Fourier-CPPNs for Image Synthesis

TL;DR

This paper introduces Fourier-CPPNs, an extension of CPPNs that explicitly model frequency information to generate images with richer detail and less smoothness, enhancing visual quality.

Contribution

The paper presents Fourier-CPPNs, a novel architecture that incorporates frequency modeling into CPPNs for improved image synthesis detail.

Findings

Fourier-CPPNs produce images with more high-frequency detail.

They outperform traditional CPPNs in visual richness.

The method enhances creative image generation.

Abstract

Compositional Pattern Producing Networks (CPPNs) are differentiable networks that independently map (x, y) pixel coordinates to (r, g, b) colour values. Recently, CPPNs have been used for creating interesting imagery for creative purposes, e.g., neural art. However their architecture biases generated images to be overly smooth, lacking high-frequency detail. In this work, we extend CPPNs to explicitly model the frequency information for each pixel output, capturing frequencies beyond the DC component. We show that our Fourier-CPPNs (F-CPPNs) provide improved visual detail for image synthesis.