Women's Attractiveness Judged by Software 348
Roland Piquepaille writes "According to Haaretz, an Israeli team of computer scientists has developed software that ranks facial attractiveness of women. Instead of identifying basic facial characteristics, this software has been designed to make aesthetic judgments — after training. The lead researcher said this program 'constitutes a substantial advance in the development of artificial intelligence.' It is interesting to note that the researchers focused on women only. Apparently, men' faces are more difficult to grade."
more average is more attractive (Score:5, Interesting)
Even beyond that... (Score:5, Interesting)
In the end, it all comes down to individual perception. Sit ten guys down with thirty pictures, and you're going to get 10 different #1's. Maybe you can teach a program to be able to say who it thinks is hot, whatever use that is. Or you could write a program that would allow a person to rate a hundred or so pictures, so that you could run a dating service that automatically pairs you up with people it thinks you'll find attractive...That's the only use I can come up with.
Symmetry (Score:5, Interesting)
I read an article a while back that made the point that one of the biggest factors in attractiveness was symmetry. The "perfect" face doesn't have any features out of alignment. There was another study that made the point that "averaging" faces produced more attractiveness, but this was actually the wrong conclusion. It was the averaging process that smoothed out features into perfect alignment.
Symmetry actually makes sense. The more messed up someone's face is from ideal, the worse their genetics could be. Of course, there are other factors such as shiny hair, clear skin, sharp cheekbones, fitness, which all factor back to health.
Re:requires external criteria (Score:2, Interesting)
Skin smoothness (Score:5, Interesting)
Re:Even beyond that... (Score:3, Interesting)
We knew that (Score:2, Interesting)
Re:requires external criteria (Score:5, Interesting)
Re:Original Paper & Obvious Criticisms (Score:3, Interesting)
Also noteworthy was that none of the top faces were celebrities. Oh, and the top face was absolutely breathtaking. I mean impossibly beautiful. Several times over the years I've poked around on the internet trying to find it. I remember at the time suspecting it was photoshopped to be perfectly symmetrical, but it was more than that. This face was otherworldly - and to have such an astonishingly perfect face not belong to a celebrity? Weird. Could be some model I'm just unaware of though I guess.
Something ommitted (Score:5, Interesting)
An intelligent woman looks highly attractive when confused...you can almost see the gears working in her head, trying to figure it out. An unintelligent woman just has a dumb confused look on her face.
From what I have seen, intelligent women tend to not necessarily have more attractive facial features, but a more attractive way of showing their emotion and reaction to things. Not something that is commonly thought about.
Re:more average is more attractive (Score:1, Interesting)
It's hard to do that right. It would be interesting to see what would happen if you averaged it with the mirror image.
Re:requires external criteria (Score:4, Interesting)
I'm not an expert of male attractiveness, nor do I play one on Slashdot, but I imagine similar factors (absence of damage, proportionality of features, symmetry on the vertical plane, etc.) would play a similar enough role. That said, there have been plenty of studies showing that the "attractiveness" of a male's face corresponds to the menstrual cycle of the female: during ovulation, the "rugged and handsome" look is preferrable to "nice and well-shaved" whereas the inverse is generally true at other times.
As for "wealthy and powerful", I guess that could similarly depend on the financial and social status of the female. I prefer to consider it a truism in the same way that in the wild, it's typically the biggest, strongest, or the one with the most goodies that gets the opportunity to mate.
A side note for anyone cherishing the notion that everything is relative or personal, and there can be no standard of attractiveness. Even across disparate cultures where such things can run into the extremes, the attractiveness value of facial symmetry, to take one example, remains universal. I remember a PBS program on the subject years back that examined the faces of famous movie stars. Turns out by taking a ruler to the face of someone like Brad Pitt or Angelina Jolie, you'll discover both have nearly perfectly symmetrical faces. I imagine one could conclude there's some form of Golden Ratio [wikipedia.org] that applies, particularly to body shapes like those of Angeline Jolie.
Re:Even beyond that... (Score:3, Interesting)
You may be interested in reading about the "evolutionarily deceptive" teenage years, where soon-to-be-ugly people appear attractive to seduce a mate, and soon-to-be-beautiful people repel mates so as to avoid the wrong one.
