SAND: Spatially Adaptive Network Depth for Fast Sampling of Neural Implicit Surfaces

TL;DR

SAND introduces a spatially adaptive neural network depth framework that reduces computational costs in implicit surface modeling by focusing resources on complex regions.

Contribution

The paper proposes a novel adaptive network depth method using a volumetric depth map and a tail-MLP to improve efficiency in neural implicit representations.

Findings

Significantly faster inference times for implicit neural representations.

Efficiently allocates computational resources to complex geometric regions.

Maintains high-fidelity surface reconstructions despite reduced computation.

Abstract

Implicit neural representations are powerful for geometric modeling, but their practical use is often limited by the high computational cost of network evaluations. We observe that implicit representations require progressively lower accuracy as query points move farther from the target surface, and that even within the same iso-surface, representation difficulty varies spatially with local geometric complexity. However, conventional neural implicit models evaluate all query points with the same network depth and computational cost, ignoring this spatial variation and thereby incurring substantial computational waste. Motivated by this observation, we propose an efficient neural implicit geometry representation framework with spatially adaptive network depth (SAND). SAND leverages a volumetric network-depth map together with a tailed multi-layer perceptron (T-MLP) to model implicit representation. The volumetric depth map records, for each spatial region, the network depth required to achieve sufficient accuracy, while the T-MLP is a modified MLP designed to learn implicit functions such as signed distance functions, where an output branch, referred to as a tail, is attached to each hidden layer. This design allows network evaluation to terminate adaptively without traversing the full network and directs computational resources to geometrically important and complex regions, improving efficiency while preserving high-fidelity representations. Extensive experimental results demonstrate that our approach can significantly improve the inference-time query speed of implicit neural representations.