You must log in or register to comment.
If you are inferring 32 pixels from 1 pixel that is because the model has been trained on billions of computed pixels. You cannot infer data in vacuum. The statement is bullshit.
Yeah but it does it much faster and more efficiently (according to him).