Face Mask Detection Using Convolutional Neural Network

COVID-19 virus has affected people around the world, and it is known for its potential deadly symptoms if not treated or prevented using vaccines. This virus spreads very fast and easily as it can be airborne. One of the practical methods and solutions to stop the virus from spreading is simply using facemask. However, many people tend to wear a face mask incorrectly or in a worst-case scenario they do not wear it at all. Our work proposes a way to implement a surveillance like system to identify and recognize people wearing face masks. The development of this surveillance system is in the hope that it can be used in premises so it can help the authorities identify people who is wearing facemask incorrectly. The system can detect a face mask by implementing a CNN classification algorithm. This study focuses two CNN models with different optimizers trained with an image dataset that is labelled in two different class labels where one classes identifies as people who wear facemasks correctly and the other one identifies it as people who wears facemasks incorrectly. Furthermore, parameter tuning is implemented to increase the accuracy of the CNN model. Finally, based on the tuned model’s performance, the model with the highest accuracy was selected to be used in the surveillance system. In this study, the highest accuracy is produced by the CNN model with an Adamax optimizer that achieves 99% training accuracy.