Generative Adversarial Networks - Matchue GANs

This is the first installment of Matchue GANs, where I will be going through many different types of Generative Adversarial Networks. I will walk you through how they work, how to program them in Keras, and show you some of the results!
This episode was all about just regular GANs, or vanilla GANs as I call them. I work through the logic behind GANs, and implement python code for a GAN to generate images of flowers.

[@Matchue624]

Here is the link to the repository, as well as the data used in this episode: github.com/manicman1999/Matchue-Vanilla-GAN.
I hope you all enjoyed the video!

[@kobruh9642]

So far so good!

[@akulvohra5344]

Nice video. Good explanation, would love to see more.

[@aizek1396]

That is awesome, please make more of this.

[@nijeesh4all]

Subscribed and waiting for more

[@teidenzero]

Fantastic, keep it up!

[@aliqsaeed2878]

Hi Matchue,
Your coding is very understandable and clear (thank for simplifying issues). However, i couldn't understand how did you update the gradients of G and D? ... as i saw in other's implementations, they updated G and D gradients in the following lines of code :
gradients_of_generator = gen_tape.gradient(gen_loss, generator.trainable_variables)
gradients_of_discriminator = disc_tape.gradient(disc_loss, discriminator.trainable_variables)

generator_optimizer.apply_gradients(zip(gradients_of_generator, generator.trainable_variables))
discriminator_optimizer.apply_gradients(zip(gradients_of_discriminator, discriminator.trainable_variables))
Please, i am looking forward for your reply.
Best regards

[@teidenzero]

quick question: if I wanted to generate 512x512 images instead of 64x64, would my latent input and dense layer look like this?
#Latent input
latent_input = Input([512])
#Reshape to 4x4x512
x = Dense(4*4*512, activation = 'relu')(latent_input)
x = Reshape([4, 4, 512])(x)

[@Lazy84.20]

I would really like you to increase the font size. I’m watching this on my iPad and I can’t read the text very well. Thanks!

[@rishikksh20]

Go ahead

[@benb7528]

I was trying to run your notebook but I think you overwrote your flower data with mountains of 128X128

[@Matchue624]

Thanks for noticing! I will fix this right away!