It's real. Image completion has been an active part of the computer vision and computer graphics communities for a decade now. I'm glad to see Adobe is finally implementing this. They have good researchers working for them so it was only a matter of time. Microsoft even had a version of this in their Digital Image Pro software. It would be nice if they attributed the ideas, but that's not as flashy I suppose.

I would say that what Adobe has cooked up here is quite well engineered, but not fundamentally different from the things that have been in the literature for 6 or 7 years. I would also guess that they've chosen their examples well, and you'll be disappointed when you try it in other situations.

Here's some of the examples of this from the literature:
1999: http://graphics.cs.cmu.edu/people/efros/research/EfrosLeung.html (scroll to the bottom, this is the one that started the texture / image completion craze in the vision / graphics communities)
2004: http://research.microsoft.com/apps/pubs/default.aspx?id=67276 (more sophisticated heuristics about how to propagate texture)
2007: http://graphics.cs.cmu.edu/projects/scene-completion/ (searches the internet for similar scenes to use to patch up holes in images)

I was in attendance when Randy Pausch gave a dinner talk to prospective CMU Ph.D. students. It was probably the best talk I've ever seen -- driving home the point that computation is the most powerful force for changing the world in the future.

I wish I had a video of that

Sredni Vashtar went forth,
His thoughts were red thoughts and his teeth were white.
His enemies called for peace, but he brought them death.
Sredni Vashtar the Beautiful.

*promote

And R.I.P. Randy Pausch

I love this guy, I had previously watched his videos and loved how indeed he does bridge that cost gap and allows people to tinker and play with some pretty cool technologies.

The other thing I love about him is that he's very well spoken, he gets across his points really well, explains things really well and is a pleasure to listen to.

I look forward to seeing what his next big thing may be (also, check out his website to see some of his other works including a community experience using a giant slingshot and a condemned building).

>> ^bamdrew:
Good times. But, I say, what if you want to move around? or not take picture at eye level? or not look ridiculous? what then, I ask to no-one in particular?!
you guys are familiar with the other poor-man's steadycam, right? with the weights and whatnot? http://www.cs.cmu.edu/~johnny/steadycam/


But that completely defeats the purpose of this one. It's big and probably weighs a ton. You can't just slip it in your pocket. And as far as looking ridiculous? I think a tiny string looks FAR less ridiculous than that huge contraption.

Good times. But, I say, what if you want to move around? or not take picture at eye level? or not look ridiculous? what then, I ask to no-one in particular?!

you guys are familiar with the other poor-man's steadycam, right? with the weights and whatnot? http://www.cs.cmu.edu/~johnny/steadycam/

Keepon is soooo cute! I found more about him at http://www.cmu.edu/robou.

This robot vid is sweet! Check out http://www.cmu.edu/robou.

Watch the full version. Randy is a great speaker, and his ability to motivate is part of the reason that I chose CMU for grad school.

http://www.cs.cmu.edu/~marekm/projects/beatbots/

from youtube:

His is a live-action performance from the annual Carnegie Mellon University Entertainment Technology Center's (http://www.etc.cmu.edu) Building Virtual Worlds Show. The performance is from December 6th, 2006, and students have 2 weeks to brainstorm, develop, and implement a virtual world, which in this case, was required of us to make an entertaining experience for a large audience.

The performance is given in McConomy Auditorium in Carnegie Mellon, via the technology of the Playmotion. The playmotion is a motion sensor technology that receives data on the user's head and two hands.

Members of the team include:
Paul Capriolo (programmer)
M.E. Chung (designer/fx)
Carlos Pineda (audio/stage mgmt/script)
Jake Rheinfrank (artist)

Run Wrake has done some very strange stuff, but this is quite pleasant. Its very Grimm's Fairy Tale.

http://www.cs.cmu.edu/~spok/grimmtmp/ if you're not familiar with how crazy and weird most of the grimm's fairy tales are, that site is worth checking out.

The image here demonstrates how the system breaks things down into ground/billboard/sky.

Don't think this would work so well for pr0n because it doesn't really handle curves

I tried installing this but it didn't find some DLL's properly in the matlab libs..despite setting the PATH.

w00t!!!! It's totally something that you can download!!! Check it: http://www.cs.cmu.edu/~dhoiem/projects/popup/

I haven't gotten started yet because I'm a bit intimidated by this 100mb "matlab" thing, but it's interesting that you seem to feed it this bunch of pictures with predefined "3d-ness" and then maybe it learns how pictures map onto 3d stuff, and then it goes on and can do your pictures... or something? Anyway, I plan on fooling around a bit with it in the next few days and see if I can make anything cool with it.

I love how it works even with the painting of the ship! I wonder if you could use it to make virtual pop-up books out of your family photo album or something. I wonder if they are considering creating something like that either as a consumer software package, or perhaps as a small business.

In any case, really neat technology.