Did the leaves fall off? Something not working right? Let us know how we can improve the Photosynth experience.
use the technology of cloud computing to remotely process the synths. this seems to be what seadragon.com dose if your have trouble ask them how?
:) Except that Photosynth's processing load would be up to 1200 times as heavy, simply for the image conversion and then you would add on the feature extraction, indexing, matching, and reconstruction computation on top of that heavy load.
Also, since people wouldn't feel the need to restrict themselves on the number of images used in a synth ("After all a bunch of servers can synth far more pictures than my PC can!"), the processing load would rise even higher.
Their choice to not launch Photosynth as a cloud computed service was very much intentional and one that they believed strongly in. They've also said that cloud computing is very much neccesary for some cases (i.e. to make synths from mobile devices) and will be coming in the future.
One thing to keep in mind is that linking synths together has been on the roadmap for quite a while and so far no one has said how that is to be computed. It may well be on Azure like Seadragon.com.
Even though Photosynth has left Live Labs (where they were formed and where Seadragon is still developed), I consider it quite telling that Live Labs is now pursuing server-side conversion of images to dzi format. Doesn't that sound like testing the waters for server side synthing to you?
Considering that the University of Washington is furthering their Photo Tourism work (which Photosynth is based on) and this year's new presentation from them is "Building Rome in a Day" http://grail.cs.washington.edu/rome/ (massive photo sets processed on a server farm in less than 24 hours) and also considering that not only Microsoft Research is sponsoring their research now, but Google as well, I would be very much surprised to find Photosynth asleep at the wheel. The only major growth we've seen this year so far are the forums, a color scheme update, first release of Silverlight viewer, and search improvements.
We have 2 more weeks until the anniversery of the first public release of Photosynth and until the end of the year, really, to hear what else the Photosynth guys and gals have been slaving away on. Granted porting the viewer to Silverlight, adding highlights to it, etc. is all hard work and defintely takes time, but I'm hoping to hear one big headline for them for 2009 and it doesn't feel like I have yet.
Server-side synthing could mean synth-linking, mobile photosynthing, Photosynth for Mac + Linux, enhancing current synths, providing Photosynth services to photo sharing sites, etc. etc. I really hope that no one like Apple or Google beats Photosynth to the punch on those scores. Apple's has already talked about Augmented Reality support for apps in iPhone 3GS 3.1 firmware and I know once Google starts down this road it's going to be 100% cloud mania from them and people will sing their praises. I'm hoping the Virtual Earth guys get there first.
For the reasons why they didn't go with cloud computing from day one, please watch these videos:
Hi - I work on the Photosynth team, and this question comes up often, so let me summarize our thinking. Cloud synthing sounds great, and will definitely be the way to go if and when we tackle the problem of synthing directly from mobile devices.
For desktop computers though, we've discovered that in almost cases people's upload bandwidth is such that uploading takes longer than the actual synthing. Since we do both simultaneously right now, there would be usally be no time saving for our users if the synthing were done in the cloud.
Secondly, as Nathanael says, the compute required for a big synth (~300 images+) is really substantial, and as you've probably noticed, it scales non-linearly. (i.e. synthing 400 images takes more than twice as long as synthing 200 images). Doing this with in the cloud would get very expensive, very quickly, as we scale up.
But, there *are* clever things we can do in the cloud. We're definitely working on this. Stay tuned!