Face Mask Wearing Detection Using Support Vector Machine (SVM)

Abstract

As an effort to prevent the spread of the Covid-19, various countries have implemented health protocol policies such as work-from-home, social distancing, and face mask-wearing in public places. However, monitoring compliance with the policy is still difficult, especially for the face mask policy. It is still managed by humans and is costly. Thus, this research proposes a face mask-wearing detection using a soft-margin Support Vector Machine (SVM). There are three main stages: feature selection and preprocessing, model training, and evaluation. During the first stage, the dataset of 3833 images (1915 images with face masks and 1918 images without face masks) was prepared to be used in the training stage. The training stage was conducted using SVM added with the soft-margin objective to overcome images that could not be separated linearly. At the final stage, evaluation was conducted using a confusion matrix with 10 folds cross-validation. Based on the experiments, the proposed method shows a performance accuracy of 91.7%, a precision of 90.3%, recall of 93.5%, and an F-measure of 91.8%. Our method also worked fast, taking only 0.025 seconds to process a new image. It is 7.12 times faster than Deep Learning which requires 0.18 seconds for one classification.