Transposer: Universal Texture Synthesis Using Feature Maps as Transposed Convolution Filter

TL;DR

Transposer introduces a novel texture synthesis method using feature maps as transposed convolution filters, enabling fast, generalizable, and high-quality synthesis of both regular and irregular textures, including large outputs.

Contribution

It proposes a new approach that treats encoded feature maps as transposed convolution filters, allowing for efficient, one-pass synthesis of unseen textures without retraining.

Findings

Achieves state-of-the-art quality in texture synthesis

Enables real-time synthesis of unseen textures

Supports diverse and arbitrarily large texture outputs

Abstract

Conventional CNNs for texture synthesis consist of a sequence of (de)-convolution and up/down-sampling layers, where each layer operates locally and lacks the ability to capture the long-term structural dependency required by texture synthesis. Thus, they often simply enlarge the input texture, rather than perform reasonable synthesis. As a compromise, many recent methods sacrifice generalizability by training and testing on the same single (or fixed set of) texture image(s), resulting in huge re-training time costs for unseen images. In this work, based on the discovery that the assembling/stitching operation in traditional texture synthesis is analogous to a transposed convolution operation, we propose a novel way of using transposed convolution operation. Specifically, we directly treat the whole encoded feature map of the input texture as transposed convolution filters and the features' self-similarity map, which captures the auto-correlation information, as input to the transposed convolution. Such a design allows our framework, once trained, to be generalizable to perform synthesis of unseen textures with a single forward pass in nearly real-time. Our method achieves state-of-the-art texture synthesis quality based on various metrics. While self-similarity helps preserve the input textures' regular structural patterns, our framework can also take random noise maps for irregular input textures instead of self-similarity maps as transposed convolution inputs. It allows to get more diverse results as well as generate arbitrarily large texture outputs by directly sampling large noise maps in a single pass as well.