What is StyleGAN?

Course Q&A Generative Adversarial Networks (GANS) Build Better Generative Adversarial Networks
1

The lecture and the programming assignment did not touch on discriminator, but all were about generator.

Is StyleGAN just a type of generator that can work with all types of discriminators and be plugged in other GANs like DCGAN ?

2

Hi mc04xkf!
Hope you are doing well.
The generator in StyleGAN is designed to generate high-quality images with high fidelity and diversity. It is like, we are trying to improve the generator architecture starting from simple generators to more complex generators like that of the StyleGAN’s generator as we progress through the specialization. StyleGAN helps us to generate better or high-quality images so it’s very important to understand its generator’s architecture (this is where we have to make some major changes to generate better-quality images) and that’s why more focus is given to explaining the generator architecture.
Also, StyleGAN is not a type of generator - it is a type of GAN (which comprises both generator and discriminator)
Please go through the lectures carefully, it will be very interesting to learn about StyleGANs.
Have a great day!
Regards,
Nithin

3

thanks Nithin for your reply.

Are there any differences between the discriminator of DCGAN and StyleGAN ?

4

Well, it is not strictly necessary to have different discriminators for different GAN architectures. But the choice of discriminator architecture, loss functions etc can have a significant impact on the performance of the GAN. It need not necessarily be the same or different. We have to find it out which architecture will perform better with respect to our problem. Even if you see, and read the StyleGAN paper, the complete emphasis is on the generator as this is where they are bringing out the difference. Essentially at the end of the day, the role of the discriminator is just to identify if an image is real or fake, we can easily make the discriminators strong! but the real challenge is to make the generator capable of competing with the discriminator and to be able to generate high-quality images. Hope that you get the intuition behind it.If not feel free to ask your queries.
Have a great day!
Regards,
Nithin

1 Like