Affective Evaluation and Face Detection-Based Matching System for Personalized Filters in Photo Booths

With the rapid growth in demand for personalized image creation and decoration in the photo booth market, current systems often apply uniform filters that overlook individual characteristics such as gender, age, and skin tone, resulting in mismatched or less satisfying user experiences. This study proposes a personalized filter recommendation system that integrates affective evaluation and facial attribute recognition. Using the Semantic Differential method, 160 participants evaluated 18 filters across eight user profiles defined by gender, age, and skin tone. Principal Component Analysis revealed clear affective dimensions aligned with demographic traits, enabling filter suggestions tailored to each user group. A YOLOv8 model trained on the FairFace dataset achieved 0.85 mAP and 30 FPS, supporting real-time classification of user attributes. The combined system enhances both personalization and emotional satisfaction in photo-taking experiences. Future work will explore a broader range of filters and incorporate more photographic variables to further refine recommendations.