Semantic Image Inpainting with Deep Generative Models

TL;DR

This paper introduces a novel semantic image inpainting method that leverages deep generative models to fill large missing regions by conditioning on available data, outperforming existing techniques in realism and accuracy.

Contribution

The proposed method enables inpainting of large missing regions without requiring specific training on hole structures, using a search for the closest encoding in the latent space.

Findings

Successfully predicts large missing regions with high photorealism

Outperforms state-of-the-art methods in experiments

Works irrespective of missing content structure

Abstract

Semantic image inpainting is a challenging task where large missing regions have to be filled based on the available visual data. Existing methods which extract information from only a single image generally produce unsatisfactory results due to the lack of high level context. In this paper, we propose a novel method for semantic image inpainting, which generates the missing content by conditioning on the available data. Given a trained generative model, we search for the closest encoding of the corrupted image in the latent image manifold using our context and prior losses. This encoding is then passed through the generative model to infer the missing content. In our method, inference is possible irrespective of how the missing content is structured, while the state-of-the-art learning based method requires specific information about the holes in the training phase. Experiments on three datasets show that our method successfully predicts information in large missing regions and achieves pixel-level photorealism, significantly outperforming the state-of-the-art methods.