Facial Recognition for People Wearing Masks

The Chinese facial recognition company Hanwang claims it can recognize people wearing masks:

The company now says its masked facial recognition program has reached 95 percent accuracy in lab tests, and even claims that it is more accurate in real life, where its cameras take multiple photos of a person if the first attempt to identify them fails.

[…]

Counter-intuitively, training facial recognition algorithms to recognize masked faces involves throwing data away. A team at the University of Bradford published a study last year showing they could train a facial recognition program to accurately recognize half-faces by deleting parts of the photos they used to train the software.

When a facial recognition program tries to recognize a person, it takes a photo of the person to be identified, and reduces it down to a bundle, or vector, of numbers that describes the relative positions of features on the face.

[…]

Hanwang’s system works for masked faces by trying to guess what all the faces in its existing database of photographs would look like if they were masked.

Posted on March 25, 2020 at 6:33 AM11 Comments

Comments

David Leppik March 25, 2020 10:30 AM

It’s 95% accuracy with up to 50,000 faces, but actually trained with a database of 2,000 people, so the 50,000 is speculative. That’s 1 in 20 false positives with a population well under a million. Reading between the lines, it may also require people to aim their faces at the camera. Good enough for unlocking a door in a small office, not as useful for mass surveillance. (That’s what cell phones are for!)

Also, it’s using plain face masks, and since it’s still trained to work without masks, I’m sure it could be fooled with one of these masks.

I really wonder how much information there is in a face for accurate recognition once you remove the nose and mouth. That’s why I’m skeptical on this one.

That said, there are plenty of other biometrics, such as body shape and walking gait, which have not been explored because face recognition works well enough.

Mark March 25, 2020 12:12 PM

So if this is accurate, it would mean that just the forehead and eyes (including eyebrows) is sufficient for accurate facial recognition? Assuming ‘face’ doesn’t include hairstyle, ears, etc.

La Abeja March 25, 2020 12:41 PM

@Mark

Assuming ‘face’ doesn’t include hairstyle, ears, etc.

The Finnish word for “face” is kasvot, always plural.

It is related to kasvu, “growth,” from kasvaa, “to grow,” certainly in reference to facial hair.

However the ending “-o” sometimes carries a certain privative or existential meaning.

Heikki is a common man’s name in Finnish, but heikko means “weak,” as a man deprived of his strength.

The noun huone means house, home or “room” in Finnish, but huono means “poor” or “homely” — as deprived of such means or abode.

So kasvot actually refers to the quality of the beard or facial hair in an existential sense, as perhaps in the smoothness of a woman’s face, or in that if a man shaved you could see his “face.”

The word is somewhat majestic or royal in a certain sense as well, kasvoista kasvoihin, said when people are having a close conversation.

lurker March 25, 2020 2:24 PM

There’s a lot of information in eyes, besides colour which used to be the identifying factor in last century ID summaries. Size, shape, distance apart, angle of inner-outer corner lines, vertical position, &c. Also from last century was the newspaper trick of de-identifying a photo with a narrow rectangle block over the eyes.

Alejandro March 25, 2020 4:40 PM

I don’t have the link handy at the moment, but as I recall this system is applied to a large group of captive factory employees who had to put up with multiple mugshot photos and are somewhat staged for the cameras to work. Changing clothes can fake it out, I think.

The implication: it doesn’t work in open public areas.

Curious March 26, 2020 4:21 AM

I have zero faith in all of this being ethical and also anything “ethical” can’t be kept
a secret the way I see it, and so presumably nation states just do what they want anyway. Just wanted to say that as I don’t trust my own country, and certainly not trusting China, USA, UK, or France for that matter, to really behave ethically and without abuse and I don’t care what the legality of it all, or what an interpretation of legality would be.

I am no expert on such issues, but some things come to mind as I sit here:

1) Id’ing vs tracking (cameras or supplementary surveillance efforts following people around even if you can’t properly id individuals)

2) Potential mass arrests, in needing to treat basically everybody as a suspect. I imagine it would be like a fish net, to catch the largest fish so to speak, but not knowing where they are found.

3) Parallel construction problem? Learning too much and can be abused?

4) Mass surveillance as a self fulfilling prophecy of sorts. As if the nation state must have more cameras and become ever more intrusive, or it won’t work in an ethical and restricted way.

I feel a little dirty mentioning the word ‘ethical’, as I don’t believe ethics is any guarantee for good behavior, lawful behavior, or good intentions, and certainly not something that is to benefit everybody and is ruled by special interests like those involved in warfare for example (just one example).

Catherine Johnson April 21, 2020 5:28 PM

This technique is very famous these days and recently some incident occurred with my https://www.lakenormanhardscapes.com/ team fellow and this new technology of facial recognition helped him a lot to trace the root of the matter.This is very informative in criminal detection.

Leave a comment

Login

Allowed HTML <a href="URL"> • <em> <cite> <i> • <strong> <b> • <sub> <sup> • <ul> <ol> <li> • <blockquote> <pre> Markdown Extra syntax via https://michelf.ca/projects/php-markdown/extra/

Sidebar photo of Bruce Schneier by Joe MacInnis.