It really is like Minority Report!
June 6, 2007 7:34 PM   Subscribe

Photosynth. Blaise Aguera y Arcas (second one down) does a live demo (with some subtle humor) of the product we've discussed previously. Via the wonderful Ted (mentioned a few times).
posted by allkindsoftime (24 comments total) 7 users marked this as a favorite
 
Ahh I wondered how long it would take for this to show up here.

Now the question is, will Microsoft be able to turn this frankly brilliant research into a useful product? My money is on Google putting their Street View data into a system like this and giving us a 3D-textured city view before this gets to the product stage.
posted by Skorgu at 7:53 PM on June 6, 2007


Skorgu, maybe, but I don't think Google currently has the technology, and MS is set to do it a LOT better.

I'm still stuck on one single thing - if you have a photo of yourself on the web (say on a company website), there is a chance this technology could link that photo (which is attached to your name) to every other photo of you on the web (drunken cross dressing and the like) that is not labeled with your name. This is, at heart, the next evolution in search engines.

Pause and give that a thought.
posted by Muddler at 8:54 PM on June 6, 2007


Holy shit. I interviewed with Blaise a few years ago, when I was extremely dissatisfied with my erstwhile job. I decided not to make the leap because his li'l company (pre-acquisition) seemed held together by coffee and hope.

His demo has gotten better, to say the least. Kickass stuff.

I've since moved on to very gratifying work, but clearly, if I'd joined up with Sand Codex, I'd be very busy very now and likely rather rich.

So it goes.
posted by gurple at 9:10 PM on June 6, 2007


er, very busy right now, that is.
posted by gurple at 9:14 PM on June 6, 2007


Wow.
posted by CitrusFreak12 at 9:15 PM on June 6, 2007


Muddler, see riya. (well, they started as face-recognition. i guess they diversified at some point)
posted by misterbrandt at 9:30 PM on June 6, 2007


I liked the zooming thing. Does anyone have a link for that?
posted by smackfu at 9:56 PM on June 6, 2007


Disclaimer: I work in the group that put this together.

We just launched a new set of synths that you can play with yourself (under XP SP2 or Vista), which we built in partnership with the BBC. They show some beautiful historic landmarks in the UK.

Tip: space bar advances to a nearby photo, L,R,U,D arrows makes an approximate move in those directions, and the mouse wheel will zoom like crazy.

I know that a lot of people will hate us for only supporting Windows, but you should consider a couple of things: (1) it really is much easier for us to stick with a single target platform, (2) We moved some of the base technology to Silverlight as fast as we could so that it would be cross platform, and (3) I am typing this on a Mac.

Finally, this technology would not have been possible without some brilliant research from MSR and UW, and Noah Snavely in particular.
posted by dr.flakenstein at 11:20 PM on June 6, 2007 [1 favorite]


So, flakenstein, are you guys all eastside now, or did you get to keep your cool Ballard space?
posted by gurple at 11:27 PM on June 6, 2007


Part east and west. The entire team that was in Ballard is now in an even cooler space in the Smith Tower.
posted by dr.flakenstein at 12:26 AM on June 7, 2007


OK, yeah that is pretty freaking amazing (been playing with the bbc stuff for an hour). I don't understand it hardly at all and it was a shame that Blaise went through it so quickly. This is really the sort of multitech futurism where some long film intro would be very helpful. But colour me impressed.
posted by peacay at 12:42 AM on June 7, 2007


This is the kind of stuff that truly gets me excited. I am enthralled by the idea that a vast pool of images can be accessed to view parts of the world I will never get to. Can you imagine the entire globe being able to be navigated in this fashion?

All that's missing, if I didn't see it, is a chronological aspect, so you could also view a space over time.
posted by maxwelton at 12:46 AM on June 7, 2007


Sweet merciful crap. I've watched some Google Tech Talks etc. that left li'l ol' layman-me flabbergasted...but I've never heard an audience burst into spontaneous laughter/applause like I heard in this video. That's endorsement enough for me. Plus, it just LOOKS and SOUNDS like FUN. Yes, please!

