Algorithms Determine Mona Lisa's True Emotions

caffeinemessiah writes "The BBC reports that researchers at UIUC and the University of Amsterdam, Holland have used "emotion recognition" software to determine Mona Lisa's true emotions. The algorithm is based on a library of neutral face images of young women and determined that Mona Lisa was 83% happy and 9% disgusted." From the article: "The program, developed with researchers at the University of Illinois, US, draws on a database of young female faces to derive an average 'neutral' expression. The software uses this average expression as the standard for comparisons. The New Scientist says that software capable of recognising emotions just by looking at photographs could lead to PCs that adjust their response depending on the user's mood. "

How this probably works ...

[eldavojohn]

You can find corpuses [face-rec.org] of human faces taken with different emotions displayed.

Once you've either collected them yourself or downloaded them, you need to use a process called eigenanalysis which is basically fancy talk for analyzing a large dataset with multiple classes (emotions) using matrix decomposition.

I've actually worked on many projects involving this and the result is an eigenface (or eigenmask) [mit.edu] that allows you to transform the space that the original image is in and classify it using any of a number of algoirthms that use euclidean distance.

I know I left out a lot but there are many papers out there that you can find on citeseer [psu.edu] and white papers floating around out there [ucsb.edu] that provide a lot of reading material on this.

There are also strategies which require tagging certain features as points on the face (like corners of eyes, corners of mouth, center of eye, etc) and then using the relative distances between all these points to determine what classification you would give a new face. The problem with this is that it requires a lot of hand work to prepare the training set.

Hope this helps anyone who wants to learn more about the actual process used to accomplish this recognition.

I'm.....

[Anonymous Coward]

22% hungry and 88% constipated

Re:You know

[greenguy]

But he wouldn't have had much interest in her (except as a model he was commissioned to paint), as he was gay.

As for the article... I think these folks just need a hobby.

Re:And...

[Flashbck]

And to modify your sig: Those who can't RTFA, complain about stupid crud

Quote from the third paragraph in the article:
It concluded that the subject was 83% happy, 9% disgusted, 6% fearful and 2% angry, New Scientist magazine was told.

Re:You know

[Surt]

Documentation for that claim:

http://www.bbc.co.uk/science/leonardo/gallery/mona lisa.shtml [bbc.co.uk]

http://en.wikipedia.org/wiki/Mona_Lisa [wikipedia.org]

Finally!

[anocelot]

Interesting to see this idea actually working now. I think I first saw this five years ago on IBM's Alphaworks site. Ah yes, here it is.

http://www.research.ibm.com/journal/sj/393/part2/p icard.html [ibm.com]

Re:And...

[brjndr]

Actually, their determination was that the smile is "83 percent happy, 9 percent disgusted, 6 percent fearful, 2 percent angry, less than 1 percent neutral, and not surprised at all."

Re:You know

[Darkman, Walkin Dude]

Here [about.com] is an expansion on that...

What is absolutely untrue is the reference in The Da Vinci Code to Leonardo's reputation as a "flamboyant homosexual". He was not known as such. Historical evidence is sketchy about the latter, and the only thing Leonardo was "flamboyant" about was his inability to finish projects he started.

Re:So...

[Anonymous Coward]

you know. this is kind of disrespectful. saying that the monalisa had gas. how do you know that, really how? listen man, i am an anthropologist, and ill tell you something----no one ever wrote about gas in those days (during the renaissance) and no one started writing about gas until the late 19th century, approximately 1985. this quite possibly means that people didn't have gas back then because their diets were so hoarsely different. for instance, no one used to eat sodium sorbate but now everyone puts it in preservatives in the food that you and i are consume every day. heck, some people even use it to season their food. and you know something, it's a preservative and keeps stuff from going fowl and therefore makes farts stale. something or something else probably in the sodium absorbate actually causes farts so you know now that she ws not holding it in. if she was, then good for her to have manners because people know that girls avoid farting by never eating. if they never eat, how can they flasturate? seriously. anyway. be careful how you say stuff because you dont know how experts are going to come in and prove you wrong. also, as a seasoned historian, i can tell you that what i said is correct. be careful how you say stuff! or else you will be proven wrong by an expert who knows what and how he is talking about just like me and my colleagues who wrote up this article. have a nice christmas---youll need one

Re:Facial expressions are nature, not nurture

[natpoor]

This is indeed true for some expressions. I just finished teaching a class in nonverbal communication, and for part of it we looked at facial expressions. Researchers have identified 6-8 facial expressions that are the same around the world (currently), and so these are believed to be nature/hardwired into our brains. Ekman and Freisen did a lot of work on this. Here is a Google search [google.com] that has some material. However it is not always true, smiling in different cultures can mean slightly different things. As for across history, well, the ancient Greeks used those smile/frown masks for drama, so I would assume the hard wired ones have been around a long time (since evolution takes place very, very slowly).

Maybe someone else pointed this out but the computer isn't really determining her mood, the people who judged all the other faces and constructed the algorithm did it, the computer just did the calculations. But Slashdotters knew that.