Face Mask Recognition with Realistic Fabric Face Mask Data Set: A Combination Using Surface Curvature and GLCM

Lionnie, R.; Apriono, C.; Gunawan, D.

Wearing a mask is a requirement in the Covid-19 pandemic for the general public. While it is one of the several must-do actions to prevent forward spread in the Covid-19 infections, at the same time, the effect of wearing a mask in naive face recognition systems have shown lower system performance in several cases and conditions. Simultaneously, only a handful of research studies have focused on a non-medical face mask with realistic images data set. This research proposed a new data set of realistic fabric face mask data set to be evaluated using surface curvature and gray level co-occurrence matrix (GLCM). The classification applied support vector machine (SVM). One hundred seventy-six images in the data set were analyzed with various properties, resulting in several experiments. The experiments' parameters were color properties, approaches in surface curvature, i.e., Gaussian, mean and principal curvature, angle and distance in GLCM, GLCM properties, i.e., contrast, homogeneity, correlation and energy, also kernel functions in SVM. The best accuracy result, 87.5%, was derived from the combinations of these parameters. This research also improved the running time of the recognition process while maintaining the system's performance.