Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Enhancing Polygonal Building Segmentation via Oriented Corners

Mohammad Moein Sheikholeslami, Muhammad Kamran, Andreas Wichmann, Gunho Sohn

TL;DR

This work tackles the challenge of generating accurate polygonal building representations directly from overhead imagery. It introduces OriCornerNet, a direct polygonal segmentation network that predicts footprints, corner locations, and orientation vectors to guide end-to-end polygon reconstruction, supplemented by a graph convolutional refinement stage. By coupling an initialization module with geometry-aware features and orientation-based regularizers, the method produces more regular and accurate polygons while reducing reliance on post-processing. Evaluations on SpaceNet Vegas and CrowdAI-small demonstrate significant improvements over state-of-the-art approaches in both raster and vector metrics, highlighting practical impact for vectorized mapping and GIS workflows.

Abstract

The growing demand for high-resolution maps across various applications has underscored the necessity of accurately segmenting building vectors from overhead imagery. However, current deep neural networks often produce raster data outputs, leading to the need for extensive post-processing that compromises the fidelity, regularity, and simplicity of building representations. In response, this paper introduces a novel deep convolutional neural network named OriCornerNet, which directly extracts delineated building polygons from input images. Specifically, our approach involves a deep model that predicts building footprint masks, corners, and orientation vectors that indicate directions toward adjacent corners. These predictions are then used to reconstruct an initial polygon, followed by iterative refinement using a graph convolutional network that leverages semantic and geometric features. Our method inherently generates simplified polygons by initializing the refinement process with predicted corners. Also, including geometric information from oriented corners contributes to producing more regular and accurate results. Performance evaluations conducted on SpaceNet Vegas and CrowdAI-small datasets demonstrate the competitive efficacy of our approach compared to the state-of-the-art in building segmentation from overhead imagery.

Enhancing Polygonal Building Segmentation via Oriented Corners

TL;DR

This work tackles the challenge of generating accurate polygonal building representations directly from overhead imagery. It introduces OriCornerNet, a direct polygonal segmentation network that predicts footprints, corner locations, and orientation vectors to guide end-to-end polygon reconstruction, supplemented by a graph convolutional refinement stage. By coupling an initialization module with geometry-aware features and orientation-based regularizers, the method produces more regular and accurate polygons while reducing reliance on post-processing. Evaluations on SpaceNet Vegas and CrowdAI-small demonstrate significant improvements over state-of-the-art approaches in both raster and vector metrics, highlighting practical impact for vectorized mapping and GIS workflows.

Abstract

The growing demand for high-resolution maps across various applications has underscored the necessity of accurately segmenting building vectors from overhead imagery. However, current deep neural networks often produce raster data outputs, leading to the need for extensive post-processing that compromises the fidelity, regularity, and simplicity of building representations. In response, this paper introduces a novel deep convolutional neural network named OriCornerNet, which directly extracts delineated building polygons from input images. Specifically, our approach involves a deep model that predicts building footprint masks, corners, and orientation vectors that indicate directions toward adjacent corners. These predictions are then used to reconstruct an initial polygon, followed by iterative refinement using a graph convolutional network that leverages semantic and geometric features. Our method inherently generates simplified polygons by initializing the refinement process with predicted corners. Also, including geometric information from oriented corners contributes to producing more regular and accurate results. Performance evaluations conducted on SpaceNet Vegas and CrowdAI-small datasets demonstrate the competitive efficacy of our approach compared to the state-of-the-art in building segmentation from overhead imagery.
Paper Structure (11 sections, 8 equations, 4 figures, 2 tables)

This paper contains 11 sections, 8 equations, 4 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Instance Segmentation Baseline
  5. Initialization Module
  6. GCNs
  7. Experiments
  8. Datasets and Settings
  9. Evaluation Metrics
  10. Preliminary Results and Discussion
  11. Conclusion

Figures (4)

  • Figure 1: Comparison between the common approach for extracting buildings from imagery and OriCornerNet. The common approach suffers from redundant vertices and irregular polygons.
  • Figure 2: The architecture of OriCornerNet.
  • Figure 3: Oriented corners on Vegas dataset. OriCornerNet can grasp building regular shapes and detect obscured corners.
  • Figure 4: Qualitative results of Hisup, the proposed method, and ground truth on CrowdAI-small dataset.