These are some very preliminary tests with photogrammetry from open-source images, using tourist photos to reconstruct 3D models of some of the world’s most photographed sites (just for ease with initial tests). I don’t know how useful this technique will be for the future of the project, I just thought I’d start getting to grips with its limitations and potentials – the software I’m using is Agisoft’s Photoscan.
It seems to demonstrate the effect I was talking about last tutorial with the most photographed elements being the most clearly defined. The distortions in the image seem to be caused where the software fails to properly align cameras/guess lens lengths.
Obviously a Photoscan done with images intended to help build a 3D model results in a scene with considerably more detail and accuracy, although having done these experiments I think it is potentially more interesting to investigate the strange anomalies produced where the algorithm gets it wrong.