Algorithms Can Make You Pretty 288
caffeinemessiah writes "The New York Times has an interesting story on a new algorithm by researchers from Tel Aviv University that modifies a facial picture of a person to conform to standards of attractiveness. Based on a digital library of pictures of people who have been judged 'attractive,' the algorithm finds the nearest match and modifies an input picture so it conforms to the 'attractive' person's proportions. The trick, however, is that the resultant pictures are still recognizable as the original person. Here's a quick link to a representative picture of the process. Note that this is a machine-learning approach to picture modification, not a characterization of beauty, and could just as easily be used to make a person less attractive." Note: As reader Trent Waddington points out, the underlying research was mentioned in an earlier story as well.
I wonder... (Score:5, Funny)
Re:I wonder... (Score:5, Funny)
..how this would handle a goatse pic.
About the same as 20 gallons of Preparation H.
Re: (Score:2)
Intresting, I hadn't seen them all. I'm amazed and disgusted at the same time and in some weird way as always question my sexuality due to the anxiety such images put in my brain.
Re:The question is... (Score:5, Funny)
I propose a contest, put goatse man and one of those women in a thunderdome, the winner is the one that leaves with the loser stuffed inside them
Re: (Score:2)
Re: (Score:2)
I really _didnt_ want to see that.
The many faces of GOATSE.
Re: (Score:3, Informative)
goatSHE was hot thou..
Ok, I'm sold (Score:5, Funny)
Now, which port on this computer do I stuff my wife in?
Re:Ok, I'm sold (Score:5, Funny)
Funny, she was just asked which port she is supposed to stuff her husband into...
Re: (Score:2)
Better yet, how do I get this algorithm installed in other people's brains? Though then everyone would look prettier, meaning no one would look prettier...
Re:Ok, I'm sold (Score:5, Funny)
Re: (Score:3, Interesting)
It just hit me. You could have a special pair of goggles, sort of like night vision goggles- there's a camera, it adjusts the image, and then broadcasts it onto screens in front of your eyes. But instead of amplifying the light, the camera would feed images to software, which would recognize faces, and then make them pretty using this algorithm. Making this work in real time might require some major advances in
Re: (Score:2)
My grandma used to say "Pretty is as pretty does".
What did she know?
USB (Score:2)
Now, which port on this computer do I stuff my wife in?
Most digital cameras connect to a PC using USB, presenting themselves as either Picture Transfer Protocol [wikipedia.org]) devices or mass storage devices. So make her look as pretty as you can with tasteful makeup, photograph her, and let the computer do the rest.
Re: (Score:2)
woosh
My point is that on MySpace and the like, nobody knows your spouse is ugly as long as you use a program like this. So making your spouse attractive through photo manipulation is almost as effective as making your spouse attractive through more drastic measures.
Oh great, just what the world needs. (Score:5, Insightful)
A way for people to fake their online photos in a way that when you finally meet them IRL you go ... AAAGH! What ... happened.. .to you... Car accident?
Re:Oh great, just what the world needs. (Score:5, Funny)
Apparently you've never heard of the MySpace Angles [officialda...source.com].
Re: (Score:2, Funny)
So, like, anyone got a link for the download?
Re:Oh great, just what the world needs. (Score:5, Funny)
My favorite line:
"A few inches maybe"?! How big are these faces? Jesus Christ, a few inches and I'd be able to sniff my eyelids.
Making pictures more attractive (Score:2)
Apparently one of the biggest things you can do to make someone's face more attractive is to mirror one side onto the other.
From the same picture, it looks like their algorithm decided not to do much mirroring. Looks like it made her eyes smaller and her face a little rounder. Plus wrinkle smoothing and adding some skin glow.
It didn't fix some asymmetries, like her nose.
Umm...no... (Score:3, Insightful)
Human faces are not symmetric, and our brains know that even if we don't.
Mirrored faces often seem grotesque. Or at least plastic-robotic.
Re:Umm...no... (Score:4, Interesting)
Mirrored yes. But more symmetric does appear more attractive.
