Slashdot Mirror
Are 625 Pixels Enough To Identify Sex?
mikejuk writes "A Spanish research team have patented a video camera and algorithm that can tell the difference between males and females based on just a 25x25 pixel image. This means that there is enough information in such low resolution images to do the job! They also demonstrate that an old AI method, linear discriminant analysis, is as good and sometimes better than more trendy methods such as Support Vector Machines..."
But it's not that small. At least that's what my girlfriend tells me.
can do it with fewer pixels.
...on what it's an image OF.
Am I the only person imagining genitalia icons?
What'dya mean there's no BLINK tag!?
"The also demonstrates that an old AI method, linear discriminant analysis, and demonstrates that it is as good and sometimes better than more trendy mehods such as Support Vector Machines"
I think the summary accidentally forgot the
Works on a 25x25 pixel image*
(* Pixels need to be a shade of pink)
I thought this story might be believable until I looked at the page. I'm not 100% sure what gender the 2nd row, 4th from the left person is and by the way, I'm a human. So I think the rest of the title to this story is "an arbitrarily acceptable percentage of the time so oh just publish it, it sounds neat"
CSI can do it with only ONE pixel!
Chaos maximizes locally around me.
...if the right body parts are in the image.
Warning: this article may contain humor, sarcasm, parody, and perhaps even irony. Read at your own risk.
Determine gender at what precision? TFA wasn't very enlightening... indeed, listing mis-identified faces doesn't really help much here.
This is like the problem of false positives in airport scans, but without the terrorists. :P
"We have to go forth and crush every world view that doesn't believe in tolerance and free speech." - David Brin
http://totallylookslike.icanhascheezburger.com/2010/02/04/justin-bieber-totally-looks-like-ellen-page/
indicates a woman with probability greater than 0.9.
0 = female
1 = male
How'd it categorize the lead singer of Tokio Hotel?
At least it works on tv
It's not like using linear discriminant analysis is some crazy or countercultural thing. It's a common simple technique. On some data it works well, and on such data, it's not uncommon to use it. It's particularly common in image-identification type tasks, and is one of the classic approaches to face recognition.
10 PRINT CHR$(205.5+RND(1)); : GOTO 10
that an application of a standard machine learning method can be patented. They have a publication in a good journal (PAMI), but there is nothing earth-shattering in the research. As far as the comparison with SVM is concerned, non-linear SVM does beat the linear methods when there is enough data (as they acknowledge in the paper).
And yet, after watching SNL in High Def, most people can't figure out Pat's gender.
How handy. I think that it was just this week that the DoD was looking for research aimed at assisting them in blowing up slightly fewer noncombatants based on lousy aerial footage...
I believe that chatroulette already solved this problem.
Quadratic programming with an RBF or Gaussian kernel should give you the best possible separation between any two classes by design, with sufficient amount of cross validation. Sadly, this doesn't always work in practice. I spent many months working on getting SVM to classify speech datasets, but the simpler methods always reigned. Not to mention, they take a fraction of the time to train a model.
I am guessing that the parameter tweaking required for SVM in some datasets is much more sensitive than others.
inb4, you don't know how to use svm
I can't remember where I saw it, so can't give you a link, but there's a video of two people in a completely dark room with small light sources at joints and extremities. The instant they start moving, you can tell which one's the man and which one's the woman.
The US Air Force recently launched a challenge for a system "that can determine approximate age (adult, teen, child) and gender of small groups of people at a distance.", with the goal of reducing civilian casualties during UAV operations. It shouldn't be too hard to make a system that can guess ages (at least well enough for their purposes), so the research team practically netted $20k already.
Additional news coverage: http://www.wired.com/dangerroom/2011/04/boy-from-girl/
625 pixels should be enough for anyone
Be gone from my sight or prepare to feel my flaming wraith!
The algorithm is also interesting in that it proves that an older and fundamental pattern recognition technique - linear discriminant analysis is just as good as the more trendy Support Vector Machines if used correctly and much more efficient.
A bit of clarity might be useful here. Support vector machines use linear discriminants as the central part of the algorithm. These linear discriminates -- simply hyperplanes separating two regions, are defined by a subset of the data points (called the support vectors). The other key part of an SVM is that it projects the data into a high-dimensional space in which hyperplanes can appear as curves or other shapes in the original space. This higher dimensional space is determined from the data using distances between the points in the data set (it's a kernel space).
The net result of all this is that SVMs are pretty much guaranteed to always perform better in terms of misclassification error than a simple linear discriminant, as every possible linear discriminant is considered in building the SVM. But it can be slower, and it can overfit.
So what's going on here? Linear discriminant analysis is an old statistical technique (1930s) that fixes a hyperplane based on distributional assumptions about the two classes. This allows the two classes to be plotted in a simple histogram by projecting them to the normal of this hyperplane, as shown in the picture in the article. It's used all over in statistics, and it works very well when dealing with two symmetric Gaussian distributions (that's what the theory assumes).
