Research Article | Open Access
FACEMASK DETECTION USING CONVOLUTIONAL NEURAL NETWORKS (CNN) WITH MOBILENETV2 ARCHITECTURE
Dr. Ajith Jubilson E , Ch Madhurya
Pages: 2173-2183
Abstract
Coronavirus Disease (COVID-19), which flared up at the end of 2019, continued to be a source
of concern for all organizations and people till date. According to World Health Organization (WHO), around
228 billion people have been affected and seen 4 billion deaths globally. Even after considerable vaccination
worldwide, there is still a wide spread of the virus. WHO stated that, the best way to protect ourselves is to
avoid exposure to the virus, and also advised proper use of facemasks, i.e., covering the mouth and nose with a
mask in public areas and continuous sanitization will avoid exposure to the virus. Implementing the "no maskno service" method has been recommended by the governments of all countries. Even then, the goal was
obstructed by wearing a mask improperly, which led to the spread of the virus. For classification of images, a
technique in deep learning called Convolutional Neural Network (CNN) is used. CNN is a part of Artificial
Neural Network (ANN), which has an ability in recognizing and classifying the images by processing images in
video streams. An image classifier is created for identifying masked and unmasked faces from video streams.
The proposed system was carried out by making use of Python, OpenCV, and TensorFlow, which can recognize
whether a person is with proper facemask or with no facemask from live video streams. 2000 images are
collected from two different datasets, from Kaggle repository, where each dataset contains images with and
without masks. During training, the model has attained an accuracy of 98.2%.
Keywords
COVID-19, Convolutional Neural Networks, Artificial Neural Network, Face mask Detection, Generative Adversarial Network (GAN)