Hierarchical Semantic Regularization of Latent Spaces in StyleGANs

TL;DR

This paper introduces a Hierarchical Semantic Regularizer (HSR) that aligns generator representations with pretrained network features, enhancing the naturalness and attribute editing linearity of StyleGAN-generated images.

Contribution

The paper proposes HSR to improve StyleGANs by aligning hierarchical representations with pretrained features, resulting in more natural images and better attribute editing linearity.

Findings

16.19% reduction in Perceptual Path Length (PPL)

Improved attribute linearity in editing tasks

Enhanced naturalness of generated images

Abstract

Progress in GANs has enabled the generation of high-resolution photorealistic images of astonishing quality. StyleGANs allow for compelling attribute modification on such images via mathematical operations on the latent style vectors in the W/W+ space that effectively modulate the rich hierarchical representations of the generator. Such operations have recently been generalized beyond mere attribute swapping in the original StyleGAN paper to include interpolations. In spite of many significant improvements in StyleGANs, they are still seen to generate unnatural images. The quality of the generated images is predicated on two assumptions; (a) The richness of the hierarchical representations learnt by the generator, and, (b) The linearity and smoothness of the style spaces. In this work, we propose a Hierarchical Semantic Regularizer (HSR) which aligns the hierarchical representations learnt by the generator to corresponding powerful features learnt by pretrained networks on large amounts of data. HSR is shown to not only improve generator representations but also the linearity and smoothness of the latent style spaces, leading to the generation of more natural-looking style-edited images. To demonstrate improved linearity, we propose a novel metric - Attribute Linearity Score (ALS). A significant reduction in the generation of unnatural images is corroborated by improvement in the Perceptual Path Length (PPL) metric by 16.19% averaged across different standard datasets while simultaneously improving the linearity of attribute-change in the attribute editing tasks.