Leveraging the rich semantic cues from language models to enhance the performance of appearance-based gaze estimation.
