A Comparative Study of CNN Networks for Face Mask Detection
COVID-19, a virus that badly hit the world which had shown as many adverse effects as possible in terms of all the possible fields like causing illness, an economic imbalance to the nations, adverse effects in health conditions of the people, etc. It is believed and proven fact that wearing a face mask can be more sensible to minimize the possibility of being affected by covid-19. However, proper monitoring is needed in public institutions like health care centers, airports, offices to prevent the infection. Hence use of intelligent systems is the proper solution to the problem. By keeping in mind that it shouldn’t delay the working process, an accurate and quick system of face mask detection is needed. The better use of the intelligent systems in the given problem statement is obviously the computer vision when powered by artificially intelligent systems. In this paper, we present the research work we have conducted on the use of deep learning algorithms in the implementation of Face Mask Detection systems and determining the best one under the given circumstances. Here we considered the various types of CNN networks that can act as a backbone for the model by keeping in mind the economical constraints possessed. So, we implemented three types of CNN networks i.e. MobileNetV2, ResNet50V2, and Xception for face mask detection system and compared the results accordingly.
Keywords- Face detection, CNN, RCNN, ResNet, MobileNet, Object detection