Face Mask Detection System using Mobilenetv2

Abstract

In this pandemic, it is getting more and more difficult to keep a track of people who are wearing masks regularly or not. It cannot solely depend on human efforts to take care of this task and therefore there is a need to develop software that can automatically detect whether any given person is wearing a mask or not. Face Detection has evolved as a really popular problem in image processing and computer vision. Many new algorithms are being devised using convolutional architectures to form the algorithm as accurately as possible. These convolutional architectures have made it possible to extract even the pixel details. Training is performed through Fully Convolutional Neural Networks to semantically segment out the faces present in that image. Feature detection and feature extraction techniques help us identify whether a person is wearing a mask or not. The face mask detector will use a dataset of morphed masked images. Therefore, the created model will be accurate and it will also be computationally efficient and easily deployable in embedded systems since the MobileNetV2 architecture will be incorporated (Raspberry Pi, Google Coral, etc.). This framework can also be used in real-time applications that, due to the outbreak of Covid-19, require face-mask detection for safety purposes. This project can be merged with embedded application systems at airports, train stations, workplaces, schools, and public places to ensure compliance with the guidelines for public safety. The above topic is very prominent in recent times as the identification process will not only help us classify individuals but also will reduce the workforce required to do the same exponentially.