yinongh/real_world_no_image_1
收藏资源简介:
这是一个机器人操作数据集,使用LeRobot创建,用于机器人学习任务。数据集包含15个episodes、859帧、1个任务和60个视频,存储为parquet和mp4格式,帧率为30fps。数据特征包括机器人动作(9维浮点数,表示位置和旋转变化)、观测数据(如末端执行器内力、位姿、深度图像、点云、相机外参、RGB和深度图像等),以及时间戳、帧索引、episode索引等元数据。观测数据来自多个传感器,包括手腕相机和辅助Azure Kinect相机,提供丰富的视觉和物理信息,适用于机器人控制与感知研究。
This is a robotics manipulation dataset created using LeRobot, designed for robot learning tasks. It contains 15 episodes, 859 frames, 1 task, and 60 videos, stored in parquet and mp4 formats with a frame rate of 30fps. The dataset features robot actions (9-dimensional float32, representing position and rotation changes), observations (such as end-effector internal forces, pose, depth images, point clouds, camera extrinsics, RGB and depth images), and metadata like timestamps, frame indices, and episode indices. Observations are collected from multiple sensors, including a wrist camera and an auxiliary Azure Kinect camera, providing rich visual and physical information suitable for robotics control and perception research.




