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.