Technology fighting Covid-19

IA identifies people not wearing masks


What is the project about?

Why not use IA to fight Covid-19? Imagine surveillance cameras able to determine automatically whether someone is wearing a mask or not. Such a system could be used in many places such as hospitals, schools and stores, and would pave the way to many applications :

  • trigger a pop-up on a large screen to remind people they have to wear a mask.
  • make a sound everytime someone comes in without a mask.
  • open automatic doors only to persons wearing a mask.


    • Many benefits for the users of this technology:

  • participation to the fight against covid-19.
  • garantee to comply with government's mask laws.
  • employees don't have to spend time to check masks wearing.
  • a reminder from a machine is less unpleasant than a human comment.
  • the most appropriate response is chosen in advance : reminder, closing of doors, triggering of staff calls...

    • By the way, the last point may make you think about a simplified - and maybe more ethical?- version of the Social Credit System implemented in some regions in China.


    This is definitely not science fiction!

    The principle is rather simple. Surveillance cameras coupled with machine learning algorithms analyse all the people entering into, let's say a hospital. The cameras can differenciate a face with a mask from a face without one. For that to work, the machine needs to be previously trained, by having as many pictures of faces - with and without maks- as possible. This technique is called supervised machine learning.

    If you want to try out yourself!

    Some js code to help you:

    The html part:

    One of the major tool you need is ml5.js. You can learn how to master this js library here .

    Especially, have a look at image classification

    Some examples use the MobileNet pre-trained model. (like the one just above) It is not efficient for mask recognition, you will have to train your own specific model using Teachable Machine. For instance, create a model with 3 classes "nobody", "someone-with-a-mask" and "someone-without-mask". You can learn how to master TeachableMachine here.

    You can find a data set for face mask recognition here .

    JS basics to go further than mask recognition: trigger an alert message, or whatever you want... here.

    If you don't manage to make your project work, look for existing solutions (such as the programm above using Python). LeewayHertz

    Good luck and have fun!