AI in Sampler

gnoop

I tried to make material from this photo (hi res version of it) and get  this:

Looks like AI did zero de-lightng  and height reconstruction only a tad bit better  if any  than  old crazy bamp. 

   Is it what we could expect from AI so far  or it's just  not a subject it  have been trained for?    I got somewhat better results with photo of small pebbels for example  ,  not so blurry and weird  height  at least.        Maybe  Ai is not trained  well yet  to recognize  more macro forms?      Wonder would it be ever " intelligent"  enough.

Jerc

This is definitely outside of the range of things it was trained for. The Image to Material model was trained on mostly grounds, walls and fabric surfaces. A large-scale photograph like this with hard self-shadowing will not get you good results (yet).

gnoop

Thanks   Jerc  .       It's good to understand  what to expect from it