PolyTopoBench: A Benchmark for Complex Vector Polygon Generation from Remote Sensing Imagery

2026-09-26

AI Summary

This study addresses the challenge of preserving complex topological structures, such as holes and multi-ring polygons, during remote sensing image vectorization, a problem compounded by the absence of systematic evaluation protocols. To this end, we construct the first unified evaluation framework explicitly designed for complex polygons comprising both inner and outer rings. Our investigation encompasses three methodological paradigms: end-to-end vector generation, segmentation-based post-processing, and vision foundation models. We conduct a systematic benchmark of eleven representative methods across two datasets. Experimental results demonstrate that while existing approaches effectively reconstruct simple outer boundaries, their performance degrades substantially when handling polygons with holes or multi-ring configurations. By exposing these critical deficiencies in complex topology recovery, this work provides valuable insights and establishes a clear direction for future research in remote sensing vectorization.

Abstract

Vector polygon generation converts visual inputs, e.g., remote sensing (RS) images, into vectorized polygonal geometries, supporting applications such as autonomous driving, vector map construction, and remote sensing. Early pipelines predict raster masks and post-process them into polygons, which prevents end-to-end optimization and may miss small objects or introduce inaccurate vertices. Recent methods directly generate vector polygons, but most focus on simple exterior contours, while they either cannot represent complex polygons with holes or fail to preserve their topology. In this paper, we propose PolyTopoBench, a unified evaluation framework for vector polygon generation from RS images with explicit emphasis on complex polygons. PolyTopoBench evaluates both exterior and interior rings, and benchmarks 11 representative methods, including segmentation-based polygonization pipelines, vision foundation model baselines, and specialized vector polygon generators, on two RS-image datasets covering buildings, roads, vegetation, and unvegetated regions. Experiments show that existing methods often recover simple exterior boundaries but degrade substantially on polygons with holes or multiple rings. