Some Eye-Popping Research From Siggraph
jamie found links to a discriminating selection of Siggraph papers at waxy.org. Among the more captivating: automatically improving the attractiveness of faces in portraits; automatic substitution of similar faces into photographs (with potential applications such as a privacy-enhanced Google Street View); and using still photographs to enhance video of a static scene.
I like sheep.
Those altered faces are somewhere between Bunraku puppet and a prosthetic hand
"Beauty is Symmetry, and you have none"
One of the main characters in the plastic surgery show Nip/Tuck made that comment. It seems as if TFA applies said comment.
All I got back was an email that read "ROTFLMAO!"
Just add symmetry and make thinner.
I wonder how soon they will be offering the "attractiveness improvement" service to the photos of their subscribers. I don't think they have enough CPU power to improve mine, though.
I rotated the pairs of adjusted faces so they were left to right (and the faces were on their sides), and defocused my eyes as if I was looking at a 3D stereo pair of pictures to see what would happen. The slight differences made the portraits appear to me as if they had been photographed in 3D. The places that had been changed were subtly evident as a misalignment -- in the eyes of some, for example. I realize this is a fudged 3D effect, but might there be some use for it?
---
I write pointed political and business short stories at http://klurgsheld.wordpress.com/
The first two are meh-worthy, but the last one approaches magic-grade technology. Wow!
+0 Meh
I'll never trust an image or video ever again. Never. Ever. Make sure you watch the "enhance video of a static scene" clip.
Decidedly a step up from the software their rivals wrote to make people look more like Steve Wozniak. *shiver*
Somebody should send Dr. Steinman over there, I hear he's been having symmetry issues.
The video enhanced by photographs was pretty amazing. NBC will probably be interested in getting their hands on this to "enhance" the next Olympic games.
the question is in twenty years time will you trust the news you see on TV?
when cheap, easy, video editing allows this then supposedly real footage: news, family videos, wedding snaps will lose all veracity.
after every girl wants to look good for her wedding...
and before somebody says "it will never happen" this is only a logical extension of red-eye removal.
Their Ugly!... Ugly!! UUUGGLY!!!!
I'd call this karmawhoring, but seeing as the editors didn't even bother linking to claimed list at 'waxy.org'... lists of Siggraph papers have been kept by Tim Rowley and Ke-Sen Huang for years. You can find this year's list at:
http://kesen.huang.googlepages.com/sig2008.html
And an overview of all years at:
http://kesen.huang.googlepages.com/
This also includes lists of papers presented at other events such as Eurographics.
For even more fun, visit the papers' authors sites; they often also publish papers at seemingly unrelated events that contain some interesting computer graphics gems.
The video enhancement was rather impressive, i wonder how cpu cycle intensive it is. I needses to get me handses on its
This will save porn companies a bundle...
It's all about the smile. The red-haired girl suddenly looks so much better when she's smiling.
As long as there are slaughterhouses, there will be battlefields.
... was digitally removed from the video.
In fact, it managed to get so different that got titled Clone Wars.Now you know why the president is so different in real life from what you see in TV.
that last link with the video was amazing. I hope it makes it into a commercial (or open source) product and isn't just some research that gets abandoned. Hopefully canon or adobe or someone will buy the technology.
I'm still waiting to be able to buy a plenoptic camera
Seems to me they could take this a step further and implement a sort of convolution matrix, but instead of modifying neighboring pixels, they'd be making subtle modifications to facial features.
dove into that link, 10Q berry much:-)
but i've not been able (ok, only 2 googlesworth;-) to decode this 3 letter acronym: NPR, as in NPR & deformations?
the blond and redhead - didnt they feature in a large set of "posture" images I came across once (ok bad pun there lol)
look familiar and the background matches IIRC
We took this spectacularly attractive blonde chick and wrote software that "automatically" turned her neutral facial expression (or slight scowl) into a smile. While we were at it we improved the symmetry of her face. Now she's gorgeous! We're brilliant!
But that's not all. We also took this African American guy and automatically plugged in a different African American guy's face that had the exact same pose. We're Gods I tell you!
These posts express my own personal views, not those of my employer
...if Captain Janeway was a true redhead.
n/t
This space available.
I could not decide, which versions looked better. I only recognized that they ere mostly non that good looking on both fotos.
Then I looked closer, because I know a bit about the methods behind it. And they did some big errors, like copying the one side of a face to the other, when the face did not look perfectly straight into the camera. This gave some weird results. Some faces even looked quite unnatural (especially, but not solely the focus on huge foreheads.
