Image Synthesis via Semantic Composition
This work addresses image synthesis for computer vision applications, but it appears incremental as it builds on existing methods with a novel dynamic network approach.
The paper tackles the problem of synthesizing realistic images from semantic layouts by establishing dependencies between regions based on appearance correlation, resulting in improved generation performance as demonstrated through qualitative and quantitative experiments on benchmarks.
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.