Personality prediction from Five-Factor Facial Traits using Deep learning

Abstract

In this paper, a fine-tuned deep learning model is presented for the prediction of personality using facial images. It can measure personality qualities from a portrait photograph using the Five-Factor model (Big Five). To assess the efficacy of this method, a fresh corpus of 30,935 portraits with their associated personality characteristic was retrieved from an existing video resource (First Impressions, ChaLearn) and labelled with redundant pairwise comparisons to assure consistency. For each Big Five feature, the provided model will categorize these qualities into a binary class: openness to experience (O), conscientiousness (C), extraversion (E), agreeableness (A), and neuroticism (N). The experiment was conducted on three baseline models and achieved an average accuracy of 81% which shows improvement of 3% over existing state-of-art models.

URN:NBN:sciencein.jist.2023.v11.578