Aesthetics and neural network image representations

TL;DR

This paper investigates how generative neural networks encode aesthetic and symbolic properties in images, revealing that parameter perturbations can produce artistic and symbolic visual representations without training on art data.

Contribution

It demonstrates that neural network parameter space inherently encodes aesthetic and symbolic features, emerging from the network's structure rather than explicit training on art.

Findings

Perturbations lead to artistic renditions of objects.

Modifying deep semantic parts yields symbolic visual representations.

Networks trained without art data still produce aesthetic features.

Abstract

We analyze the spaces of images encoded by generative neural networks of the BigGAN architecture. We find that generic multiplicative perturbations of neural network parameters away from the photo-realistic point often lead to networks generating images which appear as "artistic renditions" of the corresponding objects. This demonstrates an emergence of aesthetic properties directly from the structure of the photo-realistic visual environment as encoded in its neural network parametrization. Moreover, modifying a deep semantic part of the neural network leads to the appearance of symbolic visual representations. None of the considered networks had any access to images of human-made art.