Gaze estimation methods based on deep learning: research status, challenge and future

Gaze estimation has been gaining countless currency in recent years, however, that is still confronted with myriad challenges. In the work, we present a brief overview of gaze-based applications, related methods and several dominant plight. First and foremost, we retrospect about the application on basis of gaze. Additionally, we classify the related method into three types: 3D Geometric model, monocular or binocular estimation and deep learning model. Subsequently, we summarize the main challenge, which is still a tough nut for researches to crack. Ultimately, we discuss about the research status and dilemmas for future researchers.