Deep Learning Method for Multi-Class Face-Mask Classification in Real-Time

The paper presents a real-time face mask detection system employing MobileNet through transfer learning. The main novelty of the paper is to even detect faces with improperly worn masks. In the context of the COVID-19 pandemic, the importance of adhering to safety measures, such as wearing face masks has become important and hence, this research provides a valuable contribution to ensuring public health. The approach combines the power of RetinaNet used for face detection and localization with MobileNet for efficient and accurate mask-wearing-style classification. The proposed system showcases promising results in terms of speed and accuracy, demonstrating its potential for practical deployment in various settings wherein face-mask compliance is crucial, such as airports, healthcare facilities, and public transportation hubs. This innovative approach to real-time face mask detection through the fusion of RetinaNet and MobileNet opens new avenues for efficient and effective monitoring of safety protocols. This work achieves multi-class classification of images into three classes: with-mask, without-mask, and incorrectly-worn mask with an accuracy of 85%.