ajaysri/lego_stack_4x4_regular_plus_correction_frontdot_backwrist_lerobot_v3
收藏资源简介:
这是一个用于机器人学习的数据集,专注于UR5机器人配合WSG50夹爪进行Lego积木堆叠任务。数据集通过LeRobot平台创建,包含遥操作和仿真数据。数据集结构包括202个完整episodes,总计80912帧,数据以parquet格式存储,视频文件以mp4格式存储,帧率为20fps。特征数据涵盖:主摄像头和腕部摄像头的视频观察(分辨率160x240,3通道RGB),机器人状态观察(7维浮点数),动作数据(7维浮点数),以及提示文本(字符串)、时间戳、帧索引、episode索引等元数据。数据集主要用于训练和评估机器人控制算法,特别是堆叠任务中的视觉-运动策略学习。
This is a dataset for robotics learning, focusing on UR5 robot with WSG50 gripper performing Lego stacking tasks. It was created using the LeRobot platform and includes teleoperation and simulation data. The dataset structure consists of 202 total episodes, with 80912 total frames, stored in parquet format for data and mp4 for video files at 20fps. Features include video observations from main and wrist cameras (resolution 160x240, 3-channel RGB), robot state observations (7-dimensional float32), action data (7-dimensional float32), and metadata such as prompt text (string), timestamp, frame index, episode index, etc. The dataset is designed for training and evaluating robot control algorithms, particularly for visual-motor policy learning in stacking tasks.




