Hey guys,

Today I wanna show my newest my project: the Mask-Detector. It is a pytorch based AI that detecs whether you wear a mask or not!

First step: search for data

The hardest problem was to find enough data. I used two Kaggle datasets which contain portraits of persons wearing a mask. So I downloaded them.
To have picture without masks, I used https://thispersondoesnotexist.com/. They create fictional faces with another AI (GAN).

The algorithm

Firstly all images were searched for faces whether they wear a mask or not. This does a neural network by opencv (pretrained). Then all found faces will be prepared and go through my own neural network (4 conv layers). Finally I should have a boolean which contains a True when someone wears a mask.

But where should someone use this toll?

In my home land (Germany), we have nowadays a law to wear our masks in shops and other public areas. But someones do not want a mask or just forget it. So a staff member for example has to convice them to wear a mask. This is super annoying why I got this idea.
To collect the data we need (the faces), we can use the builtin cameras in stores.

Results

On my test data I have got 97% correct classifications. This is brilliant!!

If you interested to try it, you can view my GitHub Repo (https://github.com/Christopher-06/Mask-Detector). There is also an instruction how to use it!

Do you have any questions? Just ask in the comments!

Thanks, Christopher