Slashdot Mirror
Racist Facial Recognition Software
An anonymous reader writes "A black man found that his HP facial-tracking recognition software wouldn't work. Then he discovered it worked fine for a white co-worker. From the article: 'HP's Tony Welch thanked Desi and Wanda, the video's creators, and promised that he and the team at HP were looking into why the camera was behaving the way it was. "The technology we use is built on standard algorithms that measure the difference in intensity of contrast between the eyes and the upper cheek and nose," he said. "We believe that the camera might have difficulty 'seeing' contrast in conditions where there is insufficient foreground lighting."'"
With this extensive testing performed on the camera (only 2 subjects), i wouldn't dare to say for sure that it is because he is black. They two are different on more things than skin color (i.e., sex).
Anyway if I had to bet I would also bet on "blackness", but I'm just pointing some flaws in their straightforward conclusions.
I do not think HP is out right racist. But his skin color does have an effect. His eyes, cheek, and nose are of similar color.
Therefore the recognition software will interpret no difference in variables, because it can not see any contrast. This makes triangulation impossible. Simply a bug in software design.
A possible solution would be an adjuster to the contrast ratio for the recognition points. Allowing the user to adjust it. Though hard to teach to people who prefer plug-n-play style of settings.
it just doesn't trust white people and need to keep an eye on them~
It's also pretty clear the the creators of the software where white.
The Kruger Dunning explains most post on
When the robot wars come, then we'll see who's laughing...
Notice how slashdot jumps on the item as inflammatory of a way as possible.... calling it racist. Way to be fair slashdot.
Luckily, a similar situation was addressed and resolved in a Better off Ted episode.
They're describing the Viola-Jones algorithm. Because it doesn't use colour (frames are converted to monochrome and normalised to account for different lighting conditions) it usually works well regardless of skin colour.
software developed and tested on the faces of pasty white software engineers.
The recognition software is looking for contrast spots, such as shadow from the nose, eye-sockets, and ultimately the head shape, if these criteria isn't met - then it has a problem.
It's the same with eg. Sony PS3's Playstation EYE - and eg. the new game EYEPET, you try that on at home against a dark floor, dark carpet, no matter how much light you put on, the pet won't react to you.
Same with me - I'm a white guy, and unfortunately my floor is sorta white colored (wood, flesh like...if you're slightly yellow in the skin, don't worry - bad indoor light WILL look pale yellowish) ;)
Even with 200 watts of lighting - I had problems with my EYEPET and PS camera, it's virtually useless and incredibly annoying.
Trust me buddy, there's no difference with HP's camera software, the same stuff. Wait until MicroSofts Project NATAL comes out, then you're going to be pleased, it will reckognize ALL colors! Because it has an extra infrared camera!
You're kind of funny though ;) You have a talent for talking on tv, not everybody have that gift, you do!
What this world is coming to - is for you and me to decide.
As I was watching this, I got "Find an African Date" AD from Youtube... Something strangeis going on here.
...this was an "edge case" they obviously did not test for!
I hate being bipolar; it's awesome!
No, this is funny and ironic. It isn't inflammatory. Most racism singles out "people of color", and this facial recognizer matches only white faces.
is the person that submitted this. It is DIFFICULT to get this kind of thing done correctly. Heck, America has spent BILLIONS since 9/11 trying to handle facial recognition. And we are still nowhere near close. To the original submitter, please keep your racism to yourself.
I prefer the "u" in honour as it seems to be missing these days.
SOMEBODY developed this facial tracking software, and HP vetted it for installation on thousands or millions of computers.
Either this problem will come as a complete surprise to them, or they knew about it and released it anyway. Both alternatives are pretty upsetting.
Because either there were no test cases involving black people -- for an algorithm that depends on skin contrast, you'd think this would be a no-brainer -- or they knew there was a problem, but never expected black people to buy it.
It's supposed to be completely automatic, but actually you have to press this button.
A project that doesn't get outsourced to India and look what happens.
You'd think this story would bring them out of the woodwork...
When in fact Skynet becomes self aware, this will be a good thing for the dude :)
1. get a HP laptop
2. tell ppl the laptop track ppl's soul
3. claim u can get their soul back for them for a fee
5. ???
6. PROFIT!!!
Beer tap?
What kind of entertaining robot doesn't serve beer?
I mean come on this is 2010.
Not to be offensive, but how greasy is this guy's face? You'd think there would be *more* contrast between his skin and his corneas and teeth, which would make a general "face" easier to pick out (even if it were a Cheshire-cat caricature, it's a baseline). However, there's a noticable glare off his forehead that I'd imagine would skew the results (as it'd significantly take a chunk out of his otherwise-round head to the point it wouldn't even hit on a fuzzy match).
My suggestion is to get this guy some Noxzema and try again.
Never attribute to Hanlon that which can be adequately attributed to Heinlein.
If you can't see his face, you are racists! http://www.epicfail.com/wp-content/uploads/2009/11/wedding-photo-dark-fail.jpg
lmao
racism
1 : a belief that race is the primary determinant of human traits and capacities and that racial differences produce an inherent superiority of a particular race
I haven't asked the camera but I don't believe it has such a belief.
Since when does all-in-one PC = robot?
Thank you for helping us help you help us all.