oxe-auge/fractal20220817_data_train_40000_42500_augmented

fractal20220817_data_train_40000_42500_augmented是一个用于机器人学习和跨实体策略学习的大规模机器人增强数据集。数据集包含了多种机器人的图像、关节位置、末端执行器位置等信息，并且提供了自然语言指令。数据集的特征包括原始机器人和增强机器人的图像、关节角度、末端执行器位置和误差等。数据集的结构为Parquet格式，分割为训练集，包含2500个episodes和108277帧。数据集的引用政策要求同时引用原始数据集和本数据集。

fractal20220817_data_train_40000_42500_augmented is a large-scale robot augmentation dataset for robotics learning and cross-embodiment policy learning. The dataset includes images, joint positions, end-effector positions, and natural language instructions for multiple robots. The features of the dataset include images of the original and augmented robots, joint angles, end-effector positions, and errors. The dataset is structured in Parquet format, split into training sets, containing 2,500 episodes and 108,277 frames. The citation policy requires citing both the original dataset and this dataset.