Visual Indeterminacy in GAN Art

TL;DR

This paper investigates how GAN-generated art often features visual indeterminacy, where images appear real but lack coherent spatial interpretation, highlighting a key aspect of modern GAN art.

Contribution

It introduces the concept of visual indeterminacy in GAN art and analyzes its origins as a consequence of the model's synthesis process.

Findings

GAN models tend to produce indeterminate images.

Indeterminacy is common in modern representational art.

Indeterminacy arises from the combination of diverse object classes in synthesis.

Abstract

This paper explores visual indeterminacy as a description for artwork created with Generative Adversarial Networks (GANs). Visual indeterminacy describes images which appear to depict real scenes, but, on closer examination, defy coherent spatial interpretation. GAN models seem to be predisposed to producing indeterminate images, and indeterminacy is a key feature of much modern representational art, as well as most GAN art. It is hypothesized that indeterminacy is a consequence of a powerful-but-imperfect image synthesis model that must combine general classes of objects, scenes, and textures.