WEARING MASK DETECTION SYSTEM FOR STUDENTS

Abstract

The use of masks in society is still a challenge. According to my observation, In US, for example, there are residents who are reluctant to use masks. Several reasons are the people’s health condition, misinformation and misinterpretation, politics, beliefs, mental health conditions and herd immunity. Another thing that becomes a challenge is that people are getting tired of using masks. Another challenge is the limited authorities’ personnel which resulted the monitoring of masks usage becomes less and less effective. To overcome such problem of ineffective monitoring, this theies proposes a method to detect face mask through image that can be produced by cameras or image files. To detect, I use and compare the method using classification method called as MobileNetV2, ResNet50V2, and Inception-V3 ain deciding whether a face image wears a mask or not. After 20 epochs, the level of training accuracy and validation accuracy reached 99% but when test in special case correct accuracy reached 73.335% In this study, I compared several pre-trained artificial neural methods in comparing the accuracy of each method suitable for detecting the use of mask. I will start with the systematics of writing by introducing the dataset and methods used in the augmentation process to increase the information from the image to be studied. Then, I will discuss about the transfer learning method using the MobileNetV2 algorithm, compared with ResNet50V2, and Inception-V3 to be combined with facial image recognition model called Caffe Model. All in all, I will discuss the results and discuss the performance of each transfer learning method as well as program implementation.