What is Digg?176 Comments
- jfanaian, on 10/12/2007, -7/+43Thats pretty freaking sweet, I can't wait to be able to play with that.
- mephitix, on 10/12/2007, -1/+32Eh, they're thinking from a high-level point of view since this is a prototype. It's always best to think outside of the box and then try to fit it into your strategy instead of always worrying about resources and never being able to think creatively.
- dec0ded, on 10/12/2007, -3/+29My only question is how are the initial 3d environments created, is it dynamic or will someone have to spend time placing each image separately?
Looks like they already answered my question..
"Each photo is processed by computer vision algorithms to extract hundreds of distinctive features, like the corner of a window frame or a doorhandle. Then, photos that share features are linked together in a web. When a feature’s found in multiple images, its 3D position can be calculated. It’s similar to depth perception—what your brain does to perceive the 3D positions of things in your field of view based on their images in both of your eyes. Photosynth’s 3D model is just the cloud of points showing where those features are in space." - uzusan, on 10/12/2007, -4/+24Anyone else notice the bottles of Microsoft "Brain Wash" on the desks in the video?
- chriskzoo, on 10/12/2007, -5/+24Why? It'll be so much more fun to watch Linux nuts (yes, I know Windows software can be open source too) complain that a MS product is not available to them.
- inactive, on 10/12/2007, -7/+25You're a moron.
- inactive, on 10/12/2007, -2/+19I hope user photos are approved and not automatically inserted. If you had a connecting photo, you could insert tubgirl or goatse into the middle, while the seams of the photo still matches the scene, and voila! Millions lose their lunch.
- shranko, on 10/12/2007, -2/+17Very cool tech. The guy who figures out how to integrate it with porn will be the next millionaire.
- inactive, on 10/12/2007, -6/+20This looks incredible...
- inactive, on 10/12/2007, -3/+16In case someone gets too excited... That's not the software, that's just more video demonstrations.
- cathode, on 10/12/2007, -11/+23flickr? it's like old ***** for egotistic photography wannabees. I have better things to do than to look at someone's photo of some internet cafe. at least this thing would have real interest value, be dynamic, and actually allow you to walk around an object.
- axentrix, on 10/12/2007, -6/+17It is..The smallest is about 1A MB, not FF
- ConceptJunkie, on 10/12/2007, -3/+14Microsoft Research is working on tons of really cool stuff, 98% of which you never hear about. This is a perfect example, although hopefully this will turn out to be more than vapor. Despite everything you hear about Microsoft (and believe me, I'm no fan of the company in general), their research group does some groundbreaking stuff.
- inactive, on 10/12/2007, -0/+10You mean "this would be more useful THAN websites like Flickr"? :>
- Double-Z, on 10/12/2007, -3/+13Don't upload your pictures to it then, simple.
- axentrix, on 10/12/2007, -2/+11That's a totally awesome idea, but will a more large-scale version, like a whole city, work as intended?
How cool would google-earth be with something like this? - rhinopig, on 10/12/2007, -0/+9Yeah. While in theory i guess you could link any 2 images that overlapped (like the brain), my question is what about time and lighting and movement?
For example in there demo they just used one (large) set of photos taken at the same time, but how well would it work with 1 picture taken at noon and one at night? what about on a sunny day and and a rainy one? How do the people in the foreground effect it's ability to link images? construction scaffolding? etc.
Just linking a bunch of photos from one photo shoot seems pretty trivial, but their apparent goal of linking any 2 pictures regardless of who took them, when, at what quality, etc. seems much more far off (not impossible, but the video didn't show anything to that effect. just said 'what if'). - jals, on 10/12/2007, -1/+10I'll believe it when I use it. Great idea but come on, it'll never work in practice with real people's photos which don't look as perfect and tidy as the ones shown.
- mahoutie, on 10/12/2007, -3/+12Can't believe I'm actually digging a microsoft product!
