oxe-auge/fractal20220817_data_train_0_2500_augmented

fractal20220817_data_train_0_2500_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，总计107077帧，帧率为3fps。数据集涵盖了多种机器人（images, jaco, kinova3, kuka_iiwa, panda, sawyer, ur5e, xarm7）的图像、关节位置、末端执行器位置等信息，并提供了自然语言指令。数据以parquet格式存储，每个文件包含256×320×3的图像数据和相关的机器人状态信息。数据集还提供了详细的特征描述，包括图像、关节角度、末端执行器位置等。

fractal20220817_data_train_0_2500_augmented is a large-scale dataset for robot augmentation learning and cross-embodiment policy learning. The dataset contains 2500 episodes, totaling 107077 frames at 3fps. It covers images, joint positions, end-effector positions, and other information for multiple robots (images, jaco, kinova3, kuka_iiwa, panda, sawyer, ur5e, xarm7), along with natural language instructions. The data is stored in parquet format, with each file containing 256×320×3 image data and related robot state information. The dataset also provides detailed feature descriptions, including images, joint angles, end-effector positions, etc.