Material Editing Using a Physically Based Rendering Network

TL;DR

This paper introduces a physically based rendering network that predicts intrinsic image properties from a single image to enable realistic material editing, leveraging a differentiable rendering layer for end-to-end training.

Contribution

It presents a novel end-to-end network architecture with a differentiable rendering layer that accurately disentangles physical properties for material editing in images.

Findings

The network successfully predicts shape, illumination, and material from a single image.

It enables realistic and visually plausible material editing examples.

The approach outperforms existing methods in disentangling intrinsic properties.

Abstract

The ability to edit materials of objects in images is desirable by many content creators. However, this is an extremely challenging task as it requires to disentangle intrinsic physical properties of an image. We propose an end-to-end network architecture that replicates the forward image formation process to accomplish this task. Specifically, given a single image, the network first predicts intrinsic properties, i.e. shape, illumination, and material, which are then provided to a rendering layer. This layer performs in-network image synthesis, thereby enabling the network to understand the physics behind the image formation process. The proposed rendering layer is fully differentiable, supports both diffuse and specular materials, and thus can be applicable in a variety of problem settings. We demonstrate a rich set of visually plausible material editing examples and provide an extensive comparative study.