oxe-aug/kaist_nonprehensile_augmented

kaist_nonprehensile_augmented数据集是一个机器人增强数据集，包含多种机器人（如google_robot、jaco、kinova3等）的增强图像和相关信息。数据集总共包含201个剧集，32429帧图像，分为训练集。每个剧集包含机器人的关节位置、末端执行器位置、姿态误差等信息。该数据集用于扩展跨机器学习策略学习，支持模拟到现实的零样本迁移。

The kaist_nonprehensile_augmented dataset is a robot augmentation dataset containing augmented images and related information for various robots (such as google_robot, jaco, kinova3, etc.). The dataset includes a total of 201 episodes, 32,429 frames of images, split into a training set. Each episode contains information such as the robots joint positions, end-effector positions, pose errors, etc. This dataset is used to extend cross-embodiment policy learning, supporting zero-shot sim-to-real transfer.