iPersonalityReader: An Application to Assess the Personality of a Person via Face Image using Pre-trained Convolutional Neural Networks

Abstract

The iPersonalityReader is a web-based app which is capable to assess the personality from the outer appearance of a person. To assess the personality, we proposed a convolutional neural network-based model with transfer learning. VGG-16 pre-trained deep learning model is used as a feature extractor to extract facial features. The extracted features are fed into six parallel “ReLu” layers to predict “Big-Five” personality values including “interview” factor. The proposed model achieved a mean squared error which is below 0.03 for all the personality traits including the “interview” factor.