Multi-Class Boundary Extraction from Implicit Representations

TL;DR

This paper introduces a novel 2D boundary extraction algorithm for multi-class implicit representations that ensures topological correctness and water-tightness, addressing a gap in existing single-class surface extraction methods.

Contribution

It presents the first multi-class boundary extraction method from implicit neural representations with guarantees on topology and detail control.

Findings

Successfully applied to geological data, demonstrating adaptability.

Ensures topological correctness and water-tightness in multi-class extraction.

Allows setting minimum detail constraints during approximation.

Abstract

Surface extraction from implicit neural representations modelling a single class surface is a well-known task. However, there exist no surface extraction methods from an implicit representation of multiple classes that guarantee topological correctness and no holes. In this work, we lay the groundwork by introducing a 2D boundary extraction algorithm for the multi-class case focusing on topological consistency and water-tightness, which also allows for setting minimum detail restraint on the approximation. Finally, we evaluate our algorithm using geological modelling data, showcasing its adaptiveness and ability to honour complex topology.