MaskNet: CNN for Real-Time Face Mask Detection Based on Deep Learning Techniques

Nasri, I.; Karrouchi, M.; Snoussi, H.; Messaoudi, A.; Kassmi, K.

Coronavirus disease 2019 (COVID-19) is currently spreading in several countries around the world. The wearing of face masks during the COVID-19 pandemic is one of major protection method that has received varying recommendations from different public health agencies and governments. In this paper we have proposed a face mask detector based on deep learning techniques to classify each face as with mask or without mask in real-time. This study will focus on the Prajna Bhandary dataset. In this work a comparison between transfer learning and training from scratch has been provided, this comparison based on accuracy, model size and computer time. We have also analyzed the role of number of images dataset and learning rate in the classification accuracy and training time. In additions, the extracted model can achieve an accuracy of more than 98% and can be implemented in embedded system and used for many applications. © 2021, The Author(s), under exclusive license to Springer Nature Switzerland AG.