Exploring creative AI techniques for art generation
With the rise of GANs, humans are able to generate photorealistic images based on existing source material. With these developments, generating art using AI has been on the rise. Even on the NFT market, projects are advertising with AI-generated art. This project explores how we can advance the generation of art with GANs and other ConvNet based methods, focusing on inciting creativity rather than just photorealism.
I selected source material (abstract art images in the public domain) and trained a GAN (architecture and training method resembled StyleGAN2-ADA
The main limitation is computational resources. Training GANs is computationally intensive, and current GPU prices make it challenging to acquire the necessary hardware for large-scale experimentation.