FACEMASK DETECTION USING YOLO V5

Abstract

The use of facemasks is one of the obligations when carrying out activities outside the home during the COVID-19 pandemic, but despite the COVID-19 pandemic, the use facemasks is still needed. One of the supporting factors driving this is air pollution. The use of facemasks can reduce the risk of respiratory diseases, because it is important to use facemasks when carrying out activities in place with a high risk of air pollution such as industrial areas, this is done to maintain the safety of its users both in term of healh and comfort. So consistency is needed for users to use masks, through current technological developments detecting the use of masks is one of the right solutions to this problem. One of the mask detection methods used in this study is YOLO (You Only Look Once). YOLO is a method that detects objects using a single neural network consisting of several layers of convolution networks for image feature extraction, then prediction of bounding box coordinates is performed simultaneously. The YOLO v5 training model in this study was carried out with a combination of minimum values ​​on img, batch, and epoch resulting in a maximum F1 value and mAP@50 of 86%.