2026-04-23_direction_have_cap_water-lerobot-without-rinse
收藏资源简介:
该数据集由LeRobot项目创建,是一个用于机器人学研究的开源数据集。数据集记录了双手机器人(ALOHA类型)的操作数据,包含机器人的状态观察、执行的动作以及多视角的视频信息。数据集总规模包含9个完整操作序列(episodes),共计9801个数据帧,对应1个任务。数据以Parquet文件格式存储,总数据量约为200MB,同时包含约200MB的视频文件,视频帧率为50fps。数据集仅包含训练集划分。核心数据特征包括:14维的机器人关节位置观察状态(涵盖左/右侧的腰部、肩部、肘部、前臂旋转、手腕角度、手腕旋转和夹爪状态);14维的关节动作指令;14维的关节速度;14维的关节力矩。此外,数据集提供了来自四个固定摄像头的同步视频观察:一个高位摄像头、一个低位摄像头以及左、右手腕摄像头。所有视频均为彩色(3通道),分辨率480x640,采用AV1编码,帧率50fps。数据集还包含任务索引、子任务描述字符串、是否为训练数据的标志、时间戳、帧索引、episode索引和样本索引等元数据。该数据集适用于机器人模仿学习、行为克隆、视觉运动策略学习等任务的研究与模型训练,遵循Apache-2.0开源协议。
This dataset, developed by the LeRobot project, is an open-source resource for robotics research. It records operation data from ALOHA-type dual-arm robots, including robot state observations, executed actions, and multi-view video information. The dataset contains 9 complete operation episodes, totaling 9801 data frames, corresponding to a single task. The data is stored in Parquet file format, with a total size of approximately 200MB, alongside around 200MB of video files at 50fps. Only the training split is provided in this dataset. Core data features include 14-dimensional robot joint position observation states (covering left/right waist, shoulder, elbow, forearm rotation, wrist angle, wrist rotation, and gripper status), 14-dimensional joint action commands, 14-dimensional joint velocities, and 14-dimensional joint torques. In addition, the dataset provides synchronized video observations from four fixed cameras: one overhead camera, one low-position camera, and left and right wrist cameras. All videos are 3-channel color with a resolution of 480x640, encoded in AV1, and have a frame rate of 50fps. The dataset also includes metadata such as task index, subtask description strings, training data flag, timestamp, frame index, episode index, and sample index. This dataset is suitable for research and model training on tasks including robot imitation learning, behavioral cloning, and visuomotor policy learning, and follows the Apache-2.0 open-source license.




