oxe-auge/fractal20220817_data_train_82500_85000_augmented

该数据集名为fractal20220817_data_train_82500_85000_augmented，是一个用于机器人学习和增强学习的大规模数据集。数据集包含多种机器人（jaco, kinova3, kuka_iiwa, panda, sawyer, ur5e, xarm7）的图像和状态数据，包括原始图像和增强后的图像。数据集的帧率为3fps，总共有2500个episodes和107440帧。数据以parquet格式存储，包含多种特征如机器人图像、关节位置、末端执行器位置、自然语言指令等。数据集基于OXE-AugE项目，旨在通过数据增强技术提升跨机器人策略学习的效果。

The dataset is named fractal20220817_data_train_82500_85000_augmented and is a large-scale dataset for robotics and reinforcement learning. It contains image and state data for multiple robots (jaco, kinova3, kuka_iiwa, panda, sawyer, ur5e, xarm7), including both original and augmented images. The dataset has a frame rate of 3fps, with a total of 2,500 episodes and 107,440 frames. The data is stored in parquet format and includes various features such as robot images, joint positions, end-effector poses, natural language instructions, etc. The dataset is part of the OXE-AugE project, which aims to improve cross-embodiment policy learning through data augmentation techniques.