Indexing facial attractiveness and well beings using machine learning

A challenge is indexing the facial beauty by a machine as same evaluated by human beings. A question arises: Can beauty be learnt by machines? Every individual have different concept of facial beauty. Somebody can be attracted by someone but might not be by another person. In recent past, many psychologists, neurologists and other scientists have done tremendous work in this area. This work presents a study on the facial attractiveness in a machine learning context. Various techniques applied on SCUT-FBP facial images dataset for learning the facial attractiveness. From the results, we showed that facial beauty is a universal concept that a machine can learn. A model is designed that learns from the facial images with their attractiveness ratings and produced human like evaluation of attractiveness ratings. We extracted features from images and normalized all the features value. After that we proposed some techniques of machine learning like Support Vector Machine (SVM), k-Nearest Neighbor (KNN), Decision tree and Artificial Neural Network (ANN). An accuracy of 80% was obtained using KNN and 86% was obtained using ANN for multiclass classification. Accuracies of 61% and 62% were reported by SVM using linear kernel and RBF kernel respectively. These accuracies were obtained for multiclass classification. We also evaluated accuracy for binary class to reduce the effect of non-uniformity of data.