Members Nathanael Lawrence

Aug 27 2011: Since then, the Bing team has been working on increasing the speed to start building exactly the sort of service that you're describing and that Photo2Search described over half a decade ago.

2011 April 20 saw Blaise announcing Bing's Read/Write World project at the O'Reilly Where 2.0 2011 conference and they are now able to take a geotagged image and match it to other geotagged imagery very nearly in realtime. (As you point out, for speed reasons it is necessary to give the matching engine a place to start looking for matches. Trying to match each new image to every other image on the web would still take far too long.) This speed increase is similar to how Noah's Notre Dame collection that Blaise shows above took a week or more of computing time using a big network of computers whereas by 2008 August 20 when Photosynth was released it was able to do the same job on your laptop in less time than it takes to upload the photos on a typical American broadband connection.

Avi Bar-Zeev (formerly of Keyhole - Keyhole being who Google bought to acquire Google Earth) and Blaise have given several public talks about Read/Write World http://bit.ly/readwriteworld (six video presentations in that link) - both its promise to really start to deliver on some of these first dreams of Photosynth that Blaise is talking about in the 2007 TED Talk above (connecting everyone's photos, rather than just little clusters of 2,000 or less as the current version of Photosynth does) and some of the challenges in bringing that to pass.

If you want more news, Steve Seitz at the University of Washington, Noah Snavely at Cornell University, Rick Szeliski at Microsoft Research, and Yasutaka Furukawa at Google are really good guys to search for for keeping up with news in this area of internet-scale photogrammetry. Also watch Henri Astre's 'Visual Experiments' weblog for news around new techniques that other researchers are publishing.

Aug 27 2011: Nathaniel, for a reverse image search engine, check out TinEye. http://tineye.com/
They predate both Google Goggles http://www.youtube.com/watch?v=Hhgfz0zPmH4 and Bing Vision http://www.youtube.com/watch?v=hzWT9ClzeVk , but all three of those services are focused mainly on recognizing specific images (book, movie, music, or game case covers, iconic images, etc.).

You're very much thinking along lines that all the big search engines and major computer vision researchers are striving towards. I remember seeing Craig Mundie at Microsoft show a research project from Microsoft Research Asia called Photo2Search back at the Microsoft Financial Analyst Meeting 2006 (actually it was the last thing out of his mouth before he called Blaise up on stage for a demo of Seadragon and Photosynth - 32 weeks and two days before the TED Talk above was filmed).

What the Photosynth team found back in 2006 - 2008 was that in the photography that's already online there are little clusters of photography that connect, but even on most landmarks most people shoot it from the same angle, so you don't get much of a 3D model. When they launched Photosynth in 2008, they decided to make it its own photo sharing site where everyone who uploaded would be encouraged to change how they photograph things in order to make a good synth (e.g. walk all the way around a building).

This is better on a per-landmark basis, but there still isn't enough photography out there to connect each landmark to those surrounding itself. By 2009 Blaise was talking about Microsoft and others gathering a 'trellis' of imagery which users' 'grapes' of own content (photos, video) could then latch onto. They did a proof of concept with matching geotagged creative comons photos from flickr to Bing Streetside panoramas and showed it at TED 2010, but it was only in San Francisco, Seattle, and Vancouver Canada. http://bit.ly/streetsidephotos

Aug 22 2011: Daniel, the guys at Bing continue to work on this. I don't know if you've seen their recent mobile panorama app for iOS ( http://www.youtube.com/watch?v=6BbuPPOVXQo&hd=1 ), but the live image tracking in the viewfinder on there when you're shooting a new panorama is very much related to the live mobile video being tracked against Bing Maps imagery seen in Blaise's demo. You can read some more of Blaise's thoughts on it on his blog ( http://styleisviolence.com/photosynth-app/ ) and hear him talk about it in his more recent talks about what they're doing with Read/Write World ( http://bit.ly/readwriteworld ).

I won't say that I'm happy with current bandwidth on wireless networks, but the situation is helped by the fact that all Photosynth and Bing Maps imagery is stored in a tiled multi-resolution format, meaning that the mobile user only needs to load whatever fits on their screen at the moment (and those screens are smaller, so that helps). For more on that, see the Seadragon portion of Blaise's 2007 TED Talk ( http://www.ted.com/talks/blaise_aguera_y_arcas_demos_photosynth.html ).

