Real Image Inversion via Segments

TL;DR

This paper introduces a segmentation-based method for editing real images with GANs, enabling more precise local modifications that preserve the original image structure and natural appearance.

Contribution

The approach segments images to improve the accuracy of latent code estimation and local editing, enhancing the natural look of edited images compared to previous methods.

Findings

Improved local editing accuracy with segmentation.

Better preservation of original image structures.

Enhanced natural appearance of edited images.

Abstract

In this short report, we present a simple, yet effective approach to editing real images via generative adversarial networks (GAN). Unlike previous techniques, that treat all editing tasks as an operation that affects pixel values in the entire image in our approach we cut up the image into a set of smaller segments. For those segments corresponding latent codes of a generative network can be estimated with greater accuracy due to the lower number of constraints. When codes are altered by the user the content in the image is manipulated locally while the rest of it remains unaffected. Thanks to this property the final edited image better retains the original structures and thus helps to preserve natural look.