CAGE: Continuity-Aware edGE Network Unlocks Robust Floorplan Reconstruction

2025-09-18

AI Summary

Traditional polygonal layout reconstruction methods are highly sensitive to point cloud noise and occlusion, often yielding fragmented layouts and topological inconsistencies; while existing line-grouping approaches improve robustness, they suffer from insufficient geometric detail recovery. This paper proposes an edge-centric vectorized floorplan reconstruction framework: it explicitly models wall segments as directed continuous edges and directly generates structurally coherent, topologically valid indoor layouts from point cloud density maps. We introduce native edge modeling and a dual-query Transformer decoder, enabling collaborative optimization between perturbed and latent queries to ensure geometric continuity and watertight room boundaries. Integrated denoising training and structure-aware edge grouping enable end-to-end vectorized generation. Our method achieves state-of-the-art performance on Structured3D and SceneCAD: 99.1% room F1-score, 89.7% corner detection accuracy, and 89.3% angle estimation accuracy, with strong cross-dataset generalization capability.

Technology Category

• Planning, Routing, and Scheduling: Planning with Language Models
• Computer Vision: 3D Computer Vision
• Intelligent Robots: State Estimation

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Systems and Infrastructure for Web, Mobile and WoT: Cloud, edge and content delivery systems for the Web

Abstract

We present extbf{CAGE} ( extit{Continuity-Aware edGE}) network, a extcolor{red}{robust} framework for reconstructing vector floorplans directly from point-cloud density maps. Traditional corner-based polygon representations are highly sensitive to noise and incomplete observations, often resulting in fragmented or implausible layouts. Recent line grouping methods leverage structural cues to improve robustness but still struggle to recover fine geometric details. To address these limitations, we propose a extit{native} edge-centric formulation, modeling each wall segment as a directed, geometrically continuous edge. This representation enables inference of coherent floorplan structures, ensuring watertight, topologically valid room boundaries while improving robustness and reducing artifacts. Towards this design, we develop a dual-query transformer decoder that integrates perturbed and latent queries within a denoising framework, which not only stabilizes optimization but also accelerates convergence. Extensive experiments on Structured3D and SceneCAD show that extbf{CAGE} achieves state-of-the-art performance, with F1 scores of 99.1% (rooms), 91.7% (corners), and 89.3% (angles). The method also demonstrates strong cross-dataset generalization, underscoring the efficacy of our architectural innovations. Code and pretrained models will be released upon acceptance.

Problem

• Robust vector floorplan reconstruction from point-cloud maps
• Addressing noise sensitivity in corner-based polygon representations
• Recovering fine geometric details and ensuring topological validity

Innovation

• Edge-centric formulation for continuity
• Dual-query transformer decoder integration
• Denoising framework stabilizes optimization