It's enough to make me glide giddily past my initial grumpy-point that the speaker sounds suspiciously similar to the text-to-speech 'bot Bruce from my old MacInTalk days. "I sure like being inside this fancy computer."

I kid, actually. As swept away as I am by the software demonstration... as entranced as I am by the audience response... I'm even more enchanted by Mr. A y A's genuine, playful JOY at what he's helped create. LOVE it.

So I feel that I'm doing him a disservice by comparing his tone to that of some automated, synthetic speak-drone that awakens at a keystroke, ready to belch forth its scripted phonemes in a rickety pastiche of human conversation.

Speaking of which...hi, dr.flakenstein, how've you been??
posted by unregistered_animagus at 12:59 AM on June 7, 2007


So I feel that I'm doing him a disservice by comparing his tone to that of some automated, synthetic speak-drone that awakens at a keystroke, ready to belch forth its scripted phonemes in a rickety pastiche of human conversation.

I noticed his way of speaking, too. Doesn't strike me as robotic per se, but really blue-blood, listen-to-my-perfect-diction kind of way. Rich-kid speak.

But wow, speaker aside, this is amazing stuff.
posted by zardoz at 1:48 AM on June 7, 2007


Man, did I jive with the whole idea of social networking that this represents. From my standpoint, I can't wait for this technology because as more and more people dive into their grandparents' archives and scan in old photos, more and more "who is that person?" questions will be resolved simply through the collective making a connection to that person's face. I could finally prove to my mother that she can't really do anything about the second-chin effect, as it appears to be the fault of some Norwegian ancestor from 1600.
posted by thanotopsis at 4:33 AM on June 7, 2007


Yeah, I saw this last week and it's just so damn cool I made anyone I thought might care about it sit and watch it. dr.flakenstein, can you confirm/ deny one thing for me: in the part where he shows the poster of Notre Dame, is the system linking other photos to the individual parts of the cathedral within the poster? Because if it is, I give up. That's far too cool.
posted by yerfatma at 5:15 AM on June 7, 2007


yerfatma, my understanding of the video is that yes, indeed, it is recognizing the poster as just another photo of the cathedral and putting it in the proper place in the unified model.
posted by Skorgu at 7:12 AM on June 7, 2007


how about a shout-out for steve seitz!? i know he worked on this stuff but i didnt see his name come up anywhere.
posted by joeblough at 8:38 AM on June 7, 2007


zardoz, re. "rich kid speak" - you couldn't be more wrong about Blaise. He comes from a *very* modest background, and while he went to Princeton he was something of a charity case. I think what you hearing is but a hint of his brilliance (which is considerable).

animagus, same old same old.

joeblough, right you are, but Noah just deserves a little extra props since this is his Ph.D. work. Rick Szeliski, of MSR, with those other two are the main three contributors to virtual photo tourism (the research project responsible for the machine vision techniques that Photosynth uses).
posted by dr.flakenstein at 11:00 AM on June 7, 2007


well get that mac port up and running!! i assume it won't run under parallels given the video card requirements? maybe its worth a try.
posted by joeblough at 11:16 AM on June 7, 2007


its very cool, but a little tough on the digestive track
posted by sswiller at 1:11 PM on June 7, 2007


my understanding of the video is that yes, indeed, it is recognizing the poster as just another photo of the cathedral and putting it in the proper place in the unified model.

Oh, I get that part, but what I want to know is if it's also relating all the other photos to their place in the image of Notre Dame in the poster.
posted by yerfatma at 1:19 PM on June 8, 2007


yerfatman, for the parts that are visible, yes.
posted by dr.flakenstein at 7:12 PM on June 8, 2007


/splodes
posted by yerfatma at 12:09 PM on June 12, 2007


« Older The Beast of Omaha   |   Jump back, wanna kiss myself! Newer »


This thread has been archived and is closed to new comments