hindupuravinash/the-gan-zoo

the-gan-zoo is a long, hand-curated list of named GANs. A GAN, short for generative adversarial network, is a kind of machine learning model where two neural networks compete: one tries to generate fake data that looks real, and the other tries to tell real data from fake. GANs are used to produce images, video, audio, and other content. New variations of the basic idea show up in research papers all the time, and researchers tend to give each variant a clever name. The README opens with a casual note from the author about how hard it is to keep up with the flood of new GAN papers, especially given the creative names. The repository started as a fun exercise to track them all. The header image is a play on a zoo poster, and there is a chart of the cumulative number of named GANs over time included in the repo. The core content is a flat list, alphabetised by name, of every GAN the author has come across. Each entry has the short name, the title of the paper, and a link to the paper, usually on arXiv. When code is available, a link to the GitHub repository is included as well. Names range from 3D-GAN for object shape generation, to AC-GAN for conditional image synthesis, to AnoGAN for anomaly detection, to AttnGAN for text-to-image generation. The list keeps going through hundreds of entries. The same data is also available in a tab-separated file called gans.tsv inside the repository, which the README points to as an alternative way to filter by year or search by title. The author welcomes contributions, asking people to either open a pull request against gans.tsv or file an issue if they spot a missing paper. For a non-technical reader, the repository is a research index, not a piece of software. You cannot run anything from it directly. Its value is as a reference: if you hear about a paper, or want to find existing work on a particular GAN application, you can scan this list to see which papers are out there and where to read them. The README also points to the author's weekly AI newsletter and Twitter for related coverage.