Is there any way to measure, say, the bumpiness or reflectivity of objects in real scenes in order to model them as accurately as possible?
Like, for example, with this scene: it’s stunning for sure, but almost looks too perfect. In a movie, for example, a scene like this would break suspension of disbelief. Maybe with more accurate measurement of the reflectivity of that lemon tart, or the coffee mug?
I think the notion of accuracy becomes somewhat subjective after a certain point of realism. For example, even for photography, different cameras, films, and filters will produce different effects. It is sort of hard to say which one is the "ground truth." Meanwhile, it is also absolutely not crazy for someone to argue that the scene in the slide is not too perfect... After all, I thought the photo was taken by a film camera from the 80s.
Given the ambiguity stated above, perhaps a more tangible task is to design a rendering pipeline that can help people achieve whatever effects they have in mind.
Some key things being modeled in this rendering:
-Rasterization of shapes onto screen space
-Perspective/occlusion/visibility problem: what objects should be in front of others -> order of rendering
-Textures mapped onto rendered shapes
-Realistic lighting from light source
-Reflection mapping onto various surfaces
-Viewing camera focus/lens mapping (the chair/table cloth are much more blurry than the cups, since the cups are in focus and the other objects are out of focus - like a real camera)
-Motion blurring (in image on next slide)