yinongh/geoadapt_task2
收藏资源简介:
该数据集是使用LeRobot创建的机器人数据集,主要用于机器人任务。数据集包含15个完整集,总计886帧,覆盖1个任务,视频数量为60个,帧率为30 FPS。数据以Parquet格式存储,分割为训练集。特征包括动作(9维浮点数组,表示位置和旋转变化)、观测数据(如末端执行器内部力、位姿、对齐的插座深度图像、初始插头深度图像、初始手腕点云、手腕相机外参矩阵、手腕相机深度和RGB图像、Azure Kinect前相机彩色图像、辅助相机左右图像和深度图像等),以及时间戳、帧索引、集索引等元数据。数据集结构详细,适用于机器人学习和控制研究。
This dataset was created using LeRobot and is designed for robotics tasks. It consists of 15 total episodes, 886 total frames, 1 total task, and 60 total videos, with a frame rate of 30 FPS. The data is stored in Parquet format and split into a training set. Features include actions (a 9-dimensional float32 array representing positional and rotational deltas), observations (such as end-effector internal forces, poses, aligned socket depth images, initial plug depth images, initial wrist point clouds, wrist camera extrinsics, wrist camera depth and RGB images, Azure Kinect front camera color images, auxiliary camera left and depth images, etc.), and metadata like timestamps, frame indices, and episode indices. The dataset is structured for use in robotics learning and control applications.




