This is true, though machine learning models will learn typical scene structures, alongside the rendering parameters. It is also common to use more than one sou

This is true, though machine learning models will learn typical scene structures, alongside the rendering parameters. It is also common to use more than one source image (from a different viewpoint, a bit like photogrammetry).