Image Synthesis via Semantic Composition

TL;DR

This paper introduces a new image synthesis method that uses semantic layouts and appearance correlations to generate realistic images with improved global structure and detail, employing a dynamic weighted network.

Contribution

It presents a novel approach combining semantic composition with a dynamic network for enhanced image synthesis, emphasizing appearance-based dependencies.

Findings

Achieves superior qualitative image quality

Outperforms existing methods quantitatively on benchmarks

Strengthens semantic relevance in generated images

Abstract

In this paper, we present a novel approach to synthesize realistic images based on their semantic layouts. It hypothesizes that for objects with similar appearance, they share similar representation. Our method establishes dependencies between regions according to their appearance correlation, yielding both spatially variant and associated representations. Conditioning on these features, we propose a dynamic weighted network constructed by spatially conditional computation (with both convolution and normalization). More than preserving semantic distinctions, the given dynamic network strengthens semantic relevance, benefiting global structure and detail synthesis. We demonstrate that our method gives the compelling generation performance qualitatively and quantitatively with extensive experiments on benchmarks.