Broad Learning and Hybrid Transfer Learning System for Face Mask Detection

Abstract

The widespread spread of the Coronavirus COVID-19 has caused a worldwide healthcare problem. This virus spreads mostly through the droplets of infected people, posing a risk to others. Although the fight against the virus has become necessary, the risk of transmission can be reduced by using face masks. In this study,  hybrid machine learning techniques is used for designing a two-stage methodology for detecting face masks on humans. The first stage is for detection where Faster RCNN and MobileNetV2 architecture transfer model with OpenCv, Keras, and Tensorflow dependencies, while the second stage is for verification of the detected results using a Broad Learning System. A frame is displayed over the individual's face, indicating whether or not the user is wearing a mask. The bounding box outline colour is set to green if the detected user is wearing a mask. The bounding box outline colour changes to red if the detected user is not wearing a mask.