/info.json)

• 代码库版本: v3.0
• 总任务数: 1
• 数据块大小: 1000
• 数据文件总大小: 100 MB
• 视频文件总大小: 500 MB

数据特征

标量特征

• timestamp (float32, shape: [1])
• frame_index (int64, shape: [1])
• episode_index (int64, shape: [1])
• index (int64, shape: [1])
• task_index (int64, shape: [1])

状态/动作类向量特征

• observation.state (float32, shape: [6], 轴: ["x", "y", "z", "roll", "pitch", "yaw"])
• observation.joints (float32, shape: [7], 轴: ["joint_0", "joint_1", "joint_2", "joint_3", "joint_4", "joint_5", "joint_6"])
• observation.gripper (float32, shape: [1], 轴: ["gripper_pos"])
• observation.joint_stamps (int64, shape: [1], 轴: ["joint_stamp_ms"])
• observation.tactile1_deform (float32, shape: [2100])
• observation.tactile2_deform (float32, shape: [2100])

深度图像特征

• observation.depth.camera1 (float16, shape: [480, 640], 轴: ["height", "width"])
• observation.depth.camera2 (float16, shape: [720, 1280], 轴: ["height", "width"])

视频特征

• observation.images.camera1 (video dtype, shape: [3, 480, 640], 轴: ["channel", "height", "width"])
• observation.images.camera2 (video dtype, shape: [3, 720, 1280], 轴: ["channel", "height", "width"])
• observation.images.tactile1_raw (video dtype, shape: [3, 700, 400], 轴: ["channel", "height", "width"])
• observation.images.tactile2_raw (video dtype, shape: [3, 700, 400], 轴: ["channel", "height", "width"])
• observation.images.tactile1_diff (video dtype, shape: [3, 700, 400], 轴: ["channel", "height", "width"])
• observation.images.tactile2_diff (video dtype, shape: [3, 700, 400], 轴: ["channel", "height", "width"])

源时间戳特征

• observation.source_timestamp.camera1_ms (int64, shape: [1])
• observation.source_timestamp.camera2_ms (int64, shape: [1])
• observation.source_timestamp.depth1_ms (int64, shape: [1])
• observation.source_timestamp.depth2_ms (int64, shape: [1])
• observation.source_timestamp.state_ms (int64, shape: [1])
• observation.source_timestamp.joint_stamps_ms (int64, shape: [1])
• observation.source_timestamp.gripper_ms (int64, shape: [1])
• observation.source_timestamp.tactile1_deform_ms (int64, shape: [1])
• observation.source_timestamp.tactile2_deform_ms (int64, shape: [1])
• observation.source_timestamp.tactile1_raw_ms (int64, shape: [1])
• observation.source_timestamp.tactile2_raw_ms (int64, shape: [1])
• observation.source_timestamp.tactile1_diff_ms (int64, shape: [1])
• observation.source_timestamp.tactile2_diff_ms (int64, shape: [1])