Performance Evaluation of Machine Learning Classifiers for Face Recognition

Abstract

The digital world, especially image processing, has been evolving due to the needs of society and the importance of digital-based system security. One of the rapidly progressing technologies is the face recognition system using artificial intelligence. It recognizes a person’s face registered in the database for verification purposes. In this study, we evaluate the face recognition systems based on machine learning classifier algorithms and Principal Component Analysis (PCA) for feature extraction. Seven machine learning algorithms were considered, i.e., Support Vector Machine (SVM), Decision Tree, K-Nearest Neighbour (KNN), Logistic Regression, Naïve Bayes, Multi-Layer Perceptron (MLP), and Convolutional Neural network (CNN). In the CNN scenario, PCA was not used since it has its feature extraction capability. The first six classifiers were set to the most optimal settings. At the same time, CNN used the LeNet-5 architecture trained with a dropout rate of 0.25, 60 epochs, batch size of 20, Adam optimizer, and cross-categorical entropy for the loss function. The input image size was 64×64×1 obtained from the Olivetti faces database. CNN, SVM, and LR achieved the three highest accuracies, i.e., 98.75%, 98.75%, and 97.50%, respectively.