Tampa Police Give Up On Face Recognition Cameras

Saint Aardvark writes "The city of Tampa has given up on their face-recognition system attached to street surveillance cameras."

Comments

Doesn't work

[JohnGrahamCumming]

If you ignore the privacy worries for a minute the most interesting thing
in this story is that the system didn't work. It didn't work in Tampa,
it didn't work in Pinellas County and it isn't working in Virgina Beach.

So you've got a dud system that's wasting police time. In Tampa they had
a full time officer using the system who could have been out on the streets
in the community that he is trying to protect understanding and interacting
with that community. If you talk to police officers, reporters, or social
workers I think you'd find that they value highly local knowledge in doing
their jobs, not all seeing all knowing eyes in the sky.

John.

Re:Doesn't work

[0123456]

"I wouldn't want to be working for that company any more though. :)"

Why? Just because it doesn't work, that doesn't mean they can't get the government to mandate its installation in all public places to catch "terrorists". That's the great thing about government contracts: it's not whether it works, it's who you know with their face in the pork trough...

Re:Doesn't work

[Znork]

Not even the fact that it doesnt work is really interesting; the fact that face recognition technology used in this way is, and always will be, worthless was known already.

Face recognition is useful when comparing small groups against a large database, or a large group against a small database since you can trim the fuzzy factors to get more false positives or more false negatives. For example, if you want to find the identity of a certain suspect in a large database you can have it spit out 10 suggestions of who it could be and eliminate the false positives manually. Or if you use it for access control you can trim it to reject as much as possible, as someone going through an access control can adjust their face for optimal lighting and try again.

But to use it to scan random people under bad conditions and compare against a large database where you dont want either false positives or false negatives is idiocy and the system will be completely useless as you'll either get dozens of random false positives each day and haul in innocent people who probably look nothing like the match or you wont get the actual matches at all.

The companies like former Visionics trying to push these systems for crowd use are selling snake oil. It doesnt work today and as the factors making it unusable cant really be significantly improved upon it wont work in the future either.

Does this mean

[Lawrence_Bird]

that they have lost face?

Yes but...

[CGP314]

But after two years, it yielded no positive identifications

I'm sorry, I didn't catch that... how many false positives did you say the system had?

A UK Solution...

[The Brain Murderer]

Here, whenever a criminal appears on the TV, their face is always partially obscured whereas everyone else looks normal.

Why not site these cameras at ports and airports as any dodgy people would appear with their faces obscured so you could just arrest them...

Another Story on the Subject in The Reg.

[markom]

The Register has a story here.

Interestingly enough, they mention successful system in Scotland being up to 70% successful in "crowd".

Re:Another Story on the Subject in The Reg.

[markom]

Spend an afternoon in London if you don't believe me that Scots have `substance dependance` issues.

Uhm, I hate to bring it to you, but London is not in Scotland...

I want cameras in New Orleans

[beacher]

Cameras on every corner. Web based cameras. Pan and zooming cameras... With some recognition software.. We could build something that dispenses beads when it recognizes ... umm...

'the closed-circuit cameras will remain'

[Chip Salzenberg]

... with more effective pattern matching software watching it: human cops. I think that's a better deterrent to crime than the flaky software they've given up on.

It broke man.

[secondsun]

Police are at a loss to explain why the software wasn't effective, since it seemed to work fine in controlled testing, Guidara said.

If I were selling you a million dollar system it would work when I showed it off too.

Finally!

[Anonymous Coward]

This is awesome ... finally I can visit Tampa again!

Re:disappointing

[curtisk]

Actually it can work fairly well in controlled enviornments/parameters

At an ATM it can/could work very well, when you're walking down the street, in motion with your head at various angls, no wonder it doesn't work. As far as PC's go thumb would probably be more likely

Advertisments

[Pompatus]

Virginia Beach, Va., installed the software on closed-circuit cameras along the city's boardwalk last summer. While it has never produced a hit or an arrest, police spokesman Sgt. Max Hayden said it performed well in controlled tests and may be a deterrent to criminals. Signs along the boardwalk inform visitors of its use.

