The FBI has confirmed it is using Amazon’s Rekognition image and video analysis software, but Amazon says it is “false” to suggest this violates the company’s self-imposed moratorium on selling facial recognition technology to US police
New Orleans did a pilot with allowing facial recognition for major felonies and it just didn’t work at all. Of the 15 requests, 9 failed to make a match and of the 6 that did return a “match,” 3 were the wrong person.
That’s a small sample size — most cities don’t release data — but it explains why cities that happily use facial recognition software don’t see reductions in crime or cleared cases. It’s just a complete waste of money and the investigators’ time. Facial recognition tech can (usually) identify friends in photos but criminals aren’t posing for fancy modern phone cameras in decent lighting. They’re using security cam stills and anyone committing a major felony probably has their face at least partially covered.
It’s like that software that’s supposed to identify gunshots but has so many false positives, police stop even bothering to follow-up after awhile. Maybe not as stupid as the NYPD buying robots but still a huge waste of resources.
Not to mention the primarily black population of Louisiana who are disproportionately targeted for crimes will be the greatest number of false positives .
Due to the common knowledge the facial recognition is severely lacking when it comes to successfully identifying black populations due to darker tones being harder to differentiate.
This would take an existing problem and magnify it by creating a larger, more automated police state.
Not to mention data privacy/security, which the LA DMV just got their data hacked in the last two years…
Due to the common knowledge the facial recognition is severely lacking when it comes to successfully identifying black populations due to darker tones being harder to differentiate.
It’s not even this, if you use high quality cameras, black skin can be easily and well photographed.
The issue is by and large most security cameras use super cheap cameras that don’t have a wide range of color they can capture, so you end up losing all definition on the faces of darker skinned individuals.
This has been an issue since cameras were first invented, and better cameras that solve this issue were created long ago.
It’s because our capitalist overlords are cheap bastards.
No I don’t think this is a training issue. Light skin physically reflects more light, which gives cameras significantly more data to work with to detect shadows/etc.
The face recognition on a phone gets around that by creating their own light, with dot projection at a light wavelength the human eye can’t see, but I’ve only ever heard of that being done for short range face recognition. CCTV cameras are too far away from the face and are not really accurate enough for anyone (including white people).
AFAIK Amazon’s system was mostly intended for their self service retail stores… that’s a different scenario entirely since you’re only comparing faces to other customers who are in the store at the same time as you. And also the stakes are much lower - if two people appear to be the same person you can just flag both customers as needing to be verified by a staff member. No big deal at all.
Using it as evidence for a crime though, will inevitably result in false convictions.
Law enforcement isn’t interested in solving crimes. It’s interested in securing convictions. False convictions do nicely to advance police and DA careers. And innocent bystanders are fine for putting warm bodies into empty prison cells.
New Orleans did a pilot with allowing facial recognition for major felonies and it just didn’t work at all. Of the 15 requests, 9 failed to make a match and of the 6 that did return a “match,” 3 were the wrong person.
That’s a small sample size — most cities don’t release data — but it explains why cities that happily use facial recognition software don’t see reductions in crime or cleared cases. It’s just a complete waste of money and the investigators’ time. Facial recognition tech can (usually) identify friends in photos but criminals aren’t posing for fancy modern phone cameras in decent lighting. They’re using security cam stills and anyone committing a major felony probably has their face at least partially covered.
It’s like that software that’s supposed to identify gunshots but has so many false positives, police stop even bothering to follow-up after awhile. Maybe not as stupid as the NYPD buying robots but still a huge waste of resources.
Not to mention the primarily black population of Louisiana who are disproportionately targeted for crimes will be the greatest number of false positives .
Due to the common knowledge the facial recognition is severely lacking when it comes to successfully identifying black populations due to darker tones being harder to differentiate.
This would take an existing problem and magnify it by creating a larger, more automated police state.
Not to mention data privacy/security, which the LA DMV just got their data hacked in the last two years…
Seriously a public safety violation all-around.
It’s not even this, if you use high quality cameras, black skin can be easily and well photographed.
The issue is by and large most security cameras use super cheap cameras that don’t have a wide range of color they can capture, so you end up losing all definition on the faces of darker skinned individuals.
This has been an issue since cameras were first invented, and better cameras that solve this issue were created long ago.
It’s because our capitalist overlords are cheap bastards.
Also possible that the systems are trained on light skinned people more?
No I don’t think this is a training issue. Light skin physically reflects more light, which gives cameras significantly more data to work with to detect shadows/etc.
The face recognition on a phone gets around that by creating their own light, with dot projection at a light wavelength the human eye can’t see, but I’ve only ever heard of that being done for short range face recognition. CCTV cameras are too far away from the face and are not really accurate enough for anyone (including white people).
AFAIK Amazon’s system was mostly intended for their self service retail stores… that’s a different scenario entirely since you’re only comparing faces to other customers who are in the store at the same time as you. And also the stakes are much lower - if two people appear to be the same person you can just flag both customers as needing to be verified by a staff member. No big deal at all.
Using it as evidence for a crime though, will inevitably result in false convictions.
Better cameras aren’t enough - you also need much brighter lighting. So bright customers would complain and get headaches.
Or alternatively, have the camera close to the customer’s face. Again, nobody wants that… though we do put up with it for ATMs / etc.
Cops and wasting resources on stupid shit that doesn’t actually help them solve crimes.
Name a more iconic duo.
Law enforcement isn’t interested in solving crimes. It’s interested in securing convictions. False convictions do nicely to advance police and DA careers. And innocent bystanders are fine for putting warm bodies into empty prison cells.