Code generated in the video can be downloaded from here:
[ Ссылка ]
Conditional Generative Adversarial Network cGAN
A GAN model generates a random image from the domain.
The relationship between points in the latent space and the generated images is hard to map.
A GAN can be trained so that both the generator and the discriminator models are conditioned on the class label (or other modalities).
As a result, the trained generator model can be used to generate images of a given type using the class label (or other condition).
GAN can be conditioned using other image modalities (image to image translation).
The conditioning is performed by feeding the class label into both the discriminator and generator as additional input layer.
A few applications:
Image-to-Image Translation: Pix2Pix GAN
CycleGAN: Transform images from one set into images that could belong to another set.
Super-resolution: Increase the resolution of images, adding detail where necessary to fill in blurry areas.
Text-to-Image Synthesis: Take text as input and produce images as described by the text.
