Machine Learning Enabled Hairstyle Recommender System Using Multilayer Perceptron

One of the most distinctive elements that enhances a woman’s facial features, in terms of esthetic assessments, is her hair. According to beauty experts, the haircut or hairstyle accounts for 70% of the total appearance of the face. Yet one of the most time-consuming choices a woman ever has to make is choosing the appropriate hairstyle or hairdo. The initiation of this work is by an approach that incorporates face recognition and landmark detection to classify face shapes into 5 different shapes: heart, long, oval, round, square. Five machine learning models, namely K-Nearest Neighbors (KNN), Random Forest classifier (RF), Gradient Boosting (GB), Linear Discriminant Analysis (LDA) and Multilayer Perceptron (MLP) classifier have been implemented in this work in order to accurately classify the user’s face into one of the five aforementioned shapes. MLP classifier yielded the highest accuracy of 88%. Furthermore, the hairstyle recommendation software is an adaptive software that evolves over time based on user feedback.