I have seen conditional GANs often applied to easier datasets like MNIST and CIFAR-10 to reasonable success, but at the same time these datasets are simple enough that naïve CNNs can fairly easily max out performance on the training set. Are there any examples of training GANs to generate features and labels for more sophisticated datasets like CIFAR-100, in which even a decent-sized vanilla CNN may converge to 70-80% performance? What kinds of discriminator/generator architectures does this involve?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.