Re:Alias-Free GAN
by pabs3 on Wednesday June 23, 2021 @11:21PM
(#27613569)
Yeah, for the goal of verifying claims that is good enough indeed.
I'm not sure it is good enough for a variety of other possibly useful use-cases though; eliminating bias in an existing model, correcting a flaw in the training code, creating a different model and proving it is better by training on the same data etc.
It would be nice if there were more public/libre data sets for ML stuff.
I'm not sure it is good enough for a variety of other possibly useful use-cases though; eliminating bias in an existing model, correcting a flaw in the training code, creating a different model and proving it is better by training on the same data etc.
It would be nice if there were more public/libre data sets for ML stuff.