Visualization of Facial Attractiveness Factors in Male and Female Images Using Convolutional Neural Network

抄録

Research on attractiveness prediction using facial features has been actively conducted in recent years. The use of convolutional neural networks (CNNs) has been reported to achieve highly accurate predictions. In this study, we investigated facial attractiveness factors by visualizing the hidden layer of the constructed CNN model and enabled the confirmation of features that are important for prediction. The features extracted from this model were moderately consistent with the findings of psychological research.