From the sample it looks like their algorithm almost completely disregarded several opportunities for increasing symmetry and improved the face in other ways.
faceresearch.org (Score:5, Insightful)
personally, i find the faceresearch.org demo [faceresearch.org] posted on Slashdot a while back the most interesting. unlike this algorithm, it actually presents some interesting findings about the psychology of aesthetic beauty.
rather than manipulating a single photo to make that person more "pretty." it allows you to average different people's head shots. and the result of this research seems to show that our perception of beauty is based on the mean range of facial geometries we're exposed to. we naturally find faces that are the most "average" attractive. but different populations have different averages, so there are still cultural differences.
another way to look at it is that instead of looking for features that define beauty, we really just have an aversion to faces that deviate too much from the cultural norm as defined by the average range of facial configurations. now, everyone has unique features that distinguish them from others, and everyone deviates from the population average in some respect, but some show a greater deviation than others, which may indicate their genetic fitness. and so our psychological attraction to average faces is an evolutionarily learned trait to help us pick the most genetically healthy individuals to mate with.
but what's interesting is that if you mix several very different faces that don't meet conventional standards of beauty, you will actually get a very attractive face as a result (try this in the demo by picking the ugliest faces out of the gallery to mix). this is probably because even though "ugly" people deviate largely from the cultural average, they all deviate in different ways, so it doesn't take two beautiful individuals to produce an attractive average.
a corollary to this effect is that a couple with drastically different looks will give birth to very attractive children. which actually works out perfectly with another evolutionary trait--that of opposites attracting. human beings (and perhaps other mammals as well) are attracted to individuals with a very different histocompatibility index to themselves. that is to say, we are attracted to individuals which are very genetically different from ourselves. we can detect people's histocompatibility with our own based on their body scent. and double blind studies have found that men and women find the body odors of individuals whose Major Histocompatibility Complex (MHC) was the most different from their own. this is to ensure that their offspring will receive a diverse set of genes, which leads to a more robust immune system and prevents inbreeding.
now, my personal theory is, men and women don't just find partners with complementary MHCs to them based on scent alone. facial features can also be an indication of genetic differences. so this may also lead to individuals being attracted to people who have very different facial features from themselves. and since the average of two drastically different faces produces a more average face, this also leads to better looking children.
Well... (Score:4, Funny)
The one on the right was hotter, so I guess it works.
I put my picture in and nothing changed.
Re:Well... (Score:5, Funny)
I put my picture in and nothing changed.
Don't worry, it's just that it's still working on it.
Cultural bias? (Score:5, Insightful)
I suspect that's why they used two different countries for their data. It's funny just how horrified some people are by the idea of hardcoded behavior in humans. It's a fight that's pretty much over at this point, and the nature and nurture camps both had a lot right and wrong.
Golden Ratio? (Score:3, Insightful)
Of course, beauty is in the eye of the beholder and when you get old looks fade and all those cliched adages.
Precious Moments (Score:5, Interesting)
I wonder how much easily this could be accomplished just by enforcing the golden ratio on a face.
If you enforce the golden ratio too far, you get Precious Moments [about.com], where the eyes are a golden ratio down the face (resulting in a huge forehead) and the neck is a golden ratio up the body (resulting in difficulty putting on clothes).
Re: (Score:2)
The golden ratio is a fraud. You can use 5/3 for all intent or purpose as your personal golden ratio. The golden ratio only has one interesting property : If (a+b)/a = a/b, then a/b is the golden ratio
Re: (Score:2)
duped... (Score:5, Informative)
Here's the original [slashdot.org]...
Here's the source [tau.ac.il]...
Re:duped... (Score:5, Funny)
Yes it was duped but this new version is much better because it was fed through the algorithm. Good job slashdot!
Re: (Score:2)
I wonder how much of that is just because she was frowning in the original vs. smiling in the result.
It's inherantly flawed (Score:2, Interesting)
Re: (Score:3, Funny)
Fox news already does this (Score:5, Interesting)
Fox "News" already does this [mediamatters.org] when they're running stories about reporters from other news outlets.
--MarkusQ
Re:Fox news already does this (Score:4, Insightful)
and could just as easily be used to make a person less attractive.
Mod this guy up -- "dirty tricks" campaigning groups, foreign/domestic propaganda agencies, and disgruntled ex's will love to have something like this.
It allows the unskilled to dispense with the airbrush and photoshop skills, makes it easier and faster, and if the program is easily available publically, more deniable (for those who previously had the means to employ artists to do the job).
Re: (Score:2)
Mmmm. Petty, petty, petty. Reminds me of why I don't watch Fox news.
I actually think (Score:5, Interesting)
That the 'before' picture is much more appealing. She has nice eyes and an interesting, engaging face. She looks like someone who would be worth talking to.
The 'after' picture looks like a generic pretty-but-not-beautiful girl. She looks like she would be interested in shopping and hairstyles. The world would be very boring if everyone looked like that.
Re:I actually think (Score:5, Insightful)
I actually think... that the girl on the right is better looking. But the girl on the left would be more likely to hook up with most slashdotters. Therefore to most slashdotters, she will be more attractive.
