Tweeted By @OriolVinyalsML

Evaluating generative models is hard! We propose Classification Accuracy Score from classifiers trained on generated data:

-Accuracy of 43% when trained purely on BigGAN samples (vs 73%)
-Naive data augmentation doesn't work (yet!)

Paper: https://t.co/dN6xfVlyqE
cc @SumanRavuri pic.twitter.com/7HsdZFGzek

— Oriol Vinyals (@OriolVinyalsML) May 28, 2019