It's true that there aren't too many practical applications of GAN's yet, but I'd argue that transfer learning is already pretty powerful. It's fairly commonpla

It's true that there aren't too many practical applications of GAN's yet, but I'd argue that transfer learning is already pretty powerful. It's fairly commonplace to approach a compute vision task by starting with VGG/AlexNet/etc and fine-tuning it on a relatively small dataset.