yinongh/no_image_task2
收藏资源简介:
该数据集是使用LeRobot创建的机器人学数据集,专注于机器人操作任务。数据集包含18个episodes、1137个frames和1个任务,所有数据均用于训练。数据以parquet格式存储,并包含72个视频。数据集的特征包括机器人动作(如delta_x、delta_y等9维向量)和多种观测数据,如末端执行器内部力、姿态、深度图像(如对齐的插座深度图像和初始插头深度图像)、点云数据(初始手腕点云)、相机外参矩阵以及来自手腕相机和辅助相机的RGB和深度图像。观测数据还包括时间戳、帧索引、episode索引等元信息。数据集的帧率为30fps,适用于机器人控制和学习任务的研究。
This robotics dataset is developed using LeRobot, focusing on robotic manipulation tasks. It contains 18 episodes, 1137 frames, and 1 single task, with all data designated for training. The data is stored in Parquet format and includes 72 videos. The dataset features robotic action data (e.g., 9-dimensional vectors such as delta_x, delta_y) and multiple types of observational data: internal forces and poses of the end-effector, depth images (e.g., aligned socket depth image and initial plug depth image), point cloud data (initial wrist point cloud), camera extrinsic matrices, as well as RGB and depth images captured by the wrist camera and auxiliary cameras. The observational data also encompasses metadata such as timestamps, frame indices, and episode indices. The dataset has a frame rate of 30fps and is suitable for research on robotic control and learning tasks.




