yinongh/adaptation_task_2
收藏资源简介:
该数据集由LeRobot创建,用于机器人技术任务,专注于记录机器人在执行任务过程中的动作和观测数据。数据集包含20个训练集片段,总计1331帧数据,帧率为30fps,涵盖1个任务和80个视频。数据特征包括机器人的9维动作向量(如位置和旋转变化)、观测数据(如末端执行器内部力、姿态矩阵、对齐的深度图像、初始深度图像、点云数据、相机外参矩阵以及来自手腕摄像头和辅助摄像头的RGB和深度图像)。数据集以Parquet文件格式组织,按片段和索引存储,支持机器人学习算法的训练和评估。
This dataset was created using LeRobot for robotics tasks, focusing on recording robot actions and observations during task execution. It includes 20 training episodes, totaling 1331 frames at 30fps, with 1 task and 80 videos. Features encompass 9-dimensional robot actions (e.g., position and rotation deltas), observations such as end-effector internal forces, pose matrices, aligned depth images, initial depth images, point cloud data, camera extrinsic matrices, and RGB/depth images from wrist and auxiliary cameras. The dataset is organized in Parquet files by episodes and indices, supporting training and evaluation of robot learning algorithms.




