遇见数据集

rollout_vla_jepa_so101_pick_toy_v9_plate_20260624_202426

收藏
Hugging Face2026-06-24 更新2026-06-24 收录
官方服务:

资源简介:

该数据集由LeRobot创建,专用于机器人学任务。数据集包含一个episode,总计1836帧,数据以parquet格式存储,视频以mp4格式存储。数据集结构包括动作(action)和观测(observation)字段:动作和状态观测均为6维浮点数组,表示机器人关节位置,具体包括肩部平移、肩部抬升、肘部弯曲、腕部弯曲、腕部旋转和夹爪位置。观测部分包含来自两个外部摄像头(exterior_1_left和exterior_2_left)的视频图像,分辨率为480x640,3通道,帧率为15fps,使用AV1编解码器。此外,数据集还包括时间戳、帧索引、episode索引、索引和任务索引等元数据。机器人类型为so_follower,数据集总大小约为300MB(其中数据部分100MB,视频部分200MB)。数据集使用Apache-2.0许可证,适用于训练和评估机器人控制模型。

This dataset was developed by LeRobot for robotics tasks. It contains one episode with a total of 1836 frames. The structured data is stored in Parquet format, while the accompanying videos are stored in MP4 format. The dataset structure includes action and observation fields: both the action and state observation are 6-dimensional floating-point arrays representing robot joint positions, including shoulder translation, shoulder elevation, elbow flexion, wrist flexion, wrist rotation, and gripper position. The observation section includes video images from two external cameras (exterior_1_left and exterior_2_left), with a resolution of 480x640, 3 color channels, a frame rate of 15 fps, and using the AV1 codec. Additionally, the dataset includes metadata such as timestamps, frame indices, episode indices, index, and task indices. The robot type is 'so_follower'. The total size of the dataset is approximately 300 MB, with 100 MB for the data and 200 MB for the videos. This dataset is licensed under the Apache-2.0 license and is suitable for training and evaluating robot control models.

提供机构:
kamada-riku
创建时间:
2026-06-24
原始信息汇总

数据集概述

  • 数据集名称:rollout_vla_jepa_so101_pick_toy_v9_plate_20260624_202426
  • 许可证:Apache-2.0
  • 任务类别:机器人学(robotics)
  • 标签:LeRobot
  • 创建工具:使用 LeRobot 创建

数据集结构

  • 数据格式:Data files 为 Parquet 格式(data/*/*.parquet),视频文件为 MP4 格式(videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4
  • 帧率:15 FPS
  • 机器人类型:so_follower

数据特征

特征名 数据类型 形状 描述
action float32 [6] 包含 6 个动作维度:shoulder_pan.pos, shoulder_lift.pos, elbow_flex.pos, wrist_flex.pos, wrist_roll.pos, gripper.pos
observation.state float32 [6] 包含与 action 相同的 6 个状态维度
observation.images.exterior_1_left video [480, 640, 3] 外部左侧摄像头 1 的视频,AV1 编码,yuv420p 像素格式,15 FPS
observation.images.exterior_2_left video [480, 640, 3] 外部左侧摄像头 2 的视频,AV1 编码,yuv420p 像素格式,15 FPS
timestamp float32 [1] 时间戳
frame_index int64 [1] 帧索引
episode_index int64 [1] 剧集索引
index int64 [1] 索引
task_index int64 [1] 任务索引

数据规模

  • 总剧集数:1
  • 总帧数:1836
  • 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

数据集划分

  • 训练集:索引范围 0:1(全部数据用于训练)
二维码
社区交流群
二维码
科研交流群
商业服务