Spatially Controllable Image Synthesis with Internal Representation Collaging

TL;DR

This paper introduces a CNN-based image editing technique that enables spatially controllable modifications of images by manipulating internal feature representations within trained GAN models, allowing for flexible semantic editing of real and artificial images.

Contribution

The paper proposes two novel methods, spatial conditional batch normalization and feature-blending, for spatially controllable image editing within GANs, applicable to various datasets and models.

Findings

Effective editing of real and artificial images demonstrated.

Methods work with any GAN with conditional normalization layers.

Code available for reproducibility.

Abstract

We present a novel CNN-based image editing strategy that allows the user to change the semantic information of an image over an arbitrary region by manipulating the feature-space representation of the image in a trained GAN model. We will present two variants of our strategy: (1) spatial conditional batch normalization (sCBN), a type of conditional batch normalization with user-specifiable spatial weight maps, and (2) feature-blending, a method of directly modifying the intermediate features. Our methods can be used to edit both artificial image and real image, and they both can be used together with any GAN with conditional normalization layers. We will demonstrate the power of our method through experiments on various types of GANs trained on different datasets. Code will be available at https://github.com/pfnet-research/neural-collage.