I have a good-quality, precise 3D scene reconstructed from photographs/video by VisualSFM. All camera parameters, including 3D position and lens details, are known.
Now, that the hardest work has been accomplished, I'm sure that obtaining real-world dimensions of the model is easy... only how...