This can also be done on panoramio.com using the "Look around" feature. For example: http://nv0.panoramio.com/navigate.php?id=1674082

This demonstration is much more impressive than the "Look around" Panoramio! You can "enter" in the picture :-)

This is way better than Photosynth, and in the future places like flickr would look amazing with this kind of stuff...
Hell, look at this! Just think what we were doing ten years ago and what can we do now! :)

Could this type of technology also be used to automatically construct fully detailed 3D models from a set of photos, including textures?

> Could this type of technology also be used
> to automatically construct fully detailed 3D models from
> a set of photos, including textures?

Such a thing is commercially available for 3D artists. I'll ask someone who I saw using this what it's called.

This is freaking cool!!!

This technology is already done or only in the making?

I like how it's a university project, partnered with Microsoft, got photos from Flickr (Yahoo!), and posted a video on YouTube (Google).

I think its been at least a year ago that I first saw this demonstrated by Microsoft. I sometimes think they waste a lot of "research" money, but this one looks really useful and it will be a shame to have it show up somewhere as yet another "Windows only" product.

One problem though is that there are not really that many places in the world that have the density of photos used for the demonstrations here. Even for touristy places you are going to have thousands of photos taken from just a few vantage points and of course only a few of the many pictures of scenic objects get uploaded to some public place where they can be analyzed with this kind of software.

For many other spots, satellite, ground and helicopter shots will be more valuable for a long time to come.

Perhaps this is a dumb question, but could they do the same sort of process with videos, by assuming each frame is a photograph?

A similar / the same (?) technology was presented in March 2007 at TED: http://www.ted.com/index.php/talks/blaise_aguera_y_arcas_demos_photosynth.html

I don't find this way of navigating through photos useful for myself, because it just focuses on the objects in the photo. When I look for a photography I'm interested in well composed photos with interesting colors taken from an interesting perspective; I'm interested in the the photography as a 2 dimensional 3d-illusion and not so much in the object.

These technology of combining hundreds or thousands of randomly taken pictures into a coherent roundview will be useful however for producing automatically rendered 3d models and textures for those. It's much more a foundation for a new user experiment than a new user experience itself.

I suppose that you already noted that your commentators are part of the Blogoscope experience, an important part. It's nice to find the original TED material, very interesting.

No, I'd love to read "Bleak House" just like that. (Like in the TED lecture.)

And thanks for the pointer.

suppositio.us

nick: yes, that's how they do videos. There's an example in the presentation, where they wave a toy in front of a video camera to get enough angles to form the 3D model.

They are coming up with some cool photo software these days. I would love to see this software team work with the Gigapan folks. They are working on a robotic panorama device where you just set your point and shoot camera into the device and turn it on and the robot will take a couple hundred pics of the scene at all different zooms and angles. This creates one hell of a panoramic image. With the two software packages combined, you could have pretty sweet 3D panoramas.

Gigapan article:
http://www.popphoto.com/photonews/3793/robotic-panorama-platform-helps-photographers-create-huge-images.html

Related: "Using Photographs to Enhance Videos of a Static Scene"
http://www.youtube.com/watch?v=1PktKqyRXIE
http://grail.cs.washington.edu/projects/videoenhancement/videoEnhancement.htm

> Could this type of technology also be used
> to automatically construct fully detailed 3D models from
> a set of photos, including textures?

To follow-up on this, someone tells me it's this software:
Strata Foto 3D 1.5
http://www.google.com/search?hl=en&q=Strata+Foto+3D+1.5