遇见数据集

eval_smolvla_sorting_action_expert_ft_121ep_step160000_A_OOD_v1

收藏
Hugging Face2026-06-03 更新2026-06-03 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,专门用于机器人控制和学习任务。它包含3个episodes,总计4035帧,对应1个任务,帧率为30fps。数据以parquet格式存储,总数据文件大小为100MB,视频文件大小为200MB。特征包括:动作数据(action),包含6个浮点型关节位置(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);观测状态(observation.state),同样包含6个浮点型关节位置;两个图像观测(observation.images.top和observation.images.wrist),均为视频格式,分辨率480x640,3通道(RGB),使用av1编解码器,无音频;以及时间戳、帧索引、episode索引、索引和任务索引等元数据。适用于机器人模拟、动作学习和视觉-动作映射等研究。

This dataset is a robotics dataset created using the LeRobot tool, specifically designed for robot control and learning tasks. It contains 3 episodes, totaling 4035 frames, corresponding to 1 task, with a frame rate of 30fps. Data is stored in parquet format, with a total data file size of 100MB and a video file size of 200MB. Features include: action data (action), comprising 6 floating-point joint positions (e.g., shoulder translation, shoulder lift, elbow bend, wrist bend, wrist rotation, and gripper position); observation state (observation.state), also with 6 floating-point joint positions; two image observations (observation.images.top and observation.images.wrist), both in video format with a resolution of 480x640, 3 channels (RGB), using the av1 codec, without audio; and metadata such as timestamps, frame index, episode index, index, and task index. It is suitable for research in robot simulation, action learning, and vision-action mapping.

提供机构:
Harumo
创建时间:
2026-06-03
原始信息汇总

数据集概述

  • 数据集名称: eval_smolvla_sorting_action_expert_ft_121ep_step160000_A_OOD_v1
  • 许可证: Apache-2.0
  • 任务类别: 机器人学 (robotics)
  • 标签: LeRobot
  • 来源: 使用 LeRobot 创建

数据集规模

  • 总片段数 (Total Episodes): 3
  • 总帧数 (Total Frames): 4035
  • 总任务数 (Total Tasks): 1
  • 帧率 (FPS): 30
  • 数据文件大小: 100 MB
  • 视频文件大小: 200 MB
  • 块大小 (Chunks Size): 1000

数据集结构

  • 机器人类型: so_follower
  • 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
  • 数据格式: Parquet 文件 + MP4 视频文件
  • 划分: 训练集包含全部 3 个片段 (train: "0:3")

特征说明

特征名称 数据类型 形状 字段说明
action float32 [6] 6 个关节动作:shoulder_pan.pos, shoulder_lift.pos, elbow_flex.pos, wrist_flex.pos, wrist_roll.pos, gripper.pos
observation.state float32 [6] 6 个关节状态:与 action 名称相同
observation.images.top video [480, 640, 3] 顶部摄像头视频,分辨率 480x640,3 通道,AV1 编码
observation.images.wrist video [480, 640, 3] 腕部摄像头视频,分辨率 480x640,3 通道,AV1 编码
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
二维码
社区交流群
二维码
科研交流群
商业服务