Open Access
ARTICLE
An Enhanced Deep Learning Model for Automatic Face Mask Detection
Qazi Mudassar Ilyas1, Muneer Ahmad2,*
1 Department of Information Systems, College of Computer Sciences and Information Technology, King Faisal University, Al-Ahsa, 31982, Saudi Arabia
2 Department of Information Systems, Faculty of Computer Science & Information Technology, Universiti Malaya, 50603, Kuala Lumpur, Malaysia
* Corresponding Author: Muneer Ahmad. Email:
Intelligent Automation & Soft Computing 2022, 31(1), 241-254. https://doi.org/10.32604/iasc.2022.018042
Received 22 February 2021; Accepted 09 May 2021; Issue published 03 September 2021
Abstract
The recent COVID-19 pandemic has had lasting and severe impacts on social gatherings and interaction among people. Local administrative bodies enforce standard operating procedures (SOPs) to combat the spread of COVID-19, with mandatory precautionary measures including use of face masks at social assembly points. In addition, the World Health Organization (WHO) strongly recommends people wear a face mask as a shield against the virus. The manual inspection of a large number of people for face mask enforcement is a challenge for law enforcement agencies. This work proposes an automatic face mask detection solution using an enhanced lightweight deep learning model. A surveillance camera is installed in a public place to detect the faces of people. We use MobileNetV2 as a lightweight feature extraction module since the current convolution neural network (CNN) architecture contains almost 62,378,344 parameters with 729 million floating operations (FLOPs) in the classification of a single object, and thus is computationally complex and unable to process a large number of face images in real time. The proposed model outperforms existing models on larger datasets of face images for automatic detection of face masks. This research implements a variety of classifiers for face mask detection: the random forest, logistic regression, K-nearest neighbor, neural network, support vector machine, and AdaBoost. Since MobileNetV2 is the lightest model, it is a realistic choice for real-time applications requiring inexpensive computation when processing large amounts of data.
Keywords
Cite This Article
Q. Mudassar Ilyas and M. Ahmad, "An enhanced deep learning model for automatic face mask detection,"
Intelligent Automation & Soft Computing, vol. 31, no.1, pp. 241–254, 2022. https://doi.org/10.32604/iasc.2022.018042