From A to Z: Supervised Transfer of Style and Content Using Deep Neural Network Generators

TL;DR

This paper introduces a neural network architecture based on a modified variational autoencoder that can generate stylistically similar images from a single input, effectively separating style from content for single-image analogy tasks.

Contribution

It presents a novel supervised training method for single-image analogies using a structured similarity objective within a VAE framework.

Findings

Achieved 22.4% lower dissimilarity in font generation compared to state-of-the-art.

Demonstrated effective style-content separation in image generation.

Validated on a challenging font synthesis task.

Abstract

We propose a new neural network architecture for solving single-image analogies - the generation of an entire set of stylistically similar images from just a single input image. Solving this problem requires separating image style from content. Our network is a modified variational autoencoder (VAE) that supports supervised training of single-image analogies and in-network evaluation of outputs with a structured similarity objective that captures pixel covariances. On the challenging task of generating a 62-letter font from a single example letter we produce images with 22.4% lower dissimilarity to the ground truth than state-of-the-art.