tamarott / SinGAN

Official pytorch implementation of the paper: "SinGAN: Learning a Generative Model from a Single Natural Image"
https://tamarott.github.io/SinGAN.htm
Other
3.31k stars 611 forks source link

Changing the type of noise used during training #135

Closed alexblnn closed 3 years ago

alexblnn commented 3 years ago

Hello, I am trying to use SinGAN for a denoising task on medical images. The noise in these images resembles salt and pepper noise rather than gaussian noise (which is used to train SinGAN if i'm not mistaken).

Do you think it would be possible to modify a part of your code to change the type of noise used during training? If so, which part should I modify? I tried to investigate this on my own to no success.

Thank you for this great paper!

alexblnn commented 3 years ago

Found the answer: need to modify the function "generate_noise" in functions.py