What is a GAN?
In this section, we will introduce GANs and briefly discuss the evolution and progression of this particular data generation method. Then, we will explain the standard architecture of a typical GAN and how they work.
The concept of GANs was introduced in the 2014 paper Generative Adversarial Networks (https://arxiv.org/abs/1406.2661), by Ian J. Goodfellow and his research team. In the same year, conditional GANs were introduced, allowing us to generate more customizable synthetic data. Then, Deep Convolutional GANs (DCGANs) were suggested in 2015, which facilitated the generation of high-resolution images. After that, CycleGANs were proposed in 2017 for unsupervised image-to-image translation tasks. This opened the door for enormous applications such as domain adaptation. StyleGAN was introduced in 2019, bringing GANs to new fields such as art and fashion.
GANs have also been showing impressive progress in the field of video synthesis. In fact, the recent work...
