Progressive Growing GAN is an extension to the GAN training process that allows for the stable training of generator models that can output large high-quality images.<\/p>\n
It involves starting with a very small image and incrementally adding blocks of layers that increase the output size of the generator model and the input size of the discriminator model until the desired image size is achieved.<\/p>\n
This approach has proven effective at generating high-quality synthetic faces that are startlingly realistic.<\/p>\n
In this post, you will discover the progressive growing generative adversarial network for generating large images.<\/p>\n
After reading this post, you will know:<\/p>\n
- \n
- GANs are effective at generating sharp images, although they are limited to small image sizes because of model stability.<\/li>\n
- Progressive growing GAN is a stable approach to training GAN models to generate large high-quality images that involves incrementally increasing the size of the model during training.<\/li>\n
- Progressive growing GAN models are capable of generating photorealistic synthetic faces and objects at high resolution that are remarkably realistic.<\/li>\n<\/ul>\n
Discover how to develop DCGANs, conditional GANs, Pix2Pix, CycleGANs, and more with Keras in my new GANs book<\/a>, with 29 step-by-step tutorials and full source code.<\/p>\n
Let\u2019s get started.<\/p>\n
![\"A](\"https:\/\/machinelearningmastery.com\/wp-content\/uploads\/2019\/08\/A-Gentle-Introduction-to-Progressive-Growing-Generative-Adversarial-Networks.jpg\")
<\/p>\nA Gentle Introduction to Progressive Growing Generative Adversarial Networks
Photo by Sandrine N\u00e9el<\/a>, some rights reserved.<\/p>\n<\/div>\nOverview<\/h2>\n
This tutorial is divided into five parts; they are:<\/p>\n
- \n
- GANs Are Generally Limited to Small Images<\/li>\n
- Generate Large Images by Progressively Adding Layers<\/li>\n
- How to Progressively Grow a GAN<\/li>\n
- Images Generated by the Progressive Growing GAN<\/li>\n
- How to Configure Progressive Growing GAN Models<\/li>\n<\/ol>\n
GANs Are Generally Limited to Small Images<\/h2>\n
Generative Adversarial Networks, or GANs for short, are an effective approach for training deep convolutional neural network models for generating synthetic images.<\/p>\n
Training a GAN model involves two models: a generator used to output synthetic images, and a discriminator model used to classify images as real or fake, which is used to train the generator model. The two models are trained together in an adversarial manner, seeking an equilibrium.<\/p>\n
Compared to other approaches, they are both fast and result in crisp images.<\/p>\n
A problem with GANs is that they are limited to small dataset sizes, often a few hundred pixels and often less than 100-pixel square images.<\/p>\n
\n
GANs produce sharp images, albeit only in fairly small resolutions and with somewhat limited variation, and the training continues to be unstable despite recent progress.<\/p>\n<\/blockquote>\n
\u2014 Progressive Growing of GANs for Improved Quality, Stability, and Variation<\/a>, 2017.<\/p>\n
Generating high-resolution images is believed to be challenging for GAN models as the generator must learn how to output both large structure and fine details at the same time.<\/p>\n
The high resolution makes any issues in the fine detail of generated images easy to spot for the discriminator and the training process fails.<\/p>\n
\n
The generation of high-resolution images is difficult because higher resolution makes it easier to tell the generated images apart from training images \u2026<\/p>\n<\/blockquote>\n
\u2014 Progressive Growing of GANs for Improved Quality, Stability, and Variation<\/a>, 2017.<\/p>\n
Large images, such as 1024-pixel square images, also require significantly more memory, which is in relatively limited supply on modern GPU hardware compared to main memory.<\/p>\n
As such, the batch size that defines the number of images used to update model weights each training iteration must be reduced to ensure that the large images fit into memory. This, in turn, introduces further instability into the training process.<\/p>\n
\n
Large resolutions also necessitate using smaller minibatches due to memory constraints, further compromising training stability.<\/p>\n<\/blockquote>\n