3D models from Flickrphotos
One of these "why haven´t I thought of it" moments occurred to me today (so even if I would have I probably would have not been able to pull it off but thats beside the point). Take A LOT of photos from one and the same object/building possibly from different angles and generate a 3D polygon model out of it.
Now where do you find tons of photos of the same thing - Flickr the pixel rubbish tip.
So thats what researchers from the TU-Darmstadt and the University of Washington with the help from the usual suspects (Mircrosoft, Adobe and other "we want to own the world" types) have done. Take 200+ photos of highly varying quality of any one building and let some pattern recognition magic be performed to figure out the point of view of each photographer automatically. Then interpolate the edges of the patterns and put it in a 3D grid, shake it all fix the polygons and out comes - a 3d model representation of the building.
Now I thought this is enormously cool until I saw the actual output. I think to get that quality of polygonal modeling it takes less then the 2h for a human with only very bad knowledge of ZBrush to make a shitty model like this. I mean these are architectural marbles that have fine details and sharp edges - to present a 3D model that looks like a bad cast of souvenir soap does not speak for your research. So as cool as the concept sounds this is WAYS of to be used for anything productive other then - maybe - having a proxy object with roughly the exact measures (of by 2 inches for Pisa tower they say), but this I could also by just having a background photo.
It does however seem to work quite nice for organic object (one would have to see the mesh cloud of one of these objects to actually get a real good observatory comment on this - something they (deliberately?) have not made public.) as can be seen on the fabric of the statue of liberty. So rather then taking 200+ photos of building in flickr they should maybe focus more on making 50+ instant photos from a person and get rid of laser scanners for ones and all? Or maybe they will use this tech to make a 3d representation of humans with 200+ surveillance camera photos of one and the same person and then clone them?
http://uwnews.washington.edu/ni/article.asp?articleID=37724