Are 625 Pixels Enough To Identify Sex? 143
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..."
Depends... (Score:5, Funny)
...on what it's an image OF.
Am I the only person imagining genitalia icons?
Re:Depends... (Score:5, Funny)
Am I the only person imagining genitalia icons?
Yes.
Re: (Score:1)
says the person with no imagination, no genitalia, or porn on the go.
Re: (Score:1)
Somebody notify the IOC!
No need for any fancy genetic testing anymore.
Re: (Score:2)
Re: (Score:3)
Wait till you see what happens when I issue the command ENHANCE, 10x
Re: (Score:1)
I'll tell you about my father, mother fucker.
Hm? (Score:5, Funny)
I think the summary accidentally forgot the
Re:Hm? (Score:5, Funny)
Re:forgot (Score:2, Funny)
(
(xkcd ftw again!)
Re: (Score:3)
)
Oh, what sad times are these when passing ruffians can open parenthesis at will on Internet forums.
Re: (Score:2)
Re: (Score:1)
Re: (Score:2)
Re: (Score:1)
How long does it take to get a 30 megapixel image from the 25x25 with CSI processing?
CSI Software runs at... the Speed of Plot [tvtropes.org] .
*sunglasses* YEEAAAHHH!
Re: (Score:2)
Pfft. With CSI processing, you can take a picture of a spoon with a one-pixel camera, and get your detailed, holographic image of the moon from the reflection off the spoon....
Re: (Score:2)
Re: (Score:2)
And the drawer is inside a building ON the Moon.
(OK, it's an equipment drawer on one of the lunar landers. Happy now?)
Let Me Fix That! (Score:2)
You forgot the period!! I can't believe that you didn't see such a glaring mistake. ;^)
Footnote (Score:5, Funny)
Works on a 25x25 pixel image*
(* Pixels need to be a shade of pink)
Re: (Score:2)
http://en.wikipedia.org/wiki/List_of_Better_Off_Ted_episodes [wikipedia.org] (See "Racial Sensitivity" episode).
Re: (Score:2)
Don't go reading too much into that statement, though.
not believable (Score:1)
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"
Re: (Score:2)
The images shown were ones the software had trouble classifying (It got them wrong). The top row are male, the bottom row are female. This is all explained in the caption.
Interestingly the article does not make any mention of error rates. And I couldn't find anything easily on either site it links to.
Re: (Score:1)
... This is all explained in the caption.
And I'm laughing my ass off. I knew /. "readers" rarely RTFA, but I always assumed they were just too lazy to follow the links. I never realized they actually click the links, but only to look at the pretty pictures. xD
Re: (Score:1)
Re: (Score:2)
What's worst is that the small photos on that page are 50x50 pixels, not 25x25.
Ha! (Score:5, Funny)
CSI can do it with only ONE pixel!
Re: (Score:1)
ENHANCE!
Re: (Score:2)
Only after they hit the Enhance button.
Re: (Score:3)
Re: (Score:2, Funny)
I can do it with ZERO bits.... with 50% accuracy!
Re: (Score:2)
Not true. There is normally a reflection in that one pixel of entire scenes of crime.
Re: (Score:1)
Determining sex should be easy... (Score:2)
...if the right body parts are in the image.
Re: (Score:2)
Not so fast, mate! There are those where some parts of the body (right?) are a bad indication of what to expect several inches further up (or down).
Make it 2x25x25 for the better.
Re: (Score:1)
May I now present transgendered people including intersexed individuals?
Re: (Score:2)
Insufficient information. (Score:3)
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
Re: (Score:3)
The article has a histogram that shows how sure the algorithm was of its predictions for both sexes. Males on the left of 0 were misclassified, and vice versa for females.
Now, the only confusing this is if that plot is for the test set of the train set. If it is for the test set then it answers your question. If it is for the train set it tells us a lot less. Pretty sloppy of them to title a graph with both :(
Can it really? (Score:2, Insightful)
http://totallylookslike.icanhascheezburger.com/2010/02/04/justin-bieber-totally-looks-like-ellen-page/ [icanhascheezburger.com]
a non-white full headscarf surrounding the face (Score:1)
indicates a woman with probability greater than 0.9.
1 bit should be enough (Score:2)
0 = female
1 = male
Re: (Score:3)
Re: (Score:1)
Hi. I'm transgendered, and your database schema has a shortcoming. :)
Re: (Score:2)
The Kaulitz Test (Score:1)
Re: (Score:2)
Plastic.
