oxe-auge/fractal20220817_data_train_77500_80000_augmented

fractal20220817_data_train_77500_80000_augmented是一个用于机器人增强学习和跨实体策略学习的大规模数据集。数据集包含2500个episodes，共109018帧，帧率为3fps。数据集支持多种机器人（images, jaco, kinova3, kuka_iiwa, panda, sawyer, ur5e, xarm7），并提供了丰富的特征，包括机器人图像、末端执行器姿态、关节位置、自然语言指令等。数据以parquet格式存储，结构清晰。数据集基于OXE-AugE项目，旨在促进机器人策略学习的跨实体泛化能力。

fractal20220817_data_train_77500_80000_augmented is a large-scale dataset for robotic reinforcement learning and cross-embodiment policy learning. The dataset contains 2,500 episodes with a total of 109,018 frames at 3 fps. It supports multiple robots (images, jaco, kinova3, kuka_iiwa, panda, sawyer, ur5e, xarm7) and provides rich features including robot images, end-effector poses, joint positions, natural language instructions, etc. The data is stored in parquet format with a clear structure. The dataset is part of the OXE-AugE project, aiming to enhance cross-embodiment generalization in robotic policy learning.