Get all your questions answered about our latest Photosynth Technical Preview.
I'm not the first person to request this idea.
For me, one of the most beautiful parts of the original Photosynth was seeing photos that I cared about in the context of a larger 3D model but no one (not even me) wants to look at all the input shots.
I know that the images from the Kinect is probably not what the photography enthusiasts on the team think of when you imagine the beautiful photography that you'd like to be host to, but why not give us tools so that the more geeky among us can use the Kinect to build out a detailed point cloud or mesh of a space which more beautiful images can then rest in the context of?
Kinect Fusion seems like an ideal way (not that carrying around a laptop or tablet with a Kinect attached is completely ideal =) to build out a space which images could reside in. http://research.microsoft.com/en-us/projects/surfacerecon/
I understand that your vision for Photosynth is to present beautiful photos in a new and compelling way, but there are some of us who would love to put in the effort to build out an entire 3D scene in good detail so that others could come along and take a much simpler set of photos and get a more correct alignment or much broader context because of the geometry that someone like me captured.
I'd love to make detailed synths of every tree in a public park and have them all link together so that then anyone's photo in that park (uploaded to Photosynth or not) would have a good chance of being geolocated by Bing, but I'm not going to suggest that any normal person wants to stare at thousands of close up photos of tree trunks and tree branches.
What are your thoughts?
I'd love to help build a database of 3D models of trees so that if someone tags the species, you could take the similarities + morph them to generate believable models of a tree seen in a single photo.
To me, this sort of thing could be fundamental for getting Bing moving more toward A.I. - able to talk shop with car fanatics or understand what botanists are talking about or watch a sports match and be able to follow the game as well as a referee.
It seems to me that comprehending so many subjects relies on a full spatial understanding of the components of a situation.
I see this sort of thing as having all sorts of implications.
What about when Windows Live Photo Gallery wants to allow users to delete something from a photo? What if it can look up other photos from that location and has a three dimensional understanding of the scene that's getting close to ours to properly fill in the hole in the photo?
Anyway... I'm rambling off topic now, so I'll stop.
I'd still appreciate being able to build out a full model of a dining room and then when someone is having a birthday party there, just focus on capturing facial expressions + memories + present those.