A comparative study of classifiers used in facial embedding classification

Face recognition, recently, has been a fast and effective method of authentication with the advent of deep learning and powerful hardware. This paper investigates different classifiers used in classifying facial embeddings and evaluates their performance. The paper also focuses on an easily deployable pipeline for face recognition using Python which can be used to develop a face recognition system on portable low-power hardware devices. The methodology discussed uses pretrained models and frameworks which results in state-of-the-art performance without the need of any powerful hardware. The proposed methodology achieves an F1 score of 0. 9947with an AUC score of 0.9997 on LFW dataset.