I think facial recognition may be fairly accurate in some cases even when wearing a mask. Somewhat anecdotal, but saw this article recently:

https://onezero.medium.com/you-can-fool-a-popular-facial-rec...

I've read the article and it's about facial detection - it detects that there is a face on the picture.

Haven't read anything about identifying the person wearing the mask.

Systems are getting pretty good at working around masks, but the error rate is still greater when masked than when unmasked.

With most systems, if you want to maximize the error rate then you should wear a solid black mask that covers the entirety of your lower face.

Having actually worked in the space, no one cares about identifying you*

They're trying to do sentiment analysis and maybe some sort of classification on age/gender

*Yes, I'm sure if you Google for an anecdote of a system that did identify users you'll find one, but it really gains nothing in the space. They already have your card data and rewards programs, usually what they care about it sentiment and demographics at a specific spot in the store

Perhaps that's true in the retail space, but I was really addressing all uses and not just retail. However, if wearing such a mask can increase the error rate when it comes to sentiment analysis and age/gender classification as well, that's gravy.

OK a mask with someone else's face