eval_act_CameraOnGripperColoredPiece

该数据集由LeRobot创建，专为机器人学习任务设计，包含10个完整的情节（episodes），总计5619帧数据，采样率为每秒30帧。数据以parquet文件和视频文件形式存储，总数据量约100MB，视频文件约200MB。数据集特征包括：6维浮点型动作向量，用于控制机器人关节（包括肩部平移、肩部升降、肘部弯曲、腕部弯曲、腕部旋转和夹爪位置）；6维浮点型观测状态，对应相同关节位置；夹爪摄像头拍摄的640x480分辨率RGB图像观测；以及时间戳、帧索引、情节索引等元数据。数据集仅包含训练集，适用于机器人控制与模仿学习研究。This dataset is created by LeRobot for robot learning tasks, containing 10 complete episodes with a total of 5619 frames sampled at 30 frames per second. The data is stored in parquet files and video files, with a total size of approximately 100MB and video files of about 200MB. Features include: 6-dimensional floating-point action vectors (controlling shoulder translation, shoulder elevation, elbow flexion, wrist flexion, wrist rotation, and gripper position), 6-dimensional floating-point observation states (corresponding to the same joint positions), RGB image observations from a gripper camera at 640x480 resolution, and metadata such as timestamps, frame indices, and episode indices. The dataset includes only a training set and is suitable for robot control and imitation learning research.

原始信息汇总

数据集概述

数据集名称: eval_act_CameraOnGripperColoredPiece

许可协议: Apache-2.0

任务类别: 机器人学 (Robotics)

相关标签: LeRobot

---

数据集结构

• 代码版本: v3.0
• 机器人类型: so101_follower
• 总回合数: 10
• 总帧数: 5619
• 总任务数: 1
• 数据块大小: 1000
• 数据文件大小: 100 MB
• 视频文件大小: 200 MB
• 帧率 (FPS): 30

数据集划分

• 训练集: 全部 10 个回合 (序号 0 到 9)

数据存储路径

• 数据文件: data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet
• 视频文件: videos/{video_key}/chunk-{chunk_index:03d}/file-{file_index:03d}.mp4

---

特征说明

特征名称	数据类型	形状	描述
action	float32	[6]	机器人动作：shoulder_pan.pos, shoulder_lift.pos, elbow_flex.pos, wrist_flex.pos, wrist_roll.pos, gripper.pos
observation.state	float32	[6]	机器人状态观测，与动作字段具有相同的6个关节位置
observation.images.onGripper	video	[640, 480, 3]	安装在夹爪上的摄像头拍摄的视频，分辨率640×480，AV1编码，30 FPS
timestamp	float32	[1]	时间戳
frame_index	int64	[1]	帧序号
episode_index	int64	[1]	回合序号
index	int64	[1]	索引
task_index	int64	[1]	任务序号