2
$\begingroup$

I'm training a DCGAN model on a 320x320 dataset of images and after an hour of training the generator started to generate (on the same latent space noise as during training) images that are identical to the dataset. For example, if my dataset is images of cars, I should expect to see unexisting designs of cars, right? Am I understanding this wrong? I know this is a very general question but I was wondering if this is what should happen and if I should try on different latent space values and then see proper results and not just copies of my dataset?

$\endgroup$

1 Answer 1

2
$\begingroup$

It might be that your dataset of images is to small. Your discriminative network might hardlearn these images at which point your generative network can only produce good images if it copies the same images of your dataset.

$\endgroup$
3
  • $\begingroup$ Exactly. My dataset consists of only 45 images. What can I do? $\endgroup$ Mar 6, 2020 at 17:45
  • $\begingroup$ I believe that question is answered here. $\endgroup$ Mar 6, 2020 at 18:16
  • $\begingroup$ Thanks! I'll look into it. $\endgroup$ Mar 6, 2020 at 18:21

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .