Within the Protection of COVID-19 Spreading: A Face Mask Detection Model Based on the Neutrosophic RGB with Deep Transfer Learning

Abstract

COVID-19's fast spread in 2020 compelled the World Health Organization (WHO) to declare COVID-19 a worldwide pandemic. According to the WHO, one of the preventative countermeasures against this type of virus is to use face masks in public places. This paper proposes a face mask detection model by extracting features based on the neutrosophic RGB with deep transfer learning. The suggested model is divided into three steps, the first step is the conversion to the neutrosophic RGB domain. This work is considered one of the first trails of applying neutrosophic RGB conversion to image domain, as it was commonly used in the conversion of grayscale images. The second step is the feature extraction using Alexnet, which has been small number of layers. The detection model is created in the third step using two traditional machine learning algorithms: decision trees classifier and Support Vector Machine (SVM). The Simulated Masked Face dataset (SMF) and the Real-World Mask Face dataset (RMF) are merged to a single dataset with two categories (a face with a mask, and a face without a mask). According to the experimental results, the SVM classifier with the True (T) neutrosophic domain achieved the highest testing accuracy with 98.37%.

Keywords

• neutrosophic rgb
• covid-19
• classical machine learning
• deep learning
• face mask detection