Re:I actually think (Score:5, Funny)
She looks like she would be interested in shopping and hairstyles. The world would be very boring if everyone looked like that.
WHY DO YOU HATE AMERICA?!
Re:I actually think (Score:4, Insightful)
That is really why I hate America...
Re: (Score:2)
"That the 'before' picture is much more appealing. She has nice eyes and an interesting, engaging face. She looks like someone who would be worth talking to."
Yeah, I'd hit it too.
"The 'after' picture looks like a generic pretty-but-not-beautiful girl. She looks like she would be interested in shopping and hairstyles. The world would be very boring if everyone looked like that."
Too hawt, must therefore be superficial hence unlikely to throw sensitive me a shot of leg. Clearly her loss.
Re: (Score:2)
Seriously?
How can you judge someone just by her appearance like that?
I'm not here to argue which picture is prettier but I know for sure that you can't make
Re:I actually think (Score:5, Insightful)
While I agree that making guesses at someone's interests based on a headshot is superficial, we disagree on the definition of sexism. The poster was comparing two women, and not contemplating offering a job to either one (as far as I can tell).
Either that, or we disagree on the definition of women.
Re: (Score:2)
I don't think that sexism is the right word for it. I think that it is a logical fallacy due to not having met any engaging, attractive people. Maybe we could call it generically-attractivism.
Re:I actually think (Score:5, Insightful)
Although there is something intellectually repellant about it, you are very naive if you think that you don't, or can't, make certain judgments about people based upon appearance.
For one thing (and generalising horribly), from a biological point of view how attractive you find someone is likely to have some correlation to whether they are likely to be a good (i.e. successful) match for you, or for propagating your genes.
From a social point of view, the way someone looks and presents themself also communicates a large amount of information to you non-verbally. To me, the 'after' picture has the subtle look of someone who is attempting to present themself in conformity to a certain standard, which is not a standard I find particularly compelling.
Finally, is it impossible to think that someone who is conventionally pretty might be exposed to a different set of experiences to someone who is not regarded as such? This might natually have some impact on personality.
All of the above are generalisations. I totally agree that ideally one should not make snap judgments about people based on appearance. Nevertheless, I maintain that everyone does it, and that it is not entirely invalid (from a logical, not moral, point of view).
As for 'sexism', it is nothing of the sort. If it's anything, it's reverse discrimination against blandly pretty people, which is probably not all that high on the list of terrible things happening in the world today.
Re: (Score:2)
Re: (Score:2)
Personally I think their major problem is they didn't nail the eyes. The new eyes in every example look porcine.
It works for the developer's picture, but not for women whose eyes are already beautiful. Hmm, I wonder whose image he based his algorithm tweaks on?
I don't recognize her (Score:2, Insightful)
It doesn't look like the same person anymore, but a completely different person with a different face while keeping the same hair and clothes.
Just as easy, huh? (Score:3, Funny)
Note that this is a machine-learning approach to picture modification, not a characterization of beauty, and could just as easily be used to make a person less attractive.
Pfft, obviously this thing hasn't had to chew on my picture. It'd be a damn good algorithm that could find it's way out of this local attractiveness minimum.
Hulk not ugly! Hulk rugged! (Score:5, Funny)
I agree with Dr. Banner, and not just because I don't want to make him angry.
By who's standard (Score:3, Insightful)
Beauty is in the eye of the beholder blah blah blah..... I say beauty is cultural. The parameters by which program works are based on a elitist 'Hollywood' culture, the fact that a 'scientist' would prescribe to such unfair generalizations is offensive to me. Yea Yea demonstrating a concept blah blah blah.
Re: (Score:3, Insightful)
Feel free to hold whatever philosophy makes you the most comfortable and ignore the science.
I liked the woman on the left (original) better, but I am a statistical anomaly. That doesn't mean my opinion of beauty is any less valid, it just means my opinions aren't shared with the majority of human beings. The person's culture has less of an effect on a person's opinion of beauty than you claim is the point that the science is trying to prove.
Maybe I'm just weird... (Score:4, Interesting)
Anyone else notice the same thing? As well, did any women notice the differences in the men a lot easier than in the women?
Re: (Score:2)
Particularly with the women it seemed to make the eyes a lot smaller - much more so than the men. I notice it also straightened the mouth out, so that a slight up-turn of a faint smile came out looking rather dour and hard. I don't know about anyone else, but I think pretty much everyone shown looked better in the "before" pics...
