oxe-auge/fractal20220817_data_train_25000_27500_augmented

该数据集是一个用于机器人学习和控制任务的大规模数据集，包含了多种机器人（如jaco、kinova3、kuka_iiwa、panda、sawyer、ur5e、xarm7）的图像、姿态、关节位置等信息。数据集的特征详细描述了每个机器人的观测数据，包括图像、末端执行器姿态、关节角度等。此外，数据集还包含了自然语言指令，用于描述任务。数据集的版本为v3.0，帧率为3fps，总共有2,500个episodes和107,878帧数据。数据集的分割为train，数据文件以parquet格式存储。

This dataset is a large-scale dataset for robotics learning and control tasks, containing images, poses, joint positions, and other information of various robots (such as jaco, kinova3, kuka_iiwa, panda, sawyer, ur5e, xarm7). The features of the dataset describe in detail the observation data of each robot, including images, end-effector poses, joint angles, etc. In addition, the dataset also includes natural language instructions to describe tasks. The version of the dataset is v3.0, the frame rate is 3fps, and there are a total of 2,500 episodes and 107,878 frames of data. The split of the dataset is train, and the data files are stored in parquet format.