The Role of Face Embeddings in Classification of Personality Traits from Portrait Images
P. Sreevidya, S. H. Krishna Veni, O.V. Ramana Murthy · TENCON 2021 - 2021 IEEE Region 10 Conference (TENCON) · 2021
There are research works going on to correlate image features with personality traits. This work proposes a deep learning frame work for classifying personality traits from Portrait images. We used a dataset with 30,736 images of hetero-geneous persons for the experimentation purpose. The influence of state-of-the-art face recognition networks were investigated for extracting the facial features. The classification of personality traits was done by applying Support Vector Machine (SVM) classifier based on Big Five Personality model. The loss functions of the selected networks are more discriminative in nature, better than the conventional Mean Square Error (MSE) which is justified through the performance matrices. The proposed method could beat the state-of-the-art results in terms of accuracy and F1-score. We summarize that facial features from Portrait images could very well classify the personality traits, rather than relying on the psychometric tests or analysis of physiological parameters.