Diversified Texture Synthesis with Feed-forward Networks

TL;DR

This paper introduces a deep generative feed-forward network that efficiently synthesizes diverse textures within a single model, overcoming limitations of previous methods in generality, diversity, and visual quality.

Contribution

The work presents a novel single network architecture capable of synthesizing multiple textures and interpolating between them, with techniques to improve convergence and diversity.

Findings

Successfully synthesizes a large variety of textures

Enables meaningful interpolation between textures

Improves diversity and visual quality of generated textures

Abstract

Recent progresses on deep discriminative and generative modeling have shown promising results on texture synthesis. However, existing feed-forward based methods trade off generality for efficiency, which suffer from many issues, such as shortage of generality (i.e., build one network per texture), lack of diversity (i.e., always produce visually identical output) and suboptimality (i.e., generate less satisfying visual effects). In this work, we focus on solving these issues for improved texture synthesis. We propose a deep generative feed-forward network which enables efficient synthesis of multiple textures within one single network and meaningful interpolation between them. Meanwhile, a suite of important techniques are introduced to achieve better convergence and diversity. With extensive experiments, we demonstrate the effectiveness of the proposed model and techniques for synthesizing a large number of textures and show its applications with the stylization.