Vision Transformer for Facial Expression-based PHQ-8/9 Regression
Introduction Predicting mental health scores like PHQ-8/9 from facial expressions is a challenging task that combines computer vision and affective computing. PHQ-9 (Patient Health Questionnaire-9) is a 9-item clinical survey for depression severity assessment Predicting mental health scores like PHQ-8/9 from facial expressions is a challenging task that combines computer vision and affective computing. PHQ-9 (Patient Health Questionnaire-9) is a 9-item clinical survey for depression severity assessment (PHQ-8 is similar but without the ninth item). Our goal is to build a Vision Transformer (ViT)-based model that estimates a person’s PHQ-8/9 score from their facial expression. To tackle data scarcity in clinical PHQ-labeled datasets, we adopt a multi-stage training strategy: ...