I'm looking a references (papers / works) for synthetic image generation from small datasets. By small dataset, I mean 10-50 images.

I assume, that the best approaches should be based GAN (cGAN ?) or Diffusion Model.

Meanwhile I've found the following:

Towards Faster and Stabilized GAN Training for High-fidelity Few-shot Image Synthesis

Image Generation From Small Datasets via Batch Statistics Adaptation


1 Answer 1


The references you stated indeed are the right way to go regarding small dataset image synthesis.

I'd research the space of few-shot image synthesis for what is out there, but something along the line of StyleGAN2 is a logical direction to look into.

In addition, I'd like to state specifically that training a synthesizer on only 10-50 images is always going to be very difficult. Is your goal going to be to make slight adaptations to the images? Or to synthesize completely new images? Image data is very high dimensional, and generalising high dimensional data from very few samples is near impossible, no matter what technique you employ.

  • $\begingroup$ I'm partially agree. It was correct few years ago, that it's very challenging to create synthetic data from 10-15 images. Recently there are works of creating an synthetic images even from single image. For example SinGan paper: tamarott.github.io/SinGAN.htm. $\endgroup$
    – Michael D
    Apr 17 at 11:33
  • $\begingroup$ Did not know of that paper, nice addition! $\endgroup$ Apr 17 at 11:47
  • $\begingroup$ It got best paper award at ICCV 2019. $\endgroup$
    – Michael D
    Apr 17 at 12:16

You must log in to answer this question.

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