- Joined
- 30 May 2024
- Messages
- 6,880
- Reaction score
- 2,778
- Country
Maybe they need to look at who produces and manufacture these cameras. Maybe its they who are racist and not the technology
Facial recognition cameras appearing to be working.
- Thread starter Elsa TLC
- Start date
- Joined
- 22 Aug 2006
- Messages
- 7,635
- Reaction score
- 1,389
- Country
I think the problem is that black faces are lower contrast, so it's more difficult to tell individuals apart.
It was more accurate picking the black suspects than the others, if it was better picking the white suspects nothing would have been said/done.
- Joined
- 22 Aug 2006
- Messages
- 7,635
- Reaction score
- 1,389
- Country
That's not what I read.
I just looked and found reports like this:
The unseen Black faces of AI algorithms - Nature
The unseen Black faces of AI algorithms An audit of commercial facial-analysis tools found that dark-skinned faces are misclassified at a much higher rate than are faces from any other group.It's LESS good at identifying dark faces correctly, for the reason I thought.
50.000 scans is meaningless
we need to know iff it individual instances off a person passing or filtered to new faces
if centrally placed a family off 4 wandering about may be recorded say 3 or 4 times so will register every time they pass but instantly deleted so recorded next time as another 4 faces??
we also need to know what measures were taken to check for those taking avoiding action and if they where checked otherwise those wanting to avoid detection and surveillance aware will steer well clear off the van
the obvious thing is 2 well-marked surveillance dummy vehicles and a genuine discreet vehicle on the avoiding route
we need to know iff it individual instances off a person passing or filtered to new faces
if centrally placed a family off 4 wandering about may be recorded say 3 or 4 times so will register every time they pass but instantly deleted so recorded next time as another 4 faces??
we also need to know what measures were taken to check for those taking avoiding action and if they where checked otherwise those wanting to avoid detection and surveillance aware will steer well clear off the van
the obvious thing is 2 well-marked surveillance dummy vehicles and a genuine discreet vehicle on the avoiding route
Last edited:
my understanding is they work on permanent features like eyes nose mouth chin ears and the spacing off each item within the face oval anything else will be off no value and will be discounted as much as possible as fully changeable
but off course it may have the ability to connect adaptable features like clothing and facial hair iff it has a track record off being connected to the face "oval" over time
- Joined
- 30 May 2024
- Messages
- 6,880
- Reaction score
- 2,778
- Country
Im thinking the more difficult identification might be able to be solved over diffrent locations if there is not 100% facial, yet clothing could join up the dots
i would think they have to keep the scope off adding surrounding detail to a minimum otherwise is a busy area where a face oval will be reasonably easy to separate quite accurately if you start including say 100mm/4" around the face you will start to get other peoples hair and clothing infringing in that area causing reduced accuracy or possible misinterpretation
now i assume it could be set up to include other feature that are fully moving in time with the face can be "connected " in a building up profile way but assume the computing power will increase possibly massively due to the far greater level off information storage and level off computing required to differentiate the information and connect the hat/coat /hair what ever to the correct face....but purely a guess ??
- Joined
- 30 May 2024
- Messages
- 6,880
- Reaction score
- 2,778
- Country
I see what what you say. Yet a fleece may have a stain or a rip on the shoulder.
This maybe enough to know that matching that could just encourage that they are not barking up the wrong tree.
