Generative Adversarial Networks
lastmod 2026-05-07created 2019-09-2643 edits
A generative adversarial network (GAN) is an artificial intelligence algorithm used in unsupervised learning, implemented by a system of two neural networks competing with each other within a zero-sum game framework. The concept was introduced in 2014 by Ian Goodfellow (Ian. j. Goodfellow).
There are several variants, such as Conditional GAN (Conditional Generative Adversarial Net, CGAN) and Cycle GAN.
Abstractor
We propose a new framework for estimating generative models via an adversarial process, in which we simultaneously train two models: a generative model G that captures the data distribution, and a discriminative model D that estimates the probability that a sample came from the training data rather than G. The training procedure for G is to maximize the probability of D making a mistake. This framework corresponds to a minimax two-player game. In the space of arbitrary functions G and D, a unique solution exists, with G recovering the training data distribution and D equal to 1/2 everywhere. In the case where G and D are defined by multilayer perceptrons, the entire system can be trained with backpropagation. There is no need for any Markov chains or unrolled approximate inference networks during either training or generation of samples. Experiments demonstrate the potential of the framework through qualitative and quantitative evaluation of the generated samples.
Documentation
- Generative Adversarial Networks
- https://arxiv.org/abs/1406.2661
- Progressive Growing of GANs for Improved Quality, Stability, and Variation
- https://arxiv.org/pdf/1710.10196.pdf
-
1710.10196.pdf - https://research.nvidia.com/publication/2017-10_Progressive-Growing-of
- https://github.com/tkarras/progressive_growing_of_gans
- https://www.reddit.com/r/MachineLearning/comments/9ybstu/r_tdls_progressive_growing_of_gans/
Derived techniques
Audio & Music
- GANSynth - Making music with GANs
- GANSynth is an algorithm for synthesizing audio with generative adversarial networks. The details can be found in the ICLR 2019 Paper. It achieves better audio quality than a standard WaveNet baselines on the NSynth Dataset, and synthesizes audio thousands of times faster.
- WaveGAN - Adversarial Audio Synthesis - https://arxiv.org/abs/1802.04208
- MusicGAN - https://github.com/HStuart18/tfworldhackathon
- Voicebox - Meta releases Voicebox, a generative AI model for speech
- AudioCraft - Meta releases generative AI for audio as open source
Image
- DefGAN
- AnoGAN
- pix2pix - Image-to-Image Translation with Conditional Adversarial Networks
- CycleGAN - Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks
- BicycleGAN - Toward Multimodal Image-to-Image Translation
- Dark Model Adaptation
- StyleGAN - A Style-Based Generator Architecture for Generative Adversarial Networks
- StyleGAN2
- StyleGAN3
- StackGAN - Text to Photo-realistic Image Synthesis with Stacked Generative Adversarial Networks
- 3D ShapeNet - Train a neural network on the dataset of this name to generate realistic shapes.
- DCGAN - Generate anime characters.
- SRGAN - Implement a neural network to generate high-resolution images.
- Age-cGAN - Lets a neural network recognize a person's face well even as the person ages.
- A Neural Algorithm of Artistic Style
- Noise2Noise - Learning Image Restoration without Clean Data
- SinGAN - Learning a Generative Model from a Single Natural Image
- WGAN
- NVIDIA Canvas - Paint realistic pictures with AI
- DragGAN - Implementation of DragGAN: Interactive Point-based Manipulation on the Generative Image Manifold
- Open Generative AI
Video
- Gen-2 - Runway releases "Gen-2", a generative AI for video
- Goku - ByteDance's flow-based video generation model
Colorization
- Github - aleju/colorizer
- This project uses GANs (generative adversarial networks) to add color to black and white images.
Apartment Building Design
- ArchiGAN - a Generative Stack for Apartment Building Design
- https://devblogs.nvidia.com/archigan-generative-stack-apartment-building-design/
Simulation
- Genesis - A Generative and Universal Physics Engine for Robotics and Beyond
UI/GUI
- v0 - A generative UI system built by Vercel
Search Engine
- phind - A generative AI search engine for developers
Game
- Korean: 스몰빌 (Smallville) - A virtual village inhabited by AI agents
Projects
- 4 Impressive GAN Libraries Every Data Scientist Should Know!
- Top Libraries For Quick Implementation Of GANs
List of libraries:
- TensorFlow-GAN (TF-GAN)
- Torch-GAN (torchgan)
- Keras-GAN
- Mimicry
- IBM-GAN toolkit
- PaddleGAN
- GAN Lab
- pygan
- VeGANs
- StudioGAN
- StyleSwin - [CVPR 2022] StyleSwin: Transformer-based GAN for High-resolution Image Generation
- Karlo - Kakao Brain's image generation model
6 GAN architectures you really should know (March 21, 2022)
The most popular GAN formulations are:
- Translating images from one domain to another (CycleGAN)
- Generating images from text descriptions (text-to-image)
- Generating high-resolution images (ProgressiveGAN), and more.
This article explains the most popular GAN architectures, in particular the 6 architectures you should know to cover a wide range of GANs (Generative Adversarial Networks).
See also
Favorite site
- Paper group/paper seminar - NIPS2016 Tutorials - Genertive Adeversarial Networks
- Advanced GANs 1
- Generating custom photo-realistic faces using AI (Generating realistic realistic faces using AI; controlled image synthesis and editing with the new TL-GAN model)
- SAMSUNG SDS - New AI technology GAN ① - AI that learns on its own (basic theory)
- SAMSUNG SDS - New AI technology GAN ② - The concept and understanding of GAN (basic theory)
- [Recommended] GAN — Some cool applications of GANs. 2 (introduces a list of applications you can build with GANs)
- Github - The GAN Zoo - A list of all named GANs!
- Github - PyTorch-GAN - PyTorch implementations of Generative Adversarial Networks
- Dissecting the ‘synthetic data’ that makes accurate AI models | NVIDIA Blog
- List of synthetic data startups and companies — 2021 (Synthetic data)
- McKinsey publishes a report on the economic potential of generative AI (66p/PDF | GeekNews
Article
- Reflections on generative AI's uncanny valley | GeekNews
- [Original] Reckoning with generative AI’s uncanny valley | Thoughtworks
- Existing mental models and concepts of generative AI are a fundamental design problem, not a secondary issue that can be ignored
- Generative AI's uncanny valley is not a problem to fix but an opportunity to reassess what we really want and expect from this technology
- [Original] The Impact of Generative AI on Critical Thinking: Self-Reported Reductions in Cognitive Effort and Confidence Effects From a Survey of Knowledge Workers - Microsoft Research
- GenAI improves the efficiency of knowledge work but carries the risk of reduced critical thinking
- Designing tools that help make critical thinking a habit and sustain it is important
- Using GenAI effectively requires overall UX consideration, including user training and feedback loops (data flywheel)
- [Original] GenAI Image Showdown
- An experimental project that tests several image generation AI models with the same prompts and evaluates their accuracy, creativity, and consistency
- A total of 14 models took part, including OpenAI 4o, Gemini 2.5 Flash, Imagen 4, Seedream 4, FLUX.1 dev, and Midjourney v7