Abstract
With the advent of the COVID-19 pandemic, face masks have become an integral part in combating the transmission of the virus. As outdoor activity increased due to more relaxed restrictions, people tend to wear face masks as a means of entry to establishments which calls for a proactive approach in regulating violators. In this project, we have created a mobile deep learning model for face mask detection.
It identifies three classes, namely: with mask, no mask, and incorrectly worn. This model would help regulators identify violators and ensure public health proactively. Additionally, the model can be deployed to CCTVs or even phones. Our implementation uses MobileNetV2 to make the training faster. Moreover, our model has a 99.26% test accuracy for our dataset and about 2.3 million parameters, which is lighter than other models in literature.
