CNN-based Person Recognition System for Masked Faces in a post-pandemic world

摘要

With everyone now covering their faces with a face mask to avoid contagion of the COVID-19 virus, it becomes an increasingly difficult challenge for the facial recognition systems to identify people wearing masks. Algorithms devised before the pandemic for facial recognition often fail in this context, and consequently a need to understand the working of facial recognition algorithms when presented with occluded faces arises. This project aims to develop a new lightweight Convolutional Neural Network-based algorithm to resolve this issue. The proposed model gives a comparable accuracy with similar models developed in the past. Further, the proposed algorithm is used to create a robust system to ensure adherence of COVID-19 protocols in a real-world environment.