A podcaster friend of mine has a standard approach to tech reviews… “What’s the problem to be solved?”
Hands up here who is using PhotoLab to edit graphics.
No? No-one?
Hands up who thinks DxO’s training set of images contained graphics images.
I don’t know how many more ways I can say this. The models are trained to find difficult subjects in complex photos. If the training data did not also contain graphics, then it isn’t going to do well with them.
Time to bail on this thread.