SurfaceNet: Adversarial SVBRDF Estimation from a Single Image

TL;DR

SurfaceNet is a novel GAN-based method that estimates detailed SVBRDF material properties from a single image, outperforming existing methods and working effectively on real data without supervision.

Contribution

The paper introduces SurfaceNet, a patch-based GAN that improves SVBRDF estimation from a single image, especially on real data, with high-resolution and detail recovery.

Findings

Outperforms existing SVBRDF reconstruction methods quantitatively and qualitatively.

Effectively generates high-quality reflectance maps from real images without supervision.

Demonstrates robustness across different illumination conditions.

Abstract

In this paper we present SurfaceNet, an approach for estimating spatially-varying bidirectional reflectance distribution function (SVBRDF) material properties from a single image. We pose the problem as an image translation task and propose a novel patch-based generative adversarial network (GAN) that is able to produce high-quality, high-resolution surface reflectance maps. The employment of the GAN paradigm has a twofold objective: 1) allowing the model to recover finer details than standard translation models; 2) reducing the domain shift between synthetic and real data distributions in an unsupervised way. An extensive evaluation, carried out on a public benchmark of synthetic and real images under different illumination conditions, shows that SurfaceNet largely outperforms existing SVBRDF reconstruction methods, both quantitatively and qualitatively. Furthermore, SurfaceNet exhibits a remarkable ability in generating high-quality maps from real samples without any supervision at training time.