BUILDING A SMART HUMAN RESOURCE MANAGEMENT SYSTEM - FACE MASKED AND ANTI SPOOFING DETECTION MODULE

Abstract

This study shows how to use Transfer Learning to detect persons wearing masks in photos and real-time detection by using OpenCV’s Deep Neural Network module. A MobileNetV2 network was used to train a binary recognition task (people with masks or without masks, image spoof or image real). The photos used as input were 224x224 pixels in size. A total of 6.591 photographs for the face masked model and 14.730 pictures for the anti-spoofing model were acquired to train and test the network. In addition, we developed a web application and API server for data augmentation and face detection to aid in the development of facial recognition and detection systems. The detection accuracy of two models were over 93%, indicating that this could be one of the methods used to restrict COVID-19 disease transmission to a bare minimum and protect public health.