Spatially-Adaptive Hash Encodings For Neural Surface Reconstruction

TL;DR

This paper introduces a learned, spatially adaptive hash encoding method for neural surface reconstruction, enabling better frequency fitting and improved detail recovery compared to fixed encodings.

Contribution

It proposes a novel learned encoding approach that adaptively selects encoding basis per space, outperforming fixed grid-based hash encodings.

Findings

Achieves state-of-the-art results on benchmark datasets.

Allows neural networks to adapt encoding basis spatially.

Improves detail recovery without noise introduction.

Abstract

Positional encodings are a common component of neural scene reconstruction methods, and provide a way to bias the learning of neural fields towards coarser or finer representations. Current neural surface reconstruction methods use a "one-size-fits-all" approach to encoding, choosing a fixed set of encoding functions, and therefore bias, across all scenes. Current state-of-the-art surface reconstruction approaches leverage grid-based multi-resolution hash encoding in order to recover high-detail geometry. We propose a learned approach which allows the network to choose its encoding basis as a function of space, by masking the contribution of features stored at separate grid resolutions. The resulting spatially adaptive approach allows the network to fit a wider range of frequencies without introducing noise. We test our approach on standard benchmark surface reconstruction datasets and achieve state-of-the-art performance on two benchmark datasets.