Deep Learning Based Face Recognition System with Smart Glasses

Abstract: Individuals with prosopagnosia have difficulty in identifying different people by their faces. Our goal is to design and develop a face recognition system with wearable glasses to recognize faces and provide identity information to users. Unlike other existing systems that run locally on glasses or cellphones, we introduce a client-server architecture system for facial identification. We designed and implemented applications both on a pair of smart glasses and a cellphone to capture images and communicate with the server. Deep Convolutional Neural Networks (CNN) were chosen to build our face recognition on the back-end system and we achieved 98.18% accuracy for face recognition. The system is designed to handle new identities and new faces without having to rebuild the model.

Full paper or download from ACM Digital Library