Attribute2Image: Conditional Image Generation from Visual Attributes

TL;DR

This paper introduces a layered variational auto-encoder model that generates realistic, diverse images from visual attributes by disentangling foreground and background, enabling attribute-conditioned image reconstruction and completion.

Contribution

It presents a novel layered generative model with disentangled latent variables learned end-to-end for attribute-based image synthesis.

Findings

Generated images are realistic and diverse.

Model achieves excellent results in image reconstruction.

Disentangled representations facilitate attribute-conditioned tasks.

Abstract

This paper investigates a novel problem of generating images from visual attributes. We model the image as a composite of foreground and background and develop a layered generative model with disentangled latent variables that can be learned end-to-end using a variational auto-encoder. We experiment with natural images of faces and birds and demonstrate that the proposed models are capable of generating realistic and diverse samples with disentangled latent representations. We use a general energy minimization algorithm for posterior inference of latent variables given novel images. Therefore, the learned generative models show excellent quantitative and visual results in the tasks of attribute-conditioned image reconstruction and completion.