Thor-Magni dataset

该数据集名为Thor-Magni，包含了五种不同的人机交互（HRI）场景，参与人数从四人到九人不等。这些场景中的人类轨迹是通过高精度运动捕捉系统记录的。数据集总共包含了41,675个轨迹片段，每个片段包含16个点。其中70%的数据用于训练，30%用于验证。该数据集被用于训练社交神经网络，其准确率达到了89.69%。它涵盖了多个参与者的HRI场景，任务的目的是机器人社交导航。

The dataset is named Thor-Magni, which includes five distinct human-robot interaction (HRI) scenarios with participant counts ranging from 4 to 9. Human trajectories in these scenarios were recorded using a high-precision motion capture system. In total, the dataset contains 41,675 trajectory segments, each consisting of 16 points. 70% of the data is allocated for training, while 30% is reserved for validation. This dataset has been utilized for training social neural networks, achieving an accuracy of 89.69%. It covers multi-participant HRI scenarios targeting robotic social navigation tasks.