This reminds me of a DUI checkpoint I saw a couple of months ago. They had not one, but TWO signs 6 and 4 blocks, respectively, that said, "DUI checkpoint ahead". There were plenty of opportunities to turn down another street and avoid it altogether.

Does it really take that much intelligence for a criminal to avoid an area where he/she might get caught? While one might be so drunk as to not be able to read the signs, I think law inforcement in these circumstances is being as stupid as these criminals. Maybe it's that think like your enemy strategy.

Re:Advertisments

[untaken_name]

This reminds me of a DUI checkpoint I saw a couple of months ago. They had not one, but TWO signs 6 and 4 blocks, respectively, that said, "DUI checkpoint ahead". There were plenty of opportunities to turn down another street and avoid it altogether.

I actually saw a pretty intelligent use of signs by cops once. lollapalooza was held in an outdoor venue near where I used to live. People leaving the show had to get on a limited-access highway and go about 2 miles before there was an exit, and everyone leaving the show had to drive to that exit. About a mile before the exit, they placed several large signs that said 'Drug checkpoint ahead. All cars will be searched.'
Of course, that would be illegal to do, and there was no 'drug checkpoint' at all. Instead, the police waited around for people to illegally u-turn across the median and then busted those people. We just kept driving, and sure enough, no checkpoint. After we made a legal u-turn at the next exit, we saw someone swerve across the median, and then saw two cops streak after them, sirens blazing. I don't think we stopped laughing the whole way home. Sure, it's an underhanded method, but anyone who knows their rights wouldn't fall for it.

NPR: 1) No tax dollars; 2) No (?) false positives

[dpbsmith]

According to a radio report on NPR, Tampa did not spend money directly on the system. The surveillance cameras were already in place (and will remain in place) and Identix provided the software on some kind of free-trial or beta basis. Of course, I'm sure a great deal of police time = money was wasted on training, etc.

The reporter discussed the issue of false positives with the interviewee, in a somewhat vague way. The reporter said, sensibly enough, something like "Isn't the problem that if you require too many measurements to match you don't get identifications, and that if you only require a few you get false positives?" The interviewee concurred. I got the impression that the police department might have insisted that the system be tuned to a level where they were not wasting time on false positives, and at that level there were simply no matches.

The reporter also asked (also sensibly) whether the apparent lack of success could have been because the system's installation was widely publicized and the bad guys knew better than to show up in Ybor City. Interestingly enough, the interviewee said something like "If I believed that, it would be a great thing and I'd want to keep the system in place forever." I was, however, left with the distinct impression that the interviewee did NOT believe that.

Common problem with recognition systems.

[AlecC]

This system seems to have tripped across a common problem with all id recognition systems - face, retina, voice, fingerprint, whatever. That is that they are used in two completely different modes.

One mode is the verification mode: this person claims to be Mr XYZ: is he? For this purpose, you only have one identity to match. If the answer comes out "maybe" instead of "yes" or "no", you can take another photo/scan/whatever. You can use extremely number intensive checking techniques because you are only trying to match ONE face/eye/... to ONE record. And the people being checked have at least some incentive to help their system (remove glasses, get a rescan when they have hair cut or grow beard). Systems can be made to do this very reliably in this mode - call it mode 1.

You can scale this up a little bit, while maintaining reliability. A car, for example, might recognise the voices of four registered drivers and adjust itself to suit, or a secure area form a few tens of people. Call this mode 1A.

The second mode is when you are trying to detect any one of a large list of possible people in a huge crowd, when they may have changed their characteristics significantly, either intentionalyy or unintentionally. Call this mode 2.

The trouble is that a lot of people assume that, if you can scale from 1 to 1A, the scaling from 1A to 2 will be linear. Which it won't. As well as the linear scaling of vastly more records to match (a linear scaleing), there is the the no-rescan, chjanged face, uncooperative facto, which acts quadratically with the fist. This means the problem explodes uncontrollably very soon.

Some of the people making this assumption should know better.