AtlasGS: Atlanta-world Guided Surface Reconstruction with Implicit Structured Gaussians

2025-10-28

AI Summary

Existing 3D reconstruction methods for indoor and urban scenes suffer from global geometric inconsistency in low-texture regions, loss of high-frequency details, and discontinuities or computational inefficiency inherent to implicit signed distance function (SDF) or Gaussian splatting approaches. To address these issues, we propose Atlanta-world prior-guided semantic Gaussian lattices—a novel representation that jointly leverages implicit SDF modeling and learnable structural plane regularization. By enforcing geometry-semantic co-constraints, our method enhances robustness in low-texture areas while preserving surface smoothness globally and fidelity of fine details. Crucially, it retains the differentiable, efficient rendering advantages of Gaussian splatting. Extensive experiments demonstrate that our approach significantly outperforms state-of-the-art methods across diverse indoor and urban scenes, achieving superior geometric consistency, detail reconstruction accuracy, and inference efficiency.

Technology Category

• Computer Vision: 3D Computer Vision
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Planning, Routing, and Scheduling: Replanning and Plan Repair

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

3D reconstruction of indoor and urban environments is a prominent research topic with various downstream applications. However, existing geometric priors for addressing low-texture regions in indoor and urban settings often lack global consistency. Moreover, Gaussian Splatting and implicit SDF fields often suffer from discontinuities or exhibit computational inefficiencies, resulting in a loss of detail. To address these issues, we propose an Atlanta-world guided implicit-structured Gaussian Splatting that achieves smooth indoor and urban scene reconstruction while preserving high-frequency details and rendering efficiency. By leveraging the Atlanta-world model, we ensure the accurate surface reconstruction for low-texture regions, while the proposed novel implicit-structured GS representations provide smoothness without sacrificing efficiency and high-frequency details. Specifically, we propose a semantic GS representation to predict the probability of all semantic regions and deploy a structure plane regularization with learnable plane indicators for global accurate surface reconstruction. Extensive experiments demonstrate that our method outperforms state-of-the-art approaches in both indoor and urban scenes, delivering superior surface reconstruction quality.

Problem

• Addressing low-texture region reconstruction with global consistency
• Resolving discontinuities in Gaussian Splatting and implicit SDF fields
• Improving computational efficiency while preserving high-frequency details

Innovation

• Atlanta-world model guides surface reconstruction
• Implicit-structured Gaussian Splatting ensures smooth reconstruction
• Semantic Gaussian representation with plane regularization