I. Introduction
Image processing has been one of the pioneered field of computer science research which enables humans to perform tasks faster and better using the modern Artificial Intelligence methods like Machine Learning. Essentially, an image is used in the a digital format and passed as an input to Convolution Neural Networks [1] which perform Convolution operation and subsequent neural networks can use it for classification or clustering purposes. The arrival of COVID-19 comes with difficult challenges for health systems globally. The continuous mutations in the virus structure led to the emergence of different variants and forced countries to take strict measures like restrictions on travel of individuals, locking down cities or nations, isolating the infected individual, and quarantine of individuals who are suspected to have been exposed to the virus and positive cases, advising the sanitization of hands at regular intervals, temperature monitoring, mandatory use of face masks, and following social distancing in public spaces [2]. The task of monitoring a huge number of individuals is becoming an increasingly difficult task. The monitoring is the process to detect anyone who is not wearing a face mask. Study on working of different machine learning based CNN’s [1] has been done by Mahdianpari [3] for land cover remote sensing which inspired us to perform this study for face mask usage. In this study, we compare different widely used methods for this Face-Mask identification Problem, using their complexity, execution time, performance. Rather than using trivial data-sets with only binary classes, namely, masked faces and unmasked faces, as used in in hybrid deep transfer learning model [4], we use a third class of improper face masks thereby making it a multi-class classification problem. Different ma- chine learning models based on convolution operations are used to get better insights on their working and which one performs better in an low compute, low quality and real time environment like surveillance.