kinova3_pickplace

该数据集是一个机器人学习数据集，使用LeRobot工具创建。它包含51个episodes，总计23901帧数据，分为102个视频。数据以parquet格式存储，帧率为20fps，并包含训练集分割。数据集的特征包括：从agent视角和手腕视角采集的224x224x3分辨率视频图像（编码为h264），13维的浮点状态向量，7维的浮点动作向量，以及时间戳、帧索引、episode索引等元数据。这些数据用于机器人任务，如动作学习和状态观察。This dataset is a robot learning dataset created using the LeRobot tool. It contains 51 episodes, totaling 23901 frames of data, divided into 102 videos. The data is stored in parquet format with a frame rate of 20fps and includes a training set split. The features of the dataset include: 224x224x3 resolution video images (encoded as h264) captured from the agents perspective and wrist perspective, a 13-dimensional floating-point state vector, a 7-dimensional floating-point action vector, and metadata such as timestamps, frame indices, and episode indices. This data is used for robot tasks such as action learning and state observation.

原始信息汇总

数据集概述

• 数据集名称: kinova3_pickplace
• 许可证: Apache-2.0
• 任务类别: 机器人学 (Robotics)
• 标签: LeRobot

数据集规模

• 总片段数 (Episodes): 51
• 总帧数 (Frames): 23,901
• 总任务数: 1
• 总视频数: 102
• 帧率 (FPS): 20
• 数据路径: data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet
• 视频路径: videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4
• 训练/测试划分: 全部51个片段用于训练 (train: 0:51)

数据集特征 (Features)

特征名	数据类型	形状	说明
observation.images.agentview	视频 (h264)	[224, 224, 3]	机器人视角摄像头图像 (20 FPS)
observation.images.wrist	视频 (h264)	[224, 224, 3]	腕部摄像头图像 (20 FPS)
observation.state	float32	[13]	机器人状态 (13维)
action	float32	[7]	动作指令 (7维)
timestamp	float32	[1]	时间戳
frame_index	int64	[1]	帧索引
episode_index	int64	[1]	片段索引
index	int64	[1]	全局索引
task_index	int64	[1]	任务索引

创建工具

数据集使用 LeRobot 创建。