Human2Sim2Robot Demonstration Dataset

该数据集是从一段单一的人体展示灵巧操作任务的RGB-D视频中提取的，用于训练机器人的强化学习策略。该数据集包含了包括零食盒、水壶和盘子在内的多种物体，使得在无需大规模数据收集的情况下，也能促进强化学习策略的训练。这些任务涵盖了灵巧操作，包括抓取、非抓握操作以及多步骤任务。

This dataset is extracted from a single RGB-D video demonstrating humans performing dexterous manipulation tasks, and is used to train reinforcement learning policies for robots. It includes various objects such as snack boxes, kettles and plates, which facilitates the training of reinforcement learning policies without the need for large-scale data collection. These tasks cover dexterous manipulation, including grasping, non-prehensile manipulation and multi-step tasks.