Automated Face Mask Compliance Detection Using A Hybrid Efficientnet And Vision Transformer Architecture

Abstract

Face masks are still a way to stop the spread of diseases that are in the air when we are in public. We can use machines to check if people are wearing face masks and this can help us keep an eye on things without needing someone to always be watching. This research is about a way of using computers to detect if people are wearing face masks correctly. We use two types of computer models called EfficientNet and Vision Transformer to look at pictures of faces and figure out if someone is wearing a mask or not. EfficientNet is good at looking at the details of a face and Vision Transformer is good at understanding the picture and how things are related. We combine the information from both models.Then use it to decide if someone is wearing a mask not wearing a mask or wearing a mask incorrectly. This new way of doing things is meant to work even when it is hard to see like when the light is not good or when someones face is turned away. We tried this way and it worked better than other ways that use computers to look at pictures. This means we can use it in life to help keep people safe in places like hospitals, schools and, on public transportation.