遇见数据集

cell1_20260519_mohammed_loge-color-sorting_20260519_175316

收藏
Hugging Face2026-05-22 更新2026-05-22 收录
官方服务:

资源简介:

该数据集是一个机器人学数据集,使用LeRobot工具创建,专门针对starpilot_yam_gripper机器人类型。数据集包含6个episodes,总计2221帧,涉及1个任务(可能为颜色分类或类似操作)。数据以parquet文件格式存储,视频文件为mp4格式,帧率为30fps。数据集特征包括:动作数据(14维浮点数,控制两个机械臂的x、y、z位置、rx、ry、rz旋转和夹爪宽度);观察状态数据(32维浮点数,包括机械臂的编码器角度、IMU加速度和陀螺仪数据、位姿(位置和四元数)、夹爪开合状态和距离);多个摄像头图像数据,包括左腕、右腕和基座摄像头的RGB图像和深度图(分辨率分别为480x640和768x1024,3通道);以及时间戳、帧索引、episode索引等元数据。数据集主要用于机器人控制、视觉感知和强化学习研究。

This is a robotics dataset created using the LeRobot tool, specifically tailored for the starpilot_yam_gripper robotic platform. The dataset contains 6 episodes with a total of 2221 frames, focusing on 1 task (potentially color classification or analogous operations). The data is stored in Parquet file format, with supporting video files in MP4 format at a frame rate of 30 fps. The dataset features the following components: 1. Action data: 14-dimensional floating-point vectors used to control the x, y, and z positions, rx, ry, and rz rotations of the two robotic arms, as well as the gripper width; 2. Observation state data: 32-dimensional floating-point vectors including robotic arm encoder angles, IMU acceleration and gyroscope data, poses (position and quaternion), gripper open/close status and distance metrics; 3. Multiple camera image streams, including RGB images and depth maps from the left wrist, right wrist, and base-mounted cameras, with resolutions of 480×640 and 768×1024 respectively and 3 color channels; 4. Metadata such as timestamps, frame indices, and episode indices. This dataset is primarily intended for research in robotic control, visual perception, and reinforcement learning.

提供机构:
nodogoro
创建时间:
2026-05-22
原始信息汇总

数据集概述

数据集名称: cell1_20260519_mohammed_loge-color-sorting_20260519_175316

许可协议: Apache-2.0

任务类型: 机器人学 (Robotics)

数据集创建工具: LeRobot (https://github.com/huggingface/lerobot)


数据集结构

版本: v3.0

机器人类型: starpilot_yam_gripper

基本统计:

  • 总片段数: 6
  • 总帧数: 2221
  • 总任务数: 1
  • 块大小: 1000
  • 数据文件大小: 100 MB
  • 视频文件大小: 200 MB
  • 帧率 (FPS): 30

数据划分:

  • 训练集: 片段索引 0 到 5 (共6个片段)

数据路径:

  • data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet
  • videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4

特征详情

1. 动作 (action)

  • 数据类型: float32
  • 维度: 14
  • 包含字段: arm1_x, arm1_y, arm1_z, arm1_rx, arm1_ry, arm1_rz, arm1_grip_width, arm2_x, arm2_y, arm2_z, arm2_rx, arm2_ry, arm2_rz, arm2_grip_width

2. 观测状态 (observation.state)

  • 数据类型: float32
  • 维度: 32
  • 包含字段(机械臂1及机械臂2): enc_deg, imu_ax, imu_ay, imu_az, imu_gx, imu_gy, imu_gz, pose_x, pose_y, pose_z, pose_qx, pose_qy, pose_qz, pose_qw, gripper_open, gripper_distance_m

3. 观测图像 (observation.images)

  • 摄像头来源及参数:
    • left_wrist_0_camera_rgb_image: 分辨率 480x640, 3通道, h264编码, 30fps
    • left_wrist_0_camera_rgb_image_depth: 分辨率 480x640, 3通道, h264编码, 30fps
    • right_wrist_0_camera_rgb_image: 分辨率 480x640, 3通道, h264编码, 30fps
    • right_wrist_0_camera_rgb_image_depth: 分辨率 480x640, 3通道, h264编码, 30fps
    • base_0_camera_rgb_image: 分辨率 480x640, 3通道, h264编码, 30fps
    • base_1_camera_rgb_image: 分辨率 768x1024, 3通道, h264编码, 30fps

4. 其他特征

  • timestamp (float32, 1维)
  • frame_index (int64, 1维)
  • episode_index (int64, 1维)
  • index (int64, 1维)
  • task_index (int64, 1维)

可视化

交互式数据集可视化入口:


引用信息

BibTeX: 暂无提供。

二维码
社区交流群
二维码
科研交流群
商业服务