Re: (Score:2, Informative)
There really was a lot more difference in most of the women's faces as compared to the men so it isn't just that you are more sensitive to women's faces. If you subtract the 2 images in GIMP you can more easily see what changed. Naomi Weinstein and Martina Eckstut had large changes to the size and location of their eyes. Alison Bruce who I found very pretty was basically unchanged. James Franco was also untouched and the only real change to Woody Allen was the frames on his glasses. Micheal Cera's eye
Re: (Score:2)
Re:Maybe I'm just weird... (Score:5, Funny)
Political ad applications abound (Score:2)
If you can go both positive and negative, and in a subtle way, you might see this applied to still photos (or perhaps video eventually) of the various candidates - a little bump up for yours, a little bump down for theirs. It already happens; this would just push the envelope a little farther. Most people will never see the candidates up close an personal, so it's likely to go mostly unnoticed if done well.
I am not impressed (Score:2)
In the sample picture, we see what I would consider to be an average "horse-faced" woman. In the other picture, we see what looks like the star actress of Terminator: The Sarah Conner Chronicles.
The second is most certainly NOT recognizable as the first. They share the same coloring, but that's about it. While I have found that beauty truly is in the eye of the beholder, I also find that facial recognition methods are in the mind of the beholder as well. Whether consciously or not, people vary from pers
Lame (Score:3, Insightful)
The brown-eyed girl looks plain now. Not ugly, just plain. The before picture had a more expressing face.
The pictures on this page (http://www.cs.tau.ac.il/~tommer/beautification2008/) are absolutely lame. The "before" pics had people in a neutral to a tiny bit of sad face(look at the lips). The new pics simple lift the corners of the lips and tada, better results. That's not better, that's cheap. Since the days of tell-sell I have realised that the before/after contruct was purely based on non-smiling/smiling people because it's that much of a change. This algorithm fails and should not be touted as the best thing since sliced bread.
Also, it makes Woody Allen look like someone who is 90.
I have a better algorythm, (Score:4, Funny)
It does much the same, but leaves the face alone and photoshops it onto an image taken from a fashion magazine.
It's still recognizable as the same person, but they look a lot better.
This technique is so powerful, that if you choose the right magazine (eg, Playboy, Hustler etc) that the test subjects don't even notice if you cut the original face out badly.
Three out of Four test subjects said "What Face" when asked about this irregularity and two left the test early with the new pictures, no doubt impressed by the quality of my algorythm.
GrpA
Re: (Score:2)
Ahh, the scientific "no-blind" study.
There is a simpler, proven method (Score:4, Insightful)
Re: (Score:3, Informative)
And here is the corresponding scientific study. [wkhealth.com]
Smiling (Score:2)
Did anyone else notice that in nearly all the exemplars, the algorithm tweaked the mouth so that it was smiling more?
What does that tell you about attractiveness?
Aikon-
Machine learning... (Score:3, Funny)
Obviously that's not needed around here.
Re: (Score:3, Funny)
Quite. The algorithm round here involves growing beards and buying all your tshirts from thinkgeek..
What About the Traditional Way? (Score:3, Funny)
Just gather massive amounts of wealth, and you're *always* attractive.
Except for the "gathering" part, it's so simple!
My head just exploded. (Score:5, Funny)
"Before and after of Alison Bruce. The software program is based on the responses of 68 men and women, age 25 to 40, from Israel and Germany, who viewed photographs of white male and female faces and picked the most attractive ones."
So a bunch of Germans and Jews got together and sorted through a bunch of people to determine which ones were better?
Wow. Just wow.
Re:My head just exploded. (Score:4, Interesting)
That was the subtlest Godwin I've ever seen. Alright everyone, thread over. Move along.
How to look better *without* a computer (Score:4, Insightful)
I've seen the full video and looked at the article from the SIGGRAPH materials. All of the "after" pictures except one did look more or less better than the "before" picture, but there was one consistent change I noticed -- many of the subjects, especially among the female photos, appeared to be frowning or pouting in the original picture, and the modified picture turned up the corners of the mouth into more of a smile.
This tells me that simply smiling can enhance one's attractiveness a great deal!
Re: (Score:3, Funny)
So that explains why I am smitten with The Joker!
Am I the only one... (Score:4, Insightful)
Portrait Professional (Score:2)
and if they smile? (Score:2)
This is really interesting software, but I think it makes a mistake in making the assumption that the one picture being made "prettier" will result in a prettier face if the subject smiles. Imagine each of these faces if they were smiling, the before and after.
Interesting followup research maybe.... though practical application... hmmmm.
Re:See what happens when you put Hillary Clinton's (Score:5, Funny)
Re: (Score:3, Interesting)
Hey! You stole my joke! I was thinking exactly the same thing. (GIRL: "You don't look like your photo." ME: "Yeah well it's a few years old.")