Re: (Score:2)
ternary logic
Enhance! (Score:2)
weird positioning on LDA (Score:3)
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.
Re: (Score:2)
That's right. Moreover, in the paper they say that for larger data set SVM with an RBF kernel performs best.
it is puzzling (Score:3)
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).
How convenient... (Score:2)
The thing about SVM.. (Score:1)
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
Humans can do it with 12 points of light (Score:2)
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.
Re: (Score:2)
No, the key word here is points. For a single frame, you only need the x and y coordinates of 12 points.
The Air Force will like this (Score:2)
Additional news coverage: http://www.wired.com/dangerroom/2011/04/boy-from-girl/ [wired.com]
Like the man said... (Score:1)
Re: (Score:2)
Re: (Score:1)
of course, coming from people who call themselves "micro" and "soft".
SVMs vs. LDA (Score:5, Informative)
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.
Re: (Score:1)
Re: (Score:2)
There are quite a few people who don't know anything about it. They don't know what an SVM is, nor what differentiates it from linear separation (aka Perceptrons). So any explanation is more than welcome, and the GP got rightly modded up. Perhaps an even more *obvious* explanation is needed. Why don't you write one?
Re: (Score:1)
Thanks for the explanation! I would mod your post up if I had points.
What It Looks Like (Score:3)
Re: (Score:2)
But.. (Score:2)
Re: (Score:2)
"Gender" is a grammatical term. "Sex" is a biological term. So yes, he did mean sex.
Are 625 Pixels Enough To Identify Sex? (Score:2)
I believe ... (Score:2)
Low bandwidth porn (Score:2)
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!
Saturday Night Live: (Score:2)
Maybe we'll finally answer the question in the skit theme song.
"Is it a man, or is it a woman? It's Pat!"
Re: (Score:2)
A few times a year I see a person who I can't readily determine the gender of. I'd like to see if this algorithm can teach me a thing or two (I won't be so crass as to photograph the person and run PatApp on the image).
Bet it can't tell a dog from a cat (Score:2)
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.
not AI at all... (Score:1)
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...
Re: (Score:2)
Comment removed (Score:3)
Re: (Score:2)
Even so quite a few transsexuals undergo various forms of plastic surgery that certainly can change a lot of skeletal features. This can range from rhinoplasty, forehead contouring, chin reductions etc...
Re: (Score:2)
Re:Statistical anomalies (Score:4, Insightful)
It really differs among IS people. I am a hermaphrodite yet there is no way to tell this while I'm still wearing clothes. Everyone identifies me as being a regular female, even at the swimming pool. There are heaps of 'regular' women who would get IDed by this system as being men, making it inaccurate for regular men and women, and a huge mess for IS people. As for TS people, most MtF TSs I have seen would be identified as being male, and most FtM TSs as being female. As said, unless you are going to modify the skeletal structure of the face etc. taking hormones doesn't magically transform you into the other gender/sex.
Xkcd ruined my life (Score:1)
Had think about this somehow http://xkcd.com/598/ [xkcd.com] .
people aren't that easy to figure out (Score:2)
sometimes it's hard to tell if that tranny is a guy or a girl. not that i have much experience with that, nosiree.
I am disappointed (Score:2)
Identifying sex could be a good thing.
But this was merely about identifying gender.
Obligatory (Score:1)
Zoom! Enhance!
Re: (Score:2)
Just because the researchers are programmers... (Score:1)
... 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 [isna.org]? And it's definitely, definitely not gender, which you could only ascertain by asking that indivi
Don't need that many bits (Score:3)
Yes. (Score:2)
Take any image, resize it to 25x25, and I can tell you without a doubt if the people in it are having sex.
Ladyboys (Score:1)
Re: (Score:2, Funny)
Here you go:
Male: |
Female: O
Re:I am pretty sure that I... (Score:5, Funny)
...so can anybody from the old BBS days.
Re: (Score:1)
Re: (Score:1)
Re: (Score:2)
telnet bbs.buanzo.org i still have my own bbs running developed in linux, using kernels 1.x to 2.x, starting at libc5.
Those are far more characters than needed to identify you as male.
Re: (Score:1)
Re: (Score:3)
can do it with fewer pixels.
Dunno about this. I've seen some people in the full 3d glory of real life that I could not discern the gender of.
Re: (Score:1)
If you can't tell that those two very obvious women are female, then your Republicanism has poisoned your sex senses.
Re: (Score:1)
also it's not gay if he's hot