I need to go lie down now.... - Hydroxyl, on 10/12/2007, -2/+10http://download.microsoft.com/download/3/9/0/3902e391-734e-470f-806b-4d779f07749f/PS_Overview_1280x720_5M.wmv
Look at it in hi-def, it's even more crazy... - inactive, on 10/12/2007, -2/+9Yeah, it's just like VRML, wise guy, except that social software thing and digital imaging. And you need no 3D scanner for using it, only a digicam everyone has nowadays. It will be a great success.
- Mr.Scientist, on 10/12/2007, -0/+7@rhinopig: There's a pretty nifty algorithm called SIFT (Scale invariant feature transform: http://en.wikipedia.org/wiki/Scale-invariant_feature_transform ). That algorithm detects characteristic image features regardless of image rotation and resolution and is relatively robust against illumination variation. It seems like an ideal candidate for what Photosynth does. They take a bunch of pictures, extract characteristic features and match them with features in other pictures. Then they have a huge equation system of rays through the scene with camera and feature positions as unknowns and feature matches as constraints. After solving that equation system, they get the point cloud that you see in the video. Then it's only a matter of selecting the right images (filter for camera position and angle) and showing images at the appropriate depth. I bet the devil is in the details, but in a nutshell that's what's going on.
- chesterjosiah, on 10/12/2007, -1/+7Actually, Hex 1A converts to Decimal 26.
Anyway, I noticed the file size typos too. I just assumed they typed an extra 5 but actually meant 25mb. The 320x180 file is actually 26mb. Looks like they're just rounding to the nearest 5mb. - blaisea, on 10/12/2007, -1/+7This is Blaise, the Photosynth architect. It’s great that our project’s generating so much discussion. I wanted to clear up a couple of things that might not be so obvious from the videos. Some posts raise the smart point that the collections we show off were all from one guy, one trip to Rome, one camera. Noah Snavely, the author of the original prototype software (a Microsoft Research/University of Washington collaboration), actually used Flickr searches to get photos. The results were pretty amazing-- for example, hundreds of shots tagged “Notre Dame”, showing everything from small details to wide shots of the cathedral, day and night, taken with devices ranging from cellphone cameras to professional, all registered together perfectly. Pictures dominated by a big face in the middle, with a bit of Notre Dame visible in the background, registered too. So did images with scaffolding during a renovation. Noah’s paper and video showing these results will be available in the next few days, if they aren’t already (he’s presenting the paper at SIGGRAPH early next week). This work and the MSR followup give us a lot of confidence about this technology working to connect many peoples’ photos. We used our own photos in the video simply to avoid any copyright questions.
I also wanted to mention that my team and I, as well as Live Labs in general, have a real startup culture. I’m not saying that as a corporate drone or a marketeer. Six months ago, I was running my startup out of a loft in Ballard, Seattle. The videos were a fun surprise to me. They certainly weren’t big-budget jobs concocted by a marketing committee. I think our videographer guy just had more fun than usual shooting and editing, since he had such visual material to work with. - bdxphoenix, on 10/12/2007, -2/+8Am I the only one who finds it odd that they use wikipedia on this page http://labs.live.com/photosynth/whatis/smartphotos.html instead of...say...encarta?
Although I'm guessing they aren't going to produce a linux version (and thus, I cannot use it), that does look like a very cool technology if they can pull it off. From all the shots of it though, I imagine that they have only done one city and it is probably very slow (they probably use the magic of video to speed up the shots).
Microsoft can come up with some good ideas, but will they be able to follow through on this or will it go the way of the WinFS? - joeshlub, on 10/12/2007, -3/+8How many photos of the same area would this actually take to make anything worth while?