It's interesting that the "sample photo" really did not change much. It basically just shrunk the vertical resolution, thereby making her look rounder-headed, and more like a teenager. I could do that on paint very easily.
Re: (Score:3, Insightful)
I am not sure about Hillary (you could've provided a link), but if i have to go by the two pictures at the top of TFA, i'd say the algorithm isn't working very well for me. I find the original face more attractive than the result.
Maybe the algorithm works to tune an image to _someone's_ preferences, but those are different than mine. That is, beauty is still in the eye of the beholder.
What else is new?
links as requested (Score:5, Funny)
I am not sure about Hillary (you could've provided a link)
Hillary Clinton was a candidate in the U.S. Presidential election; here is some information about her [wikipedia.org] and here is her website [hillaryclinton.com]. She was also married to a former president of the U.S. Let me know if you need links or information about him too.
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
I think it would be interesting to see what the algorithm does to somebody like Halle Berry or Jessica Alba. What happens when you try to beautify somebody who might already be considered perfect-looking?
Probably almost nothing happens, or if it does, it results an a different generically pleasing set of proportions.
What happens if you try the algorithm on somebody who is actually in the algorithm's database?
Sheesh, I can answer that just from the /. article. It finds the closest match in the database, so it would likely find them, and then do almost nothing.
Re: (Score:3, Interesting)
I dunno, I think the second was more attractive, due mostly I think due to the cleanup of a few blemishes and skin tone enhancement (but that was probably poor lighting in the original, anyway)
On the other hand.. The second picture looks a lot more like Summer Glau than it does like the first picture.
I'm Already Pretty (Score:4, Insightful)
Seriously. Like, Chris Isaak, with Liz Taylor's eyes. But, without my overbite, I'd look dull.
So, there's software to make faces bland and uninteresting - go figure. I like the "before" picture girl - with the giant eyes, and super-sized mouth. Sensual, and sensitive. Those are attributes the "beautifying" stripped away...
Re: (Score:3)
I find uniqueness to be very cute and sexy on women. Big noses and glasses in particular
Re:I'm Already Pretty (Score:4, Insightful)
Algorithms can model the lowest-common denominator of attractiveness, as determined by our sample respondents.
or
Beauty by committee.
Re: (Score:3, Funny)
Beauty by committee.
If one doesn't believe that our finding markers of (relative) fertility beautiful is due to evolutionary pressures, this might be... wait for it...
Intelligent Design by Comittee.
Re:I'm Already Pretty (Score:5, Funny)
"Agreed, not all of us want our women to look like bobbleheaded dolls."
Performance, OTOH...
Re:I'm Already Pretty (Score:4, Funny)
I find uniqueness to be very cute and sexy on women. Big noses and glasses in particular ;)
I prefer big glasses and cat ears. Nyaaaaaaa! =^_^=
Re: (Score:2)
Re: (Score:3, Informative)
Some call a single imperfection on the face, like a mole, a 'beauty mark'. I don't subscribe to that myself, but it does make a face a bit more interesting.
I also prefer the 'after' picture. She not ugly beforehand, but I find the 'after' picture to be prettier.
Re:I'm Already Pretty (Score:5, Insightful)
yea, this research is completely useless. the only potential application i can see for this is to sell software to insecure individuals with low self-esteem so that they can hide behind altered photos of themselves online, further reinforcing their negative self image.
honestly, this program embodies what is most wrong with modern western culture--superficiality, vanity, and an abhorrence of eccentricity or individuality.
abhorrence of eccentricity or individuality. (Score:2)
"...abhorrence of eccentricity or individuality...
Unfortunately, I think this is perennial, not modern. But yeah - you're right. The technology of "communications" today means a more efficacious way of speedily distributing that mean towards its baseline...
Re: (Score:3, Insightful)
You're right - we should ban this type of research. After all, any avenues of research into machine learning algorithms or computer imagery that caused this abortion of a program to be created should be abandoned for all time and sealed up in that place where they put the Ark of the Covenant.
Or maybe this was just an example application to demonstrate their research. Your call!
Re: (Score:3, Insightful)
Other applications would include seeding several instances of the program with different cultures' opinions of beautiful photos and then comparing the results. Finding patterns in what people consider beautiful cou
Re:Fuck Israel (Score:5, Funny)
Re: (Score:2)
"Beauty is the person shining from within."
That's why it wouldn't alter a Goatse receiver pic.
Re: (Score:2)
Then again, my husband and I often disagree on which women are prettiest, and he claims that it's because men and women have different standards, so maybe I'm just looking at her with the wrong set of chromosomes.