Personalized facial beauty assessment: a meta-learning approach

Automatic facial beauty assessment has recently attracted a growing interest and achieved impressive results. However, despite the obvious subjectivity of beauty perception, most studies are addressed to predict generic or universal beauty and only few works investigate an individual's preferences in facial attractiveness. Unlike universal beauty assessment, an effective personalized method is required to produce a reasonable accuracy on a small amount of training images as the number of annotated samples from an individual is limited in real-world applications. In this work, a novel personalized facial beauty assessment approach based on meta-learning is introduced. First of all, beauty preferences shared by an extensive number of individuals are learnt during meta-training. Then, the model is adapted to a new individual with a few rated image samples in the meta-testing phase. The experiments are conducted on a facial beauty dataset that includes faces of various ethnic, gender, age groups and rated by hundreds of volunteers with different social and cultural backgrounds. The results demonstrate that the proposed method is capable of effectively learning personal beauty preferences from a limited number of annotated images and outperforms the facial beauty prediction state-of-the-art on quantitative comparisons.