image synthesis remains the main objective of GAN
The resolution and quality:
Progressive growing of GANs for improved quality, stability, and variation
Spectral normalization for generative adversarial networks
Large scale GAN training for high fidelity natural image synthesis
Understanding of synthesis process
- generator perspective
GAN dissection: Visualizing and understanding generative adversarial networks
- latent space perspective
latent space interpolations
Generative adversarial interpolative autoencoding: adversarial training on latent space interpolations encourage convex latent distributions.
Feature-based metrics for exploring the latent space of generative models
GAN Loss
Regularization
hyperparameters
The GAN landscape: Losses, architectures, regularization, and normalization
Are GANs created equal? A large-scale study
Large scale GAN training for high fidelity natural image synthesis
interpretable attributes in the latent space