five

so101_cap_sort_20260703_145131

收藏
Hugging Face2026-07-03 更新2026-07-03 收录
下载链接:
https://huggingface.co/datasets/fouad1233/so101_cap_sort_20260703_145131
下载链接
链接失效反馈
官方服务:
资源简介:
该数据集由LeRobot创建,专注于机器人技术领域,适用于机器人控制任务。数据集包含50个总片段和40613个总帧,数据以parquet文件格式存储,帧率为30 fps。特征包括动作(action),表示机械臂的6个关节位置(如肩部平移、肩部提升、肘部弯曲、腕部弯曲、腕部旋转和夹爪位置);观察状态(observation.state),同样包含6个关节位置;以及观察图像,包括基础摄像头(base_0_rgb)和左腕摄像头(left_wrist_0_rgb)的RGB视频,分辨率为480x640,3通道。此外,还包含时间戳、帧索引、片段索引、索引和任务索引等元数据。数据集用于模拟或实际机器人操作场景,支持训练和评估机器人学习算法。

This dataset was developed by LeRobot, targeting the robotics domain and suitable for robotic control tasks. It consists of 50 total segments and 40613 total frames, stored in Parquet file format with a frame rate of 30 fps. Its features include the action field, which represents the 6 joint positions of the robotic arm (e.g., shoulder translation, shoulder lift, elbow flexion, wrist flexion, wrist rotation, and gripper position); the observation.state field, which also contains the 6 joint positions; and observation images, which are RGB videos captured by the base camera (base_0_rgb) and left wrist camera (left_wrist_0_rgb) with a resolution of 480×640 and 3 color channels. Additionally, it encompasses metadata such as timestamps, frame indices, segment indices, entry indices, and task indices. This dataset is designed for simulated or real robotic manipulation scenarios, and supports the training and evaluation of robotic learning algorithms.
提供机构:
fouad1233
创建时间:
2026-07-03
原始信息汇总

数据集概况

  • 数据集名称:so101_cap_sort_20260703_145131
  • 许可证:Apache-2.0
  • 任务类别:机器人(Robotics)
  • 标签:LeRobot
  • 创建工具:LeRobot(https://github.com/huggingface/lerobot)

数据集结构

属性
代码库版本 v3.0
帧率(FPS) 30
总片段数(总episodes) 50
总帧数(总frames) 40613
总任务数 1
数据块大小 1000
数据文件大小(MB) 100
视频文件大小(MB) 200
机器人类型 so_follower
数据路径模板 data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet
视频路径模板 videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4
训练集划分 索引 0 到 49(共50个片段)

特征(Features)

特征键 数据类型 形状 说明
action float32 [6] 6个关节的动作指令(肩部、肘部、腕部、夹爪位置)
observation.state float32 [6] 6个关节的观测状态(与动作命名相同)
observation.images.base_0_rgb video [480, 640, 3] 基础摄像头RGB视频,分辨率480x640,AV1编码,30FPS
observation.images.left_wrist_0_rgb video [480, 640, 3] 左腕摄像头RGB视频,分辨率480x640,AV1编码,30FPS
timestamp float32 [1] 时间戳
frame_index int64 [1] 帧索引
episode_index int64 [1] 片段索引
index int64 [1] 索引
task_index int64 [1] 任务索引

数据特征详情

  • 关节动作/状态名称(6维):

    • shoulder_pan.pos
    • shoulder_lift.pos
    • elbow_flex.pos
    • wrist_flex.pos
    • wrist_roll.pos
    • gripper.pos
  • 视频编码属性

    • 编解码器:AV1
    • 像素格式:yuv420p
    • 帧率:30 FPS
    • 通道数:3
    • 无音频
二维码
社区交流群
二维码
科研交流群
商业服务