Re:Woman scientists will retaliate... (Score:5, Interesting)
It takes time to build a friendship, then a sexual interest, then a romantic relationship-- the guys I know who have stable relationships do not let their women hang around alone with other men. When their feelers go up they chase the guy off- usually just by saying very mildly that they are not comfortable with the situation.
In my case, they had a business relationship first.. then started meeting for lunch-- then I stopped being copied on emails-- then his wife stopped being copied on emails-- then they acknowledged something deeper than friendship (this is about 5 months in-- like I said, I got to read every email from both of them once I stopped the blind trust thing), then they started having sex-- then she had some work conferences (i.e. trips to his ranch)-- (now we are at 7 months)-- then they had a full out affair- he dropped hundreds if not thousands on jewelry, flowers, dinners and THEN he found out about me (she told him there was no one else-- he was married so there was no conflict right away) -- THEN she fought him for three months to keep both of us-- finally she told me and tried to keep both of us but as you would imagine, she was 60% him / 40% me and sliding towards him by then or else she would have cut him off instead of trying to keep both of us.
You are right- we are all individuals who control our own destinies and we have no strong control others. But it take time to slide from loving someone to being willing to lie and betray them. If you catch them early, then you can stop things before they are too far along.
If he had not been so damn wealthy I do not think it would have been an issue. It was like the second sentence out of her mouth when she broke the news to me. The universe had sent a wealthy man to take care of her. It was right after she said crying that she had had an affair with someone and she didn't want to lose me.
And I make a good income and wasn't stingy on sharing it and had proposed. She was gloriously happy while at the same time she was being a complete skank. She and he started out with the idea that it would be a discrete little side thing that they would do during the day and "no one would get hurt". His family is hurt... I'm torn all to hell... his wife is hurt. The two of them lied to everyone. I damn near had a nervous breakdown over it because there was almost no warning. I knew she was under stress and consoled her and she told me it was about her business- I trusted her completely at that point. The stress was apparently really that she was fighting with him to keep it all secret and under wraps.
I wouldn't be posting but she tried to open up contact with me again last week after I had successfully ended contact with her for several weeks and that attempt opened up all the pain again.
Re:Even beyond that... (Score:3, Interesting)
What would be useful for a computer dating service would be for you to rate each girl as you see them in terms of attractiveness, similar interests, etc. and use an algorithm like this to then filter out women that you probably wouldn't be interested in. Since each person's definition of beauty differs, it really needs to learn an individual's preference. Ideally, this could be combined with latent semantic analysis of the text that the potential dates typed as part of their profile to further improve matching accuracy. Man, I should totally design and patent that.... That and five bucks would buy me a cup of coffee at Starbucks.
very linear perception of beauty (Score:2, Interesting)
Until she starts acting "cute".
Beauty and attractiveness are somewhat separate concepts, but, as my mother used to say, "Beauty is as beauty does."
Look at Anni-Frid Lyngstad and Agnetha Fältskog, or, for that matter, the Wilson sisters. I don't know about you, but I didn't think they were either attractive or beautiful until I'd been listening to their music for several years.
Or, for that matter, look at the Mona Lisa. I'm still not sure what the fuss is there.
Any way, happy Aril 1st.
Re:Original Paper & Obvious Criticisms (Score:3, Interesting)
In the first stage, 30 human participants were asked to rate from 1-7 the beauty of several dozen pictures.
For a masters project (which this was), that's a decent sample size. For research and practice, I do not think that will suffice. Why don't they tell us how this scored some celebrities from around the world like say Iman Abdulmajid, Zsa Zsa Gabor & Angelina Jolie? I have a feeling that their system is over-trained and would perform poorly in real life. Facial beauty requires imagination and this system was hand trained on a hundred points. I don't think that's enough but I wish they would have published more results to either prove or disprove my criticisms.
The number of participants in user studies are usually pretty low. A 30 person sample size is actually pretty good. It would have been better if the number of participants exceeded the number of items being rated though. That would have made this project better. A simple case would simply have been to enlist a bunch of undergrads from some classes. Double bonus points if he got their participation in the project mandatory. That's a common technique in psych departments to get subjects.
.82 with the human judges. This corresponds to a MSE of .39, which is very good. In fact it's significantly better than the previous study this work appears to be based on, which only achieved a correlation of .6 on a similar dataset. So there goes you're argument that this isn't noteworthy.
