search

gordicaleksa / pytorch-neural-style-transfer

Reconstruction of the original paper on neural style transfer (Gatys et al.). I've additionally included reconstruction scripts which allow you to reconstruct only the content or the style of the image - for better understanding of how NST works.
https://youtube.com/c/TheAIEpiphany
MIT License
366 stars 78 forks source link

Style Transfer usage for Generative Art #10

Open garricn opened 2 years ago

garricn commented 2 years ago

Would it be possible to use this or a similar library to - instead of transferring a style to another image - build an ML model that learns the style characteristics (color, shapes, objects, etc) of a collection of inputs (e.g., N number of similar art pieces by a particular artist) which can then be used to output Y number of unique generative images? In other words, the algo transfers the style from the physical to the digital as a set of configurable parameters for generating art. This would allow non-technical artists to enter the generative space.

CycloneRing commented 7 months ago

You can use Stable Diffusion and train a LoRA (Low-RANK) for such purpose.

Andrewthekid commented 1 month ago

Hi, same question here, do you know any repository for similar applications. Maybe not with a text input, instead, we input an image and ouput an image with a fixed style?