Automated Face Mask Detection Using Dcnn and Mobilenet V3 for Covid19 Prevention

The rapid dissemination of the 2019 novel coronavirus (SARS-CoV-2) led to increased public and personal emphasis on such preventive measures, including face mask use. Even though face mask detection has been studied extensively, most previous works were limited to binary classification (wearing mask vs. No mask) and do not notice how masks are being improperly used. To bridge these gaps, in this study, we proposed a deep learning framework: the deep convolution neural network (CNN) combined with MobileNet V3 improved by Squeeze and Excitation blocks (SE-Mobilenet V3). The present work is new in two respects; first, it presents a multi-class detection framework with the ability to report correct as well as incorrect and out of vocabulary mark usage. second, it adopts integration of SE block to MobileNet V3 which enforces the feature representation strength and improves model robustness tremendously. Experiments conducted on a publicly available Kaggle dataset demonstrate that MobileNet V3 achieves 99% accuracy, surpassing the DCNN model’s 96%. Our proposed system implemented using convolution neural network which uses Keras,TensorFlow, and Scikit-learn to improve the accuracy of the algorithm. Beyond COVID-19, the proposed approach has significant relevance for public health monitoring, workplace safety compliance, and smart surveillance applications where adherence to protective equipment remains essential.