Face Mask Detection Model for the New Normal
Published on : Sunday 04-10-2020
Jahnavi Shah elaborates upon a face mask detection model for safety during Covid-19 pandemic.
On 11th March 2020 the World Health Organisation (WHO) declared Coronavirus as a global pandemic with cases crossing 100,000 mark. Within two weeks, a nation-wide lockdown was ordered as a preventive measure to keep the 1.3 billion people of India safe. Businesses, colleges, schools, and restaurants had to shut down and people were trapped inside their homes with access to only basic necessities. The world had come to a standstill. After almost nine months of being acquainted with the virus we have slowly and steadily started to pick up. Vaccines are being curated and everyone is hoping to get back to their normal life. A lot of efforts have been made to resume the life back to normal but with some changes to ensure safety, the ‘New Normal’.
In a dense country like India where the population density is 382 per sq.km, the containment of the virus was of utmost priority. While several efforts have been made by the government to sanitise and ensure social distancing in public places, a collective effort from the people is required to contain the spread. The most viable and cheap solution to reduce the chance of getting infected is the ‘Face Mask’.
Why a face mask detection system?
The image shows how a simple mask can make a difference. By wearing a mask we can significantly reduce the risk of spreading the virus to others.
Even after restrictions and fines being imposed for not wearing a mask, people still step out without one, or even if they wear it, their nose or mouth is often left uncovered. It would be strenuous to physically monitor every person to authenticate their access. We live in the 21st century where almost everything is technology-driven. The Face Mask Detection system is built on the stems of technology like Tensorflow, OpenCV, and Machine Learning. It can automatically authenticate the access of the person by a video/camera feed.
With life getting back on wheels, we need to ensure safety for one and all. This face mask detector can be deployed in any place where people flock. The live feed from the camera and the algorithm running in the background will automatically detect if the person is wearing the mask correctly and whether he should be given permission to enter the place.
How does this model work?
This Machine Learning model can be classified under Supervised Learning. In Supervised Learning, every training example has a corresponding label. The algorithm identifies the label and classifies the image accordingly. This model first identifies a Region of Interest (RoI) which in this case is a face. Further implementations are done within this specified RoI. The detector is trained on 1000+ images from the Kaggle dataset for Face Mask Detection. Also an additional of 500 images were scrapped from the net to provide better results. Since it is trained on a diverse set of images the model works correctly to identify three classes:
1. With Mask: The person is wearing a mask and it is covering his nose and mouth.
2. Mask Worn Incorrect: The person is wearing a mask but it is not covering his nose or mouth.
3. Without Mask: The person is not wearing a mask.
In all the above cases only ‘With Mask’ is considered acceptable.
Where can we use this face mask detection model?
Whenever we enter a new place outside our homes we always find a sanitiser stand and temperature guns, this face mask detection model can be used as an extra precautionary. It can be deployed in office spaces, shops, restaurants, airports, and even homes. The people entering can
face the camera, the feed will be sent to the back-end algorithm, and if the person is wearing the mask properly then he can be allowed and if not subsequent actions can be taken. It can be customised and scaled for different applications like with a speaker or even with an auto door.
The GitHub repository for the project can be found here: https://github.com/Project- SafeShop/SafeShop_Camera_MaskDetection
Jahnavi Shah is a 3rd year Computer Engineering student at Pandit Deendayal Petroleum University, Gandhinagar. She has keen interests in the fields of Data Science and Machine Learning and aspires to specialise in them. She is researching in the fields of integrating Psychology with Machine Learning to help improvise on mental health. Python is her forte, and she is also an experienced programmer in languages like C, Java and SQL.