I can tell you why they didn't use celebrities. It's completely irrelevant. First off the training data consisted of 91 caucasian (i.e. european descent, i.e. "white") women. Any results with Iman would be completely spurious since she's african (i.e. "black"). Non-whites simply don't exist in the world of machine rater. Her rating would be dependent on what weights would be assigned to skin and hair color. I would suspect that since her skin color lies outside the range of what society defines as "white," this extreme variance would either strong rate her as attractive or unattractive. Not in the middle. But as I said, it doesn't matter since she's not of the population the training set samples. Second, the faces used in the study are all neutral expression, full frontal, under controlled lighting with no makeup and jewelry. Try finding any celebrity photo like that. Third, what is the point of using celebrities? What does that give you that a random sample doesn't? The only reason why I would think that you would mention celebrities is that you believe that somehow they represent a population of highly attractive people. That's a false premise. There are plenty of unattractive celebrities, and there's celebrities that are allegedly attractive but really aren't. Case in point: Angelina Jolie. I find her absolutely hideous. She is the ugliest woman that the media tries desperately to convince me that is attractive. Sorry. No. On the likert used in this study, she's a 2. 3 at best. Now even if I am a freak, and you're implicit assertion that celebrities have above average attractiveness even under controlled conditions, what's the point of using that for testing? The system would need to learn nothing to test well on that data set. Hell, this would be all you'd need to perform well:
int ratePhoto(Photo *p) { return 7; }
Not very interesting is it?
You're also assuming that there is some sort of objective analysis here, and there clearly isn't. All you can do is measure how the system performed to the human judges. In this case, the system. The system had a Pearson's correlation of
This leads us to the discussion of why they only used a 91 image dataset. First off you're limited to what data sets are available. These datasets have apparently been used repeatedly throughout the community so apparently they're a standard set for eva
Re:very linear perception of beauty (Score:3, Interesting)
Re:Woman scientists will retaliate... (Score:3, Interesting)
Unless there's something about this guy that you don't know or aren't telling us, your ex is a gold-digger and that spiritual connection you felt was you being a hopeless romantic idealist. It's easier to write off half the human population as lying, treacherous harlots than to admit that you might have been wrong about her specifically, but it'll ruin your life more than she ever could have without your help.
Whilst I doubt your situation is unique, I can assure you it's not universal, and hopefully caution you from overextrapolating from it. Whilst I've had three partners (that I know of) cheat on me, admittedly done the same to one of them when I was much younger, and been presented with one or two opportunities to be 'the other guy' in ostensibly monogamous relationships (none of which I went through with), not a single one of those situations has never had anything to do with money.
I'm making this worse, aren't I?
The point is, there are more reasons for relationships beginning and ending, and for those beginnings and endings overlapping, than you can possibly keep track of without ending up not only alone, but alone with one or more A.V.O. [wikipedia.org]s against your name. I'm great friends with two of those three ex-partners, and the one I don't talk to is for entirely separate reasons. Turning bitter and cynical, or paranoid and stalkerish, isn't going to help your cause.
There's someone better than you, or me, at just about anything. If you make the game about money, and attract a partner who thinks the same thing, then there's going to be someone with more money, and you know what happens then. If you make it purely about physical attraction, there's going to be someone coming along who's taller or handsomer or cuddlier or whatever it takes to catch your significant other's eye. If you pick a specific criterion or criteria like that, you're going to lose. She might cheat on you, she might have the decency to dump you first, or she might stay faithful and spend the rest of her life quietly and unhappily imagining what could have been, but none of those sound especially appealing to me.
This bit is going to sound like self-help wank, but the only thing you can do better than anyone else is be YOU. It's up to you to make sure that's a good thing (i.e. staying healthy, not living in Mom's basement, soaking food stains etc.). It's up to you to make sure that you're actually putting some effort into it and allowing yourself to evolve (i.e. working towards the job you want, allowing yourself creative outputs, growing past OS fanboyism). It's up to you to find someone whose tastes intersect significantly (maybe even perfectly) with what you are, and what you're happy being, and who intersects similarly with your tastes.
Hell, I know furries who are in happy decade-old relationships; trust me, there's someone for *everyone*.