Face Recognition Using Transfer Learning on Facenet: Application to Banking Operations

Abstract

Of late, biometric systems are ubiquitous to provide user authentication and security. With the advent of AI aided computer vision, face recognition systems are the new age biometrics gaining attention. They are more robust than fingerprint scanners and provide contactless experience to users. In this paper, we discussed the applications of face recognition and face verification techniques for banking and finance services. We propose an efficient and better approach to train a face recognition model which has potential applications in banking operations among other domains. We performed face detection with Histograms of Oriented Gradients (HOG) face detection. Our approach involves transfer learning on the state-of-the-art face recognition model Facenet to extract face embeddings and a kind of Nearest Neighbors (NN) to label the face. Our approach doesn’t involve large datasets and powerful GPU computations to train the model. We performed our experimentation on the Georgia Tech Face-Database (GTFD) and achieved an accuracy of 96.67%, which is very close to human vision (97.53%) and a significant improvement over other approaches.