Geometry-Consistent Neural Shape Representation with Implicit Displacement Fields

TL;DR

This paper introduces implicit displacement fields, a new neural shape representation that decomposes 3D geometry into a smooth base surface and high-frequency details, enhancing reconstruction and transfer capabilities.

Contribution

It proposes an unsupervised, frequency-hierarchical neural representation for detailed 3D shapes based on displacement mapping, improving stability and generalization.

Findings

Superior shape reconstruction quality

Enhanced detail transfer capabilities

Improved training stability and generalizability

Abstract

We present implicit displacement fields, a novel representation for detailed 3D geometry. Inspired by a classic surface deformation technique, displacement mapping, our method represents a complex surface as a smooth base surface plus a displacement along the base's normal directions, resulting in a frequency-based shape decomposition, where the high frequency signal is constrained geometrically by the low frequency signal. Importantly, this disentanglement is unsupervised thanks to a tailored architectural design that has an innate frequency hierarchy by construction. We explore implicit displacement field surface reconstruction and detail transfer and demonstrate superior representational power, training stability and generalizability.