Hacker News Reader

Comment by muhehe | Hacker News Reader

There is a lot "X doesn't exist" sites. Is there any tutorial how to make my own?

First, you have to collect a few thousand images of the same thing (maybe more or less depending on how complex your thing is or how good the results should be). Then, you train a generative adversarial neural network on those images to generate new images. https://github.com/NVlabs/stylegan2-ada-pytorch works quite well. https://github.com/NVlabs/stylegan3 is supposedly even better, but I did not try it yet.

This is an excellent summary. Thanks!

My follow-up question is whether anything will work with an M1 yet. I'm guessing nothing from NVlabs. :-)

I would suggest getting a colab pro account. If you only use it for a month that's only $10 but it's plenty for training lots of these models (maybe 1-3 days per dataset depending on resolution and dataset diversity). https://colab.research.google.com/github/dvschultz/ml-art-co...

Those are implemented in pytorch which can use various backend including CPU-based ones.

https://pytorch.org/docs/stable/backends.html

Whether or not it would be prohibitively slow is another question, however.

This tutorial does not exist.

General adversarial networks

https://en.wikipedia.org/wiki/Generative_adversarial_network

not a tutorial, but here is how they did it: https://docs.google.com/presentation/d/18t0k17k0Xxo0BXMdjl6m...