遇见数据集

rollout_recap_stackblocks_iter0_7_20260704_192949

收藏
Hugging Face2026-07-05 更新2026-07-05 收录
官方服务:

资源简介:

该数据集由LeRobot创建,专用于机器人控制和学习任务。它包含多个特征:动作(action)为6维浮点数组,表示机器人各关节的位置(包括肩部平移、肩部提升、肘部弯曲、腕部弯曲、腕部旋转和夹爪位置);观测状态(observation.state)同样为6维浮点数组,表示机器人当前状态;观测图像包括顶部视角(observation.images.top)和腕部视角(observation.images.wrist),均为视频格式,分辨率为640x480,3通道,帧率30fps;干预标志(intervention)为布尔值,指示是否有人为干预;下一成功标志(next.success)为布尔值,表示下一步是否成功;时间戳(timestamp)为浮点数;此外还有帧索引(frame_index)、回合索引(episode_index)、索引(index)和任务索引(task_index)等整数特征。数据集规模包括2个回合、2998帧和1个任务,数据以parquet格式存储,视频以mp4格式存储。

This dataset is created by LeRobot and is primarily used for robot control and learning tasks. It includes multiple features: action is a 6-dimensional floating-point array representing the positions of robot joints (such as shoulder translation, shoulder lift, elbow bend, wrist bend, wrist rotation, and gripper position); observation.state is also a 6-dimensional floating-point array indicating the current state of the robot; observation images consist of top view (observation.images.top) and wrist view (observation.images.wrist), both in video format with a resolution of 640x480, 3 channels, and a frame rate of 30fps; intervention is a boolean flag indicating whether human intervention occurred; next.success is a boolean flag indicating whether the next step is successful; timestamp is a floating-point number; additionally, there are integer features such as frame_index, episode_index, index, and task_index. The dataset comprises 2 episodes, 2998 frames, and 1 task, with data stored in parquet format and videos in mp4 format.

提供机构:
lilkm
创建时间:
2026-07-05
原始信息汇总
  • 数据集名称: lilkm/rollout_recap_stackblocks_iter0_7_20260704_192949
  • 许可证: Apache-2.0
  • 任务类别: 机器人学(Robotics)
  • 创建工具: 使用 LeRobot 创建
  • 数据集结构:
    • 总片段数: 2
    • 总帧数: 2998
    • 总任务数: 1
    • 帧率: 30 FPS
    • 数据文件大小: 100 MB
    • 视频文件大小: 200 MB
    • 数据路径: data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet
    • 视频路径: videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4
    • 机器人类型: so_follower
    • 数据集划分: 训练集 train: 0:2
  • 特征(Features):
    • action: 6维浮点数组,包含机械臂关节位置(肩部、肘部、腕部、夹爪)
    • observation.state: 6维浮点数组,与 action 结构相同
    • observation.images.top: 顶视视频,分辨率 480x640,3通道,AV1 编码,30 FPS
    • observation.images.wrist: 腕部视频,分辨率 480x640,3通道,AV1 编码,30 FPS
    • intervention: 布尔值,指示是否干预
    • next.success: 布尔值,指示下一步是否成功
    • timestamp: 浮点时间戳
    • frame_index: 帧索引(int64)
    • episode_index: 片段索引(int64)
    • index: 全局索引(int64)
    • task_index: 任务索引(int64)
二维码
社区交流群
二维码
科研交流群
商业服务