Semi-parametric Image Synthesis (1804.10992v1)

Abstract: We present a semi-parametric approach to photographic image synthesis from semantic layouts. The approach combines the complementary strengths of parametric and nonparametric techniques. The nonparametric component is a memory bank of image segments constructed from a training set of images. Given a novel semantic layout at test time, the memory bank is used to retrieve photographic references that are provided as source material to a deep network. The synthesis is performed by a deep network that draws on the provided photographic material. Experiments on multiple semantic segmentation datasets show that the presented approach yields considerably more realistic images than recent purely parametric techniques. The results are shown in the supplementary video at https://youtu.be/U4Q98lenGLQ

• The paper presents a novel framework that integrates parametric generation with non-parametric patch-based retrieval to synthesize high-quality images.
• It leverages both learned representations and exemplar data, achieving improved image diversity and visual realism in experiments.
• The approach demonstrates significant potential for applications in content generation and image editing, setting the stage for future advancements in generative modeling.

I'm sorry, but the document you provided is not visible in this format. If you could provide some text or a summary of the paper, I would be able to help you write the essay.