Deep COVID-19 Face Mask Detection in Public Using Transfer Learning
Ravneet Kaur, Chaitanya Singla, Rishu Chhabra, Janpreet Singh · 2022 10th International Conference on Reliability, Infocom Technologies and Optimization (Trends and Future Directions) (ICRITO) · 2022
Ahstract- The face mask is necessary for crowded places to control the pernicious effect of Corona Virus (COVID-19). The government officials of various countries have mandated the usage of face masks in public places. However, inspecting unmasked people in crowded areas is very tough. To solve this issue, the research demonstrates the automatic detection of masked faces from the images using transfer learning. In the proposed works, the pre-trained models ResNet34 and ResNet50 have been used on the MAFA data set to analyze the accuracy of face mask detection. Experimental testing evaluated 91.74% accuracy for ResNet34 whereas ResNet50 outperformed and achieved 92.3% accuracy. However, the training loss is found to be minimum in Resnet50 as compared to Resnet34.