- lastberserker, on 10/12/2007, -1/+6Check also this link: http://www.microsoft.com/downloads/details.aspx?FamilyID=afe093ae-b780-4ee0-b0f1-897dfd42f0f9&DisplayLang=en
- biometricks, on 10/12/2007, -3/+7I like how that one guys vision of the future of Photosynth is "taking a tour of a paticular handbag store in Paris"
- uzusan, on 10/12/2007, -2/+6It kinda looked like this : http://www.sodaking.com/product_info.php?products_id=593
Though i wouldn't fancy paying $80 a bottle. - alwaysmc2, on 10/12/2007, -1/+5@rhinopig
I was thinking that too. They would probably only link photos that are really high resolution. As for weather, maybe they can detect that too and you can choose what you want the ambiance to be like. That'd be cool. :) - aleandro, on 10/12/2007, -1/+5Looks pretty cool. Plus live lab is now collaborating with yahoo on the IM field, imagine if flickr was used as a database for this?
- biofusion, on 10/12/2007, -4/+8Man I love how people are still complaining, it's a prototype people. Yeah Microsoft made it, get over it. I like where this is heading, Live Labs has pushed some great stuff out there. Keep it up.
- kevin.gc, on 10/12/2007, -1/+5live demo: http://phototour.cs.washington.edu
(requires java) - EtherGnat, on 10/12/2007, -1/+5It's not a purse, it's a european carryall!
- tinkafoo, on 10/12/2007, -2/+6Merge this with what Amazon had going with their Mturk/A9 Blockview HITs, and we could create a huge interactive 3D database of the location of everything under the sun.
(Yeah I'm looking at you, Amazon. You really shot yourself in the foot when you dropped the A9 HITs from Mturk..) - inactive, on 10/12/2007, -1/+4Oh he will, but not in public ;)
- seitz, on 10/12/2007, -1/+4As Blaise mentioned, Photosynth is based on "Photo Tourism", a technology jointly developed by Noah Snavely and myself (Steve Seitz) at the University of Washington, and Rick Szeliski of Microsoft Research, to be presented this week at SIGGRAPH 2006.
While you're waiting to try out photosynth yourself, you can check out a LIVE DEMO of Photo Tourism now, and read more about the technology underlying photosynth at:
http://phototour.cs.washington.edu/
See also Microsoft Research's phototourism page:
http://research.microsoft.com/IVM/PhotoTours/ - inactive, on 10/12/2007, -2/+5It's something that could be a revolution in photography. Btw, stop whining around, it's the internet.
- inactive, on 10/12/2007, -2/+5God bless 'em if it works.
The Newton was "Awesome!" as well before it came out.
Smells like Fable to me. - Sabin, on 10/12/2007, -7/+10It is..The smallest is about 11 MB, not 255
Conversion for those of you who are hexadecimally challenged and digging this guy down. - vbsurfer, on 10/12/2007, -1/+4This whole idea makes many different possibilities with other applications. Video, art, etc.
- Reno582, on 10/12/2007, -1/+4but what is it? A program? an extension of .NET? if they could implement this into AJAX then that will be a whole new way to connect users through digital pictures
- maverick999, on 10/12/2007, -2/+5Here's a direct link to the video demo:
http://labs.live.com/photosynth/videodemo.html - jdollah, on 10/12/2007, -2/+5Where's all the comments saying that MS stole this from iGhay?
- Reno582, on 10/12/2007, -2/+5This would be more useful to websites like Flickr, very cool stuff
- digitalArtform, on 10/12/2007, -0/+2It was also done more than a decade ago
- scb0825, on 10/12/2007, -1/+3that guy looks like a pot head
- yeahbuddy, on 10/12/2007, -1/+3I love how everyone has the buzz word "Labs" now.
LOL. - Magadass, on 10/12/2007, -1/+3See how, what, and why! http://channel9.msdn.com/showpost.aspx?postid=220870
- gabeN, on 10/12/2007, -3/+5I think the e coolest aspect is the web-integration... If you have a spider that crawls the web strictly for images (or possibly images + contextual data like location and subject) and can cull those images into a centralized database, the "virtual world" created in a program like this will be perpetual, and continually expanding. You needn't even upload your pictures to this program, If they are on the web, they're in Photosynth!
Bravo! -
Show 51 - 100 of 178 discussions
© Digg Inc. 2009
— Content created and posted by Digg users is