Trained Latent Space Navigation to Prevent Lack of Photorealism in Generated Images on Style-based Models

TL;DR

This paper introduces an unsupervised method for navigating the latent space of StyleGAN models to maintain photorealism during image manipulation, addressing a common challenge in generative image quality.

Contribution

The proposed method identifies local latent subspaces to enable realistic image editing while preserving photorealism, improving upon traditional latent code manipulation techniques.

Findings

Images within local latent subspaces retain photorealism after manipulation.

Method effectively applies to various style-based models.

Enhances latent code optimization for high-quality image generation.

Abstract

Recent studies on StyleGAN variants show promising performances for various generation tasks. In these models, latent codes have traditionally been manipulated and searched for the desired images. However, this approach sometimes suffers from a lack of photorealism in generated images due to a lack of knowledge about the geometry of the trained latent space. In this paper, we show a simple unsupervised method that provides well-trained local latent subspace, enabling latent code navigation while preserving the photorealism of the generated images. Specifically, the method identifies densely mapped latent spaces and restricts latent manipulations within the local latent subspace. Experimental results demonstrate that images generated within the local latent subspace maintain photorealism even when the latent codes are significantly and repeatedly manipulated. Moreover, experiments show that the method can be applied to latent code optimization for various types of style-based models. Our empirical evidence of the method will benefit applications in style-based models.