MTGS jointly models gaze following and social gaze prediction in multi-person scenarios. It leverages temporal information and interactions among individuals to enhance gaze prediction accuracy. - View it on GitHub