masked face identification using the convolutional neural network method

Abstract

            In times of a pandemic like this, masks are part of the main needs in daily activities when outside the home. Because masks can help us avoid the Covid-19 virus, it often happens among people when doing activities outside the home that they forget to wear masks, therefore the level of public awareness of the importance of wearing masks is decreasing.  This study aims to create a system that can classify people who wear masks and do not wear masks as an evaluation material for the level of public awareness of the importance of wearing masks. The total data used in this study were 600 data samples which were divided into two, namely 300 data samples wearing masks and 300 data samples not wearing masks. The CNN architecture in this study is the same as the CNN architecture in general, the  difference is the depth level of the convolution layer and pooling which consists of 5  convolution layers, 5 max pooling, and finally, 2 layers dense In the training process, it gets  the highest accuracy rate of 98%, while in the validation process it gets the highest level of  accuracy at 95%. Therefore, the results of these two processes show that the application of deep earning by utilizing the convolutional neural network can classify objects that wear masks and do not wear masks properly. The results of testing the research dataset are quite maximal by using 40 new dataset testing data to test the convolutional neural network that has been created by the researchers to get an overall accuracy result of 97.5%.