To play with the version of Bing Maps that Blaise is using in the demo, click here: http://bing.com/maps/explore/ and use the Map Apps to get Worldwide Telescope or Streetside Photos.

For more videos, see my spreadsheet here: http://docs.com/XFO

May 25 2011: Everyone who enjoyed this talk (and Blaise's 2010 Augmented Reality Maps presentation) will want to check out his recent talks about Bing's new collaborative project, entitled Read/Write World.

Here he is, announcing Read/Write World
Event: O'Reilly Where 2.0 2011
Date: 2011 April 20 (Wednesday)
Video: http://youtube.com/watch?v=4X9u4JG9H6E&hd=1

Here he is, discussing the history of Photosynth and where they are heading with Read/Write World.
Event: Queen Anne Science Café
Date: 2011 May 03 (Tuesday)
Video: http://youtube.com/watch?v=F9J4cfjau0Q&hd=1

Here he is, presenting Read/Write World and talking about how it completes the original vision of Photosynth
Event: Augmented Reality Event 2011
Date: 2011 May 18 (Wednesday)
Video: http://youtube.com/watch?v=tojFFlCUhIs&hd=1

Although Read/Write World has not yet launched yet (they aren't even in Beta yet), here are some links:
Their Website: http://readwriteworld.net
Their Facebook page: http://facebook.com/readwriteworld
Their Twitter profile: http://twitter.com/readwriteworld
Their Vimeo channel: http://vimeo.com/readwriteworld

As you will hear Blaise explain, the original vision of Photosynth was that everyone's photos would begin to link to each other. Before Photosynth launched, the team tried making synths out of photos from the entire Internet and although it worked some in a few places, in most places everyone has taken the same photo from the same spot, so 3D reconstruction doesn't work, so the Photosynth team went with a photo sharing site model and encouraged Photosynth users to shoot photos specifically for 3D reconstruction - walking all the way around buildings, etc. Photosynth as it exists is great, but once you upload your synth of a place it just sits there. Even if other people have made synths of the same park or landmark, in today's Photosynth, those synths haven't connected to each other once they're online.

Read/Write World is that missing piece of Photosynth - the connecting online.

May 25 2011: Everyone who enjoyed this talk (and Blaise's 2007 Seadragon and Photosynth presentation) will want to check out his recent talks about Bing's new collaborative project, entitled Read/Write World.

Here he is, announcing Read/Write World
Event: O'Reilly Where 2.0 2011
Date: 2011 April 20 (Wednesday)
Video: http://youtube.com/watch?v=4X9u4JG9H6E&hd=1

Here he is, discussing the history of Photosynth and where they are heading with Read/Write World.
Event: Queen Anne Science Café
Date: 2011 May 03 (Tuesday)
Video: http://youtube.com/watch?v=F9J4cfjau0Q&hd=1

Here he is, presenting Read/Write World and talking about how it completes the original vision of Photosynth
Event: Augmented Reality Event 2011
Date: 2011 May 18 (Wednesday)
Video: http://youtube.com/watch?v=tojFFlCUhIs&hd=1

Although Read/Write World has not yet launched yet (they aren't even in Beta yet), here are some links:
Their Website: http://readwriteworld.net
Their Facebook page: http://facebook.com/readwriteworld
Their Twitter profile: http://twitter.com/readwriteworld
Their Vimeo channel: http://vimeo.com/readwriteworld

In Blaise's talk above, he shows where they did a Community Tech Preview of Streetside Photos with a one-time crawl through Flickr's creative commons licensed photos in just three cities: San Francisco, Seattle, and Vancouver Canada.

What Read/Write World does is speed this process up so that when you submit an image to Read/Write World with an approximate latitude and longitude (to help it know where to start trying to match the image to existing imagery) it will match that image almost immediately and return that image's location to you.

This is very similar to the difference between the Community Tech Preview of Photosynth (where it took Live Labs multiple days to calculate a single synth and only put a few online to be viewed) and when Photosynth launched on 2008 August 20, where they had sped it up to work in less time on your laptop than it takes your photos to upload.