Thus the reason it works well here is that they've managed to transform their data in such a way that the two classes look like this sort of distribution. That's the insight here, not the choice of classifier. When the simplest model works, more complex techniques will overfit, meaning that you train on noise instead of the underlying structure of the data.
Does having a witty signature really indicate normality?
Here's what 25x25 pixel faces look like, using the example from the article: Picasa Web Albums - Paul Nickerson
I have a friend who took someone home only to find "it" wasn't the gender he was looking for! (I, on the other hand, know to pay particular attention to uh, other parts of the body... in particular the hands).
I can do it with 0 pixels and 50% accuracy.
(
(xkcd ftw again!)
My first Journal Entry ever, in 8 years! http://slashdot.org/journal/365947/aphelion-scifi-fantasy-horror-poetry-webzine
I think he means gender. Identifying sex is much, much harder. Are two people who are hugging having sex?
*** Don't be dull.***
And here I thought the title must relate to some kind of automated video analysis, you know, what-is-porn-what-is-not.
... Mick Dundee's method is more reliable.
Have gnu, will travel.
I would have thought you would need more than 625 pixels. Must have been some interesting research.
What? What do you mean: RTFA? I get all the information I need from the titles!
Maybe we'll finally answer the question in the skit theme song.
"Is it a man, or is it a woman? It's Pat!"
with any number of pixels. Color me a bit skeptical about this ... often when one looks at the training data used to train it and the test data used to test it, much is revealed about how it works.
)
Oh, what sad times are these when passing ruffians can open parenthesis at will on Internet forums.
How can I believe you when you tell me what I don't want to hear?
calling discriminant analysis an"old AI method" is like calling a typewriter "an old terminal".
Discriminant analysis was invented by Fisher and it is clearly a statistical method. The term AI would take another 20 years to be coined...
metageek
Fuck no they aren't enough pixes to detect this
I done my robe and wizard hat...
http://images.dailydawdle.com/pick-the-guy.jpg
Make each image 25x25 pixels and see if it works,
Don't fight for your country, if your country does not fight for you.
Aerosmith provides some commentary on the limitations of such a system...
http://www.youtube.com/watch?v=nf0oXY4nDxE
These things can never become truly 100% perfect as there's lots of people that will show up as statistical anomalies. There are for example people who suffer from hormonal imbalancies resulting in overly feminine looks in a male, or overly masculine looks in a female. Just as well transsexual people will be hard for these things: hormonal medication does not change skeletal features, but they change distribution of fat in the body, including face, and thus for a machine they'll like fall in the grey area between either gender. And how about intersexual people who are physically neither gender? I had a friend before who was IS and it just was really hard to tell from the looks what gender one should assume. Mentally she identified as female, but that can't obviously be told from a picture.
This also makes me wonder about the future.. I hope these "gender guessing machinery" do not become the norm in our society and public areas because they will lead to lots of issues with the aforementioned groups of people.
... 625 pixels should be enough for everyone !
Had think about this somehow http://xkcd.com/598/ .
Zooming zooming zooming.. Enhancing.. There we go. Its a .... a .... male pixel.
...but here in America we need at least 28 pixels LOL
Thank you. That would have bothered me all day otherwise.
sometimes it's hard to tell if that tranny is a guy or a girl. not that i have much experience with that, nosiree.
Remember kids, if you're not paying for the service, YOU ARE THE PRODUCT THAT IS BEING SOLD.
not the act of sex.
1 bit ...
0 = male
1 = female
Identifying sex could be a good thing.
But this was merely about identifying gender.
Are 625 pixel enough to get laid?
Zoom! Enhance!
... doesn't mean they need to deal exclusively with binaries.
TFA alludes to this issue with the "gallery of misidentifications", but doesn't get as far as asking surely the most important question: what exactly does this software claim to determine? It's clearly not "biological sex", because you can't determine that from a photo (even a full-body naked photo) -- what about the 1% of people who are born intersex? And it's definitely, definitely not gender, which you could only ascertain by asking that individual.
Whilst one potential application mentioned (analysing crowds for marketing purposes) seems vaguely sensible, any system using photos to try and identify an individual's gender is reactionary, oppressive and doomed to failure.
I can determine gender with just one or two digits, but I almost invariably get slapped for using this method.
I've abandoned my search for truth; now I'm just looking for some useful delusions.
Sex and gender are human constructs. Ask different doctors what constitutes male, female, and intersex, and they'll give you different answers. This is playing a guessing game at invented concepts. Can't wait for it to screw up and cause lawsuits.
Just wondering what does the alg say about Justin Beiber ?
Take any image, resize it to 25x25, and I can tell you without a doubt if the people in it are having sex.
You forgot the period!! I can't believe that you didn't see such a glaring mistake. ;^)
testing out my trending skills
Will it work in Thailand? I can see a market for it if it does.
Sure enough, the cow costume was hanging up next to the superhero outfit and sailors uniform. (S,Spud)
Today they are old-hat and -- by comparison with the current 'trendy' algorithms -- extremely limited.