AI-based Detection of Mask-Wearing Compliance Using Deep Neural Networks

The previous COVID-19 pandemic has underscored the critical importance of public health compliance, particularly the correct wearing of face masks in public environments, in order to mitigate airborne disease transmission. This research presents an artificial intelligence approach for the detection of correct mask-wearing versus incorrect mask-wearing. In this research, a deep neural network model, namely AlexNet, is trained using a transfer learning approach and fine-tuned on the large MaskedFace-Net dataset found on Github, comprising 137,016 images. In order to ensure robust and reliable performance evaluation of the model, a data partitioning scheme of 80% training, 10% validation, and 10% testing, as well as data augmentation techniques are employed in this research. The fine-tuned AlexNet model achieves an accuracy of 98.2% on an unseen test set of 13,701 images. These results, which are obtained from a large dataset, demonstrate the model’s robustness and its potential for deployment in public health enforcement systems.