oxe-auge/fractal20220817_data_train_12500_15000_augmented

fractal20220817_data_train_12500_15000_augmented是一个机器人增强数据集，基于v3.0版本的代码库，包含2500个集，共107293帧，帧率为3fps。数据集支持多种机器人类型，包括images、jaco、kinova3、kuka_iiwa、panda、sawyer、ur5e和xarm7。数据集提供了丰富的特征，包括各种机器人的图像数据、末端执行器位置、关节位置、自然语言指令等。数据布局采用parquet文件格式存储，视频路径为mp4格式。数据集旨在支持跨实体策略学习的机器人增强任务。

fractal20220817_data_train_12500_15000_augmented is a robotics augmentation dataset based on codebase version v3.0, containing 2,500 episodes with a total of 107,293 frames at 3 fps. The dataset supports multiple robot types, including images, jaco, kinova3, kuka_iiwa, panda, sawyer, ur5e, and xarm7. It provides rich features such as image data for various robots, end-effector positions, joint positions, natural language instructions, etc. The data layout uses parquet file format for storage, with video paths in mp4 format. The dataset is designed to support robotics augmentation tasks for cross-embodiment policy learning.