Improving Neural Indoor Surface Reconstruction with Mask-Guided Adaptive Consistency Constraints

TL;DR

This paper introduces a novel two-stage training method with mask-guided adaptive consistency constraints to improve neural implicit surface reconstruction from images, especially in complex regions, without relying on extra priors.

Contribution

It proposes a new training framework that decouples view-dependent and view-independent colors and uses adaptive constraints guided by masks to enhance detail reconstruction.

Findings

Reduces the impact of prior estimation errors.

Achieves higher quality 3D reconstructions with rich details.

Effective on both synthetic and real-world datasets.

Abstract

3D scene reconstruction from 2D images has been a long-standing task. Instead of estimating per-frame depth maps and fusing them in 3D, recent research leverages the neural implicit surface as a unified representation for 3D reconstruction. Equipped with data-driven pre-trained geometric cues, these methods have demonstrated promising performance. However, inaccurate prior estimation, which is usually inevitable, can lead to suboptimal reconstruction quality, particularly in some geometrically complex regions. In this paper, we propose a two-stage training process, decouple view-dependent and view-independent colors, and leverage two novel consistency constraints to enhance detail reconstruction performance without requiring extra priors. Additionally, we introduce an essential mask scheme to adaptively influence the selection of supervision constraints, thereby improving performance in a self-supervised paradigm. Experiments on synthetic and real-world datasets show the capability of reducing the interference from prior estimation errors and achieving high-quality scene reconstruction with rich geometric details.