Performance Evaluation of Masked Face Recognition Using Deep Learning for Covid-19 Standard of Procedure (SOP) Compliance Monitoring

• Published in: 2021 6th IEEE International Conference on Recent Advances and Innovations in Engineering (ICRAIE) Vol. 6; pp. 1 - 7
• Main Authors: Razali, Mohd Norhisham, Shafie, Ana Salwa, Hanapi, Rozita
• Format: Conference Proceeding
• Language: English
• Published: IEEE 01.12.2021
• Subjects: Convolutional neural networks; COVID-19; deep learning; face mask detection; Face recognition; Feature extraction; Performance evaluation; Support vector machines; Technological innovation; transfer learning
• DOI: 10.1109/ICRAIE52900.2021.9703986

Wearing a face mask can reduce the risk of Covid-19 transmission. As a reason, creating an effective masked face recognition model is critical for the development of an autonomous face mask wearing monitoring system. Manual way of face mask wearing monitoring is a tedious task especially in the crowd and large public areas. Furthermore, masked face recognition is complex due to variety of face mask wearing image appearances such as occlusions, calibrations, scene complexity and the types of face mask used. This paper provides the performance evaluation of the Deep Convolutional Neural Network (CNN) model and machine learning classifiers for masked face recognition. Specifically, DENSENET201, NASNETLARGE, INCEPTIONRESNETV2 and EFFICIENTNET (EFFNET) as a feature extractor. Then, the extracted features are classified by using Support Vector Machine (SVM), Linear Support Vector Machine (LSVM), Decision Tree (DT), K-nearest neighbour (KNN) and Convolutional Neural Network (CNN). The recognition model is evaluated on the face mask detection dataset. The experiment results have shown that DENSENET201-SVM and EFFNET-LSVM obtained the best classification accuracy of 0.9972. However, EFFNET-LSVM has the advantage of better computational time of feature extraction, classification as well as the size of features.