Yisti Vita Via, Eva Yulia Puspaningrum, Dicky Giancini Arwindo Kurniawan, Salamun Rohman Nudin, Nurfiana, Mochammad Syahrul Munir
As part of preventing the rapid spread of the COVID-19 virus, masks should be used besides physical distancing, avoiding crowds, setting good ventilation, cleaning hands, and more. In fact, there are still many people who do not use masks in public places. Security officers usually carry out the supervision system for the use of masks in public places. Direct interaction between security officers and many visitors in the long term will trigger a high risk of being infected. It is better if there is a system that can monitor the use of masks automatically and continuously. In this study, a deep learning algorithm with YOLOv3 architecture is applied to a web-based mask detection system. YOLOv3 is one of the best algorithms that has been widely applied and optimized by researchers for object detection. This system has been tested in two scenarios using different data sets, namely without augmentation and with augmentation. For each scenario, three experiments were carried out using varying amounts of training data, namely 10, 100, and 800. The best mAP and confidence values were obtained from experiments using 800 training data with augmentation. This mAP value is 0.99 with a confidence value range between 0.9 to 1. © 2021 IEEE.
Universitas Pembangunan Nasional Veteran Jawa Timur, Faculty of Computer Science, Department of Informatics, Surabaya, Indonesia; Universitas Negeri Surabaya, Faculty of Engineering, Departement of Informatics, Surabaya, Indonesia; Institut Informatika Dan Bisnis Darmajaya, Faculty of Computer Science, Department of Computer System, Lampung, Indonesia; Akademi Komunitas Negeri Pacitan, Departement of Tata Laksana Studio Produksi, Pacitan, Indonesia
