Discriminative Representation Learning for Face Recognition

A significant progress has been made on face recognition in recent years because of the rapid development of deep learning approaches. Deep learning approaches offer a powerful toolbox for tackling many aspects of face recognition, including the search for effective discriminative features. We compare several state-of-the-art face recognition methods and combine different modules from those methods to propose a special approach for discriminative representation learning. Our approach has a special architecture for representation learning combined with a latest design of classification loss function, making it a highly effective solution for uncontrolled face recognition. Experimented on the Labeled Faces in the Wild (LFW), the Celebrities in Frontal-Profile dataset (CFP), and the AgeDB datasets, our approach shows competitive performance to other state-of-the-art methods.