There was recent scandal [1] about companies crawling the web for pictures and then offering an easy search engine that allows you to search for pictures similar to the ones you upload.

I tried PimEyes and it was shocking and relieving and the same time. Shocking was how much of the webs they crawled (I could find pictures in really remote corners of the web) it was relieving that there were a number of false positives in the result.

My questions are:

1) How about false negatives?

2) Why is this news only blowing up now? The standard building blocks like image segmentation for faces, SIFT features and basic machine learning similarity measures were already available 10 years ago, so it seems the tech was there.

3) Some sites [2] claim unusual feats of face detection (e.g. people wearing masks). Is there any research on how much better the state-of-the-art for face recognition is than what humans are able to do? Is there any research on how much of your face you would need to hide to currently evade the state-of-the-art?

----------------------------------------------------------

[1] https://edition.cnn.com/2022/05/24/tech/cher-scarlett-facial-recognition-trauma/index.html

[2] https://eudatasharing.eu/news/pimeyes-thorough-facial-recognition-tool-accessible-everyone-raises-privacy-concerns