A Deep Learning-Powered Visual Surveillance Model for Mask Usage Governance
The use of face masks is relevant in the prevention against the spread of airborne diseases, particularly in the crowded social setting. The process of automating the identification of an adequate mask usage is not an easy one because of the differences in face coverings, lighting, and face positions.This work proposes a Face Mask Detection System that uses deep learning to automatically detect whether people are wearing masks properly, improperly, or not. The system identifies three types of faces, namely, Mask, No Mask, and Improper Mask, through visual indications of facial parts. This is done through a combined dataset of masked and unmasked face images that are used to train the model and allow the model to learn distinguishing features of various mask-wearing patterns. Face detection, normalization, and data augmentation are preprocessing methods, which enhance resistance to environmental changes. The model is combined with a real-time detection pipeline that uses a webcam feed and thus compliance can be checked instantly. This technology can assist in enforcing safety in the community, workplaces, transport centers, health centers and schools. Finally, this project will help to create intelligent surveillance systems that will aid in maintaining the health and safety of citizens by means of automated monitoring of the mask compliance.