I’m trying to build a GAN-model with a context vector as additional input, which should use RNN-layers for generating MNIST data.
For the Generator I want to slice the noise vector into four pieces and it should generate MNIST data in the same way.
For the Discriminator I want to do the same. The input should be sliced into four pieces.
I have a conditional GAN model that works not that well, but it works There is some work with the parameters to do.
Do you have any ideas or example models for a conditional GAN with RNNs or for a GAN with RNNs?
I drowned a lots of hours the last days to get by CGAN to become a CGAN with RNNs, but it’s not working. I’m missing some ideas, how I can realize the sliced input vector in addition to my context vector and how I can integrate the sliced input into the forward function.
Thank you for your help