Contents I. Introduction
COVID-19 pandemic was one of the unexpected events in our lives which we were totally unprepared for. The World Health Organization (WHO) declared COVID-19 a pandemic on March 11, 2020, as it spread across 114 countries. [1]. The pandemic caused worldwide emergency in healthcare and caused nationwide curfew. With rapid spread of the virus, lockdowns had to be declared in most of the nations. Everybody was asked to stay home. The virus spread like a wildfire and to date (12th October, 2022), the global cases count is 62.3 Cr and death count is 65. 6L. In India,nation-wide lockdown was observed for more than two months however due to no control over the spread of virus many state governments extended the lockdown.Lockdown was a temporary measure, it led to fall of economy. As the production stopped,the prices of essential goods went high. The public faced problems because of inflation. Many people lost jobs as the company faced losses. Some were stuck in foreign countries as international travels were restricted to avoid the risk of transmission. These situations led government to start the unlock phase. Limited travels were allowed with tight rules and regulations. Wearing a mask during the pandemic was a critical preventive measure [2] and is 96 % effective to stop the spread of virus. [3] Centres for Disease Control and Prevention. WHO strongly recommends wearing a mask during public and outdoor gatherings because it blocks the transmission of the virus through the nose and mouth [4]–[5].Social distancing and wearing of mask was made compulsory. In India,government restricted gathering to control the spread of virus and allowed offices and factories to operate under 50% employee. Screening booths were installed in the entrances of public places like malls, airports, hospitals, etc. It is a herculean task for a human being to do surveillance in a crowded area. The screening guard who checks the people entering are wearing mask has to risk his/her life in the front-line. Now the situation is under control and vaccination is complete. The world has returned to the normal phase and corona is a thing of the past. However, we may never know when the next wave may hit. We have been aware of the mutation capabilities of the virus and who knows what deadly variant is coming up next. Keeping this fear in mind, we are motivated to make an automatic face mask detection system.This system can be installed in entrances of public places like malls, airports, etc. or in places where surveillance is required. Computer vision and deep learning can be used to achieve this Deep-learning based object detection algorithms can be fine-tuned to perform face mask detection. Many state-of-art methods exists for object detection. This paper makes use of a hybrid data-set with three classes: Mask, No mask, and Improper mask. The dataset used contains 1000 instances of each classes which makes total of 3000 images. Most of the images were taken from kaggle’s Mask Face Dataset [36] and more were added from MAFA [26] and AIZOO [30]. The dataset was used to fine-tune YOLO v5’s’ model to detect facemask and categorize into mask, no mask and improper mask. YOLO (You Only Look Once) is a stateof-art object detection model based on CNN (Convolutional Neural Networks). YOLO v5 model was chosen for this task specific fine tuning because of its speed and performance in low resolution images. The size of YOLOv5s is 27MB which is very small compared to other object detection models. This makes it easy to integrate in small and cheap embedded devices.
