“Crop outs” in collaboration with StyleGAN2-ADA-PyTorch

Inspired by Matisse’s cut-outs series and in an attempt to generate images that surpass the quintessential GAN-aesthetic, I trained my own StyleGAN2-ADA model on a handpicked dataset of abstract prints with colorful minimal shapes and managed to make it all work with the PyTorch implementation.

These prints were generated by a custom StyleGAN2-ADA-PyTorch model that I trained on a handpicked dataset of abstract art prints and curated into the “crop outs” series (a play on Matisse’s “cut outs” which some images resemble).

The idea was to see if I could generate visuals using GAN technologies without it having the quintessential GAN aesthetic that we see most of the time. At the same time it’s a commentary on (the lack of) originality in abstract art prints.

A handful of its designs were printed and distributed as postcards during NDSM Open.

“Crop outs”in collaboration with StyleGAN2-ADA-PyTorch

Inspired by Matisse’s cut-outs series and in an attempt to generate images that surpass the quintessential GAN-aesthetic, I trained my own StyleGAN2-ADA model on a handpicked dataset of abstract prints with colorful minimal shapes and managed to make it all work with the PyTorch implementation.