'Faceless Recognition System' Can Identify You Even When You Hide Your Face (vice.com)
schwit1 quotes a report from Motherboard: By itself, the ability to instantly identify anyone just by seeing their face already creates massive power imbalances, with serious implications for free speech and political protest. But more recently, researchers have demonstrated that even when faces are blurred or otherwise obscured, algorithms can be trained to identify people by matching previously-observed patterns around their head and body. In a new paper uploaded to the ArXiv pre-print server, researchers at the Max Planck Institute in Saarbrucken, Germany demonstrate a method of identifying individuals even when most of their photos are un-tagged or obscured. The researchers' system, which they call the "Faceless Recognition System," trains a neural network on a set of photos containing both obscured and visible faces, then uses that knowledge to predict the identity of obscured faces by looking for similarities in the area around a person's head and body. As for the accuracy of the system, "even when there are only 1.25 instances of the individual's fully-visible face, the system can identify an obscured face with 69.6 percent accuracy; if there are 10 instances of an individual's face, it increases to as high as 91.5 percent."
who is that masked man!!!
if only we had the this tech available in all the superhero movies...
the new number of the beast
They'll never identify me!
“He’s not deformed, he’s just drunk!”
yup correctly identified as the faceless man.
I applaud their efforts, even though I'm a little horrified at the lengths to which we'll go in our quest to remove anonymity.
While I'm sure there are good uses this can be put to, I wonder if any of the researchers questioned whether this was really a good idea...
Now, if this system was capable of identifying the person responsible for {insert bad corporate act(s)} through the corporate veil well enough for them to be meaningfully punished, that would be incredible!
I wonder how this works with a large pool size.
Sure, it might be able to spot "Bob Smith" in a crowd if the only data it has is several shots for "Bob Smith" and maybe 10 others. How about when it has to store data on a few thousand or million people. I think at some point accuracy goes out the window as it mistakes "Bob Smith" for any of the other million or so users it has data on.
the kaonashi recognition system.
These are all pictures where the person was not trying to modify their appearance to be able to hide in plain sight. Temporary collagen injections (or even just some padding inside your cheeks), using tape (hidden under a wig) to raise the eyebrow arch vertically, or stretch the eyelids horizontally, some padding to give yourself dumbo ears, use of makeup to alter the appearance of facial bone structure via shading and highlighting, making the nose narrower using dabs of crazy glue to stick parts of the nasal septum together ...
Enough differences, and the identification will fail.
"Transparent" is a shit show that trades on every stereotype going. A man in drag is NOT a transsexual.
Pretty soon computers can be programmed to identify people by their behavioural norms -- such as walking gait and other body language.
That's good enough. Take the shot.
Look, WW II methods still work. Change clothing, reversible clothing, alter stride (the ministry of silly walks was actually based on real deep cover training), dazzle face paint, minor changes in cheeks, nose, chin, etc.
They just want you to think they can spot you.
And, yes, hoodies work. Best are backscatter hoodies with team logos that change.
Hats also work.
Facial recognition has a high failure rate in real world ops.
Anyone that tells you otherwise is trying to sell you a bill of goods.
-- Tigger warning: This post may contain tiggers! --
From Robocop: "I had a guaranteed military sale with ED 209 - renovation program, spare parts for twenty-five years... Who cares if it worked or not."
https://www.youtube.com/c/BrendaEM
By itself, the ability to instantly identify anyone just by seeing their face already "creates massive power imbalances"
91% success is impressive performance, but it maybe isn't that useful for, eg, spotting suspects on CCTV footage. For example, the London Underground carries nearly 5 million passenger journeys per day. 9% of that is 450,000. So, if we are talking false positives, nearly half a million non-suspects for humans to check every day. To put it another way, it's several non-suspects to check in every single carriage of every single train during peak times. This is why global surveillance often isn't a very good way to catch the bad guys.
Virtually serving coffee