GANs

Generative Adversarial Networks train a generator and a discriminator in a two-player game. Historically, GANs dominated perceptual image synthesis before diffusion models became the standard for controllable text-to-image systems.

Core GAN Lineage

Year	Paper	Topic	Note
2014	Generative Adversarial Networks	Original GAN	Generator-discriminator minimax framework.
2015	Unsupervised Representation Learning with Deep Convolutional GANs	DCGAN	Practical convolutional GAN architecture.
2016	Improved Techniques for Training GANs	GAN stability	Feature matching, minibatch discrimination, and practical stabilization tricks.
2017	Wasserstein GAN	WGAN	Distribution-distance view for more stable training.
2017	Improved Training of Wasserstein GANs	WGAN-GP	Gradient penalty for more reliable Wasserstein GAN training.
2017	Least Squares Generative Adversarial Networks	LSGAN	Least-squares loss for more stable adversarial training.

Conditional, High-Resolution, and Style-Based GANs

Year	Paper	Topic	Note
2016	Image-to-Image Translation with Conditional Adversarial Networks	pix2pix	Paired image-to-image translation with conditional GANs.
2017	Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks	CycleGAN	Unpaired translation through cycle consistency.
2017	Progressive Growing of GANs	Progressive GAN	High-resolution synthesis through progressive training.
2018	Large Scale GAN Training for High Fidelity Natural Image Synthesis	BigGAN	Large-scale class-conditional GAN with strong ImageNet synthesis.
2018	A Style-Based Generator Architecture for GANs	StyleGAN	Style-based generator and controllable latent directions.
2019	Analyzing and Improving the Image Quality of StyleGAN	StyleGAN2	Removes artifacts and improves high-quality synthesis.
2021	Alias-Free Generative Adversarial Networks	StyleGAN3	Improves equivariance and removes texture sticking artifacts.

Reading Path

Step	Read
1	Original GAN and DCGAN.
2	Improved GANs, WGAN, and WGAN-GP for stability.
3	pix2pix and CycleGAN for conditional translation.
4	Progressive GAN, BigGAN, StyleGAN, StyleGAN2, and StyleGAN3 for image quality and style control.