Hacker News Reader

Comment by bluedino | Hacker News Reader

People wear facemasks in camera footage now, so you have to go by their pants, belts, shoes, fingerprints..

Gait analysis is a more surveillable fingerprint anyway, since it can work with lower resolution cameras. Tag someone on a high res camera with facial recognition and as long as you've seen their gait once, you'll be able to identify them again.

Stones in our shoes, masks on our faces, strobing hoodies a boring dystopia indeed.

Discussed previously, apparently (found via Google search for `anti gait recognition shoes`, lol):

https://news.ycombinator.com/item?id=18401892

I like the comment about the Ministry of Silly Walks.

IIRC stones in shoes as a gait recognition defence were a plot point in Doctorow’s Little Brother (2008).

Yeah I read that in a novel too. It’s a common idea.

Because spies actively use the technique and it's routinely discussed in that setting.

An uncomfortable and time consuming one indeed.

I don't think gait analysis would hold itself through a trial or be bought by a jury. Maybe to correlate different videos of possibly different suspects? But if you are getting regularly videoed by security cameras while circumventing rules you are probably already wearing different shoes or pants and that can go a long way to defeat gait analysis algos.

It's quite literally more accurate than fingerprint or blood splatter analysis and those send people to prison on a daily basis - despite being little more than snake oil techniques.

Here's a handy guide: https://en.wikipedia.org/wiki/The_Ministry_of_Silly_Walks#/m...

Has gait analysis been studied by people other than the ones selling gait monitoring equipment? It smells to me like bite mark analysis, blood splatter analysis, polygraphs, and other "miracle" technologies/techniques that turned out to be horsecrap.

There are people I know that I can recognize by the way they walk, so the idea is at least plausible. I think the main problem is/will be from using the analysis inappropriately.

It's one thing to narrow down a city-sized population to a list of 50 people known to walk with a similar gait, and then follow up with other forms of investigation before drawing conclusions. It's another for a cop to look at the top match and go arrest that guy, as they've done more than once with facial recognition.

Gait analysis is absurdly over hyped. I wish people would stop trying to win arguments by mentioning at think twice about the numerous failure cases.

Gait analysis also works surprisingly well and I don't see how to make it less effective without damaging your body

Show me a security camera that does gait analysis and I will pay you $1000.

You’re offering to pay a grand to random HN contributors to wire one up.

Maybe that's the point? Maybe the person needs one of these and $1000 is well worth it, and it just needs to be

I don’t get that sense.

The person knows no cameras exist doing gait detection, but wants one. They make a random comment on HN, a forum known to be read by people that is full of people will take an idea to completion "over a weekend", as a challenge to get someone to do something for $1000. That's a ridiculously low amount for something to come into existence that would serve their need directly. Hell, there's probably people here that would do something like this for the clout.

No, I know that where security cameras are positioned and the amount of processing they currently have there will be no gait analysis being done on the camera itself. They have enough trouble telling one person from another reliably.

I was being facetious as I don't believe hackernews readership is that enterprising.

I was going to make an edit with something along the lines of "or the poster was taking the piss out of the concept being real", but left it alone ;-)

Ill do you one better, $500K and plenty of video to do that gait magic https://www.nbcnews.com/politics/justice-department/feds-boo...

https://www.newscientist.com/article/mg21528835-600-cameras-...

I thought I once saw on Reddit that target already uses this in stores. I could be wrong.

https://ggbmagazine.com/article/in-the-masked-age-operators-... Here you go. they use them in China already.

Ah yes, ggbmagazine. Clearly a Hikvision product you can buy...

Everyone walking around in huge rigid gowns/boxes of uniform size.

The privacy concerned will walk around look like Minecraft characters.

Gait analysis works as good as jeans wear marks

https://www.pnas.org/doi/10.1073/pnas.1917222117

or bite marks forensics, aka we can make it work against any guy prosecutor picks

In real world without someone to fram^^^compare to it seems pretty useless https://www.fbi.gov/contact-us/field-offices/washingtondc/ne...

There have been pretty huge leaps over partially obstructed facial recognition algorithms in the last few years -- I think state of the art is approaching no meaningful loss of match mask vs no mask.

Negative ghost rider.

It is most certainly not, unless you’re talking straight on well lit training data.

This is not my understanding, I have read at least 7 or 8 papers that seem to have for various models and techniques reduced the delta between masked and unmasked recognition to be very similar on false positive and positive rates.

https://www.mdpi.com/2076-3417/11/16/7310 https://www.sciencedirect.com/science/article/pii/S240589632... ...

That said I have no insight as to how many of these techniques have been found to scale well or have started to make it into product. It has been publicly reported that NEC’s NeoFace (a system that many police and govt use) newer versions does indeed have occlusion (mask) recognition operating at very high levels.

anyways thats just my understanding as an interested bystander -- not in the field.

How could this be true? I don't follow the field but so much of the signal is being obscured it's hard to imagine how accuracy wouldn't suffer

tldr machine learning is magic

iPhone face unlock has worked while wearing a mask for a while now.

You know how you can identify exactly which family member is walking in the door from across the house, and maybe even what mood they're in? Or how little clarity you need to identify your child or spouse on a 480p camera feed? That's what machine learning makes possible across all types of sensor input, picking out those little but distinct patterns. There's really no way to be anonymous in public once ML surveillance software is widespread.

Reply on news.ycombinator.com