Learning Compact Latent Space for Representing Neural Signed Distance Functions with High-fidelity Geometry Details

TL;DR

This paper introduces a method to represent multiple neural signed distance functions in a shared, compact latent space, enabling high-fidelity 3D shape reconstruction with improved efficiency and detail preservation.

Contribution

The paper proposes a novel framework combining generalization and overfitting strategies to enhance the representation of multiple SDFs in a compact latent space, along with a new sampling method.

Findings

Achieves higher fidelity geometry details in 3D shape reconstructions.

Improves training efficiency and reduces artifacts in SDF learning.

Outperforms recent methods in representation compactness and accuracy.

Abstract

Neural signed distance functions (SDFs) have been a vital representation to represent 3D shapes or scenes with neural networks. An SDF is an implicit function that can query signed distances at specific coordinates for recovering a 3D surface. Although implicit functions work well on a single shape or scene, they pose obstacles when analyzing multiple SDFs with high-fidelity geometry details, due to the limited information encoded in the latent space for SDFs and the loss of geometry details. To overcome these obstacles, we introduce a method to represent multiple SDFs in a common space, aiming to recover more high-fidelity geometry details with more compact latent representations. Our key idea is to take full advantage of the benefits of generalization-based and overfitting-based learning strategies, which manage to preserve high-fidelity geometry details with compact latent codes. Based on this framework, we also introduce a novel sampling strategy to sample training queries. The sampling can improve the training efficiency and eliminate artifacts caused by the influence of other SDFs. We report numerical and visual evaluations on widely used benchmarks to validate our designs and show advantages over the latest methods in terms of the representative ability and compactness.