Multiview Neural Surface Reconstruction by Disentangling Geometry and Appearance

TL;DR

This paper presents a neural network architecture for multiview 3D surface reconstruction that disentangles geometry and appearance, achieving high-fidelity results on real-world images with varying materials and lighting.

Contribution

The authors introduce a neural network that jointly learns geometry, camera parameters, and rendering, representing geometry as a neural level-set and modeling diverse lighting and materials.

Findings

State-of-the-art 3D surface reconstruction quality

High fidelity and detailed surface outputs

Effective handling of real-world noisy data

Abstract

In this work we address the challenging problem of multiview 3D surface reconstruction. We introduce a neural network architecture that simultaneously learns the unknown geometry, camera parameters, and a neural renderer that approximates the light reflected from the surface towards the camera. The geometry is represented as a zero level-set of a neural network, while the neural renderer, derived from the rendering equation, is capable of (implicitly) modeling a wide set of lighting conditions and materials. We trained our network on real world 2D images of objects with different material properties, lighting conditions, and noisy camera initializations from the DTU MVS dataset. We found our model to produce state of the art 3D surface reconstructions with high fidelity, resolution and detail.