P.S.: I'm happy that I now since the last months know trough experience, that my opinion that looks matter, were wrong.
Any sufficiently advanced intelligence is indistinguishable from stupidity.
The first two with the faces...none of them looked right after being touched up. That's probably because our brains are really really good at processing faces. They looked wrong, and creepy in some way to me. Not impressed with the results at all.
The static scene improvements with photo's though. Wow. Fooled my eyes everytime. It always looked better and when they started swapping out elements, hiding posts so you could see the flowers behind, fixing the shaky video was the only thing that looked at little off.
Making faces more attractive is easy. All you have to do to get a reasonable increase is to make them more symmetrical.
If you want yet another increase, there is a set of ratios for distances between features that uncannily applies to pretty much everyone who is widely considered attractive. Shift everything closer to those ratios, and you'll get a big improvement.
Want more? Fix skin blemishes.
Between the three of those, you can make incredible strides. I would highly encourage any interested to watch "The Human Face".
Oh, you're not stuck, you're just unable to let go of the onion rings.
Now she can look as good the morning after as she did the night before!
Red to red, black to black. Switch it on, but stand well back.
automatically improving the attractiveness of faces
Beer does that already, why do we need an algorithm?
as the subject says, I found the female face more attractive but since I may own that probably doesn't mean anything. Now I know there aren't any females on Slashdot, but are there any fat geeks pretending to be girls that can say whether or not they found the male faces more attractive simply because they were "female".
It's about making faces more attractive automatically. Doing it manually is easier than writing a program that will do it for you
Video does lie now.
Can video ever be trusted again where evidence is concerned?
They're using their grammar skills there.
slashdot can add avatars!
If they shoot at you, you can be sure it's the enemy.
True, but as we saw recently with the anthrax researcher, your "enemy" doesn't necessary work for the other guys.
FYI, a close friend was a sniper for SpecOps.
And you're still alive? Whatever happened to "I'd tell you but then I'd have to kill you"?
There are two kinds of people: 1) those who start arrays with one and 1) those who start them with zero.
Once again someone's trying to write a bloated piece of software to overcharge for something our systems already do.
See the following example for how I was able to increase the attractiveness of an already attractive Hooters girl using only Microsoft Paint. (exported via Fireworks for filesize optimization)
http://img119.imageshack.us/img119/9474/hooters4si8.jpg
Wanna fight ? Bend over, stick your head up your ass, and fight for air.
A woman once said, while referring to the writing of Chinese characters, "symmetry is not beautiful." Incidentally, she had crooked teeth and was very cute.
if Google Street View replaced everyone's face with the face of Dave Grohl.
Feed in a head-n-shoulders photo and it will automagically turn you into a supermodel
Payware but not very expensive
http://www.portraitprofessional.com/
"Don't belong. Never join. Think for yourself. Peace." V.Stone, Microsoft Corporation
A couple of items of interest displayed at Siggraph this year as well which I think have potential.
Microsoft's come up with a way of painting objects onto an object extracted from a video, then reinserted to the video that remains accurate when the camera angle's changed. Their research paper's called Unwrap Mosaics, and you can see a video on Youtube here (higher quality video on the reseach page).
A company called Image Metrics have made a video with actress Emily O'Brien, using Light Stage technology from USC's Institute of Creative Technologies (an example of this is on a Google presentation called New Techniques for Rendering Human Performances) to create a realistic animated virtual face, that has convinced an editor on VFXWorld that they've passed the uncanny valley. Article is here.
I've been thinking that it was only a matter of time until editing video would become similar to editing photos, I just though it would take a lot more time, but everything is already here. They can even create realistic hair based on photos, just think what technology we'll have in the next decade, this could be in our homes by then.
http://media1.break.com/dnet/media/2008/2/14feb6-bitter-beer-face.jpg :)
3.243F6A8885A308D313
These imaging innovations ROCK. The technology presented on these tools, will be available in commercial apps in only a few years. Damn!
Yeah, the reflections are all wrong. Definitely photoshopped.
For a while now, I thought a good solution to the captcha problem would be to ask the user to select the "beautiful" image out of a group of faces. I supposed that this was something that couldn't be automated, but apparently that has been solved. Still, it could be extended by asking the user to identify the photo of the "kitten", or the "cute" one, etc.
Computers obey me.
CAPTCHA is broken anyhow as there is no way to mitigate against the "porn site" attack.
Me lost me cookie at the disco.