Wow this is so awesome, even cooler than google street view:

http://labs.live.com/photosynth/

More info about the technology in the vid here:



Props to slashdot

Holy s*** I just went to the site and that is f***ing sweet! It will be great when that is integrated with something like google's street view.

Freaking Sweet!


It's great that google and MS are competing so heavily...

That video is f***ing amazing - its like watching those movies where you have someone in front of a computer going "enhance...enhance...enhance..." and it actually doing stuff

I thought the same, then actually did it myself at the site. Check out the artist's room that has pics in it that are 80MP big... then switch to the alternate view (one of hte buttons at the top right). You can zoom in sooo far on some of the pics.

Oh god, think of the porn applications!

I bet Britney is regretting not wearing panties with that skirt

that really is awesome

nice to see technology moving in this direction

What amaze me even more that the tech behind it is the fact that it is *very* fast. It responses almost instantly to zoom, pan & scroll. Also what's more amazing is the foot print of the plugin for firefox, 5MB. Well done Microsoft :)

this is very cool. i like the collective photo thingy at the end.

god thats f***en cool

What amaze me even more that the tech behind it is the fact that it is *very* fast.

Yeh, I wonder how that works. I don't really get how they can have soooo much data and have it respond so quickly without massive amounts of pre-caching/pre-generation of images?

Did you spell Microsoft incorrectly to be funny or because you can't spell?

^^ I can't speel heh. Fixed the title

/me discovers similarity view.

HOW DOES IT WORK!@#!@#@!#!

edit: oh I just discovered its not a Firefox extension as I first thought. What a newb. So it's like a whole application dealie thing.

last edited by trog at 18:25:05 07/Jun/07

there are many ways to measure the similarity of two images, a simple method is to compare the use of the colour spectrum in the images, do they have roughly the same quantity of red, blue, beige etc. another is texture similarity, are similar patterns of light and dark present in the images.

then more advanced algorithms could in theory recognise buildings, faces etc that are in both images

being that they have information on the photo was taken from and what angle the camera was pointing at, it looks like they use this info as well

measuring two images and figuring out how to blend them together I get - turning them into a 3d space is just science fiction!

The site has a "how do you do it" but it's too light on tech details to satisfy my curiosity:

Each photo is processed by computer vision algorithms to extract hundreds of distinctive features, like the corner of a window frame or a door handle. Photos that share features are then linked together in a web. When the same feature is 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.

isn't it obvious? Say you have 5 images, 1 front 2 from each side progressively going further around the object. It takes those, finds a certain point on that image then figures out how much it moves in each view allowing it to determine what the 3d structure might look like.

Same way you'd make a 3d clay model from a series of photographs of a subject from different angles except it's done via software rather than your brain + hands.

probably similar tech to the stuff that makes optical mouses work

Waste of time and money.


Cool, though.

this could take making game mod maps for schools to a whole nother level...

Wow. That's pretty impressive stuff. Korea was a bit boring but venice and that dudes stuido were awesome.

I thought all you guys were wankers for raving about it.. it turns out you are but it's impressive all the same.

dang it, i had that idea! wouldnt of had the skills to even start it tho!

All I see is a bunch of dots the shape of a building and some really low res pics what is so cool about this?

http://www.microsoft.com/surface/ seems to have come out again (i think it was demo'd last year some time??) hoorah for Minority report style computers!! Ok maybe not a full computer but i'd dig a coffee table that could do all this stuff

qgl thread

measuring two images and figuring out how to blend them together I get - turning them into a 3d space is just science fiction!

yeah just like de javu was science fiction...

you are a negative c*** infi lol

played around witht his today and it's f***ing awesome... i am impressed!

All I see is a bunch of dots the shape of a building and some really low res pics what is so cool about this?

watch the vid. the images are from random people, random cameras, random view points, and the software stitches em up.