Optimization of Faster R-CNN to Detect SNI Masks at Mandatory Mask Doors

Abstract

In preparing for the COVID-19 endemic period, every government and private agency will be required to comply with new rules by the government, where everyone is required to wear a mask and perform physical distancing when leaving the house for activities. This is one of the new habits that will be familiarized to the public by the government in 2022 and the following year. Due to the easy transmission of the Covid-19 virus, the selection of SNI masks is highly recommended. The purpose of this study was to classify the types of SNI and non-SNI masks so that the presence of this SNI mask cluster monitoring system could increase security in locations that require the mandatory use of masks such as in schools and the masks used can function effectively to prevent the spread and transmission of Covid -19.. From the dataset, 100 pictures were taken randomly on SNI and non-SNI mask users. Where the types of masks that are SNI have been studied on Duckbil masks, KN-45, Surgical Mask 2 ply, and Surgical 3 ply are included in SNI masks and cloth masks are not included in SNI masks at a distance of 0.5 meters, 1 meter, 1.5 meters and 2 meters from the mobile device. It can be seen on the SNI mask at a distance of 0.5 meters achieving 100% accuracy, a distance of 1 meter achieving 100% accuracy, at a distance of 1.5 meters achieving an accuracy of 95% and at a distance of 2 meters achieving 90% accuracy. Whereas non-SNI masks at a distance of 0.5 meters achieve 100% accuracy, a distance of 1 meter achieves 100% accuracy, at a distance of 1.5 meters achieves an accuracy of 95% and at a distance of 2 meters achieves an accuracy of 90%.