a) known camera positions. Combination of distance and angles defines scale. Without known camera positions you only have angles which is enough for defining shape, but not scale.
b) known distances in the object, either using markers/scale bar or transforming the final model by a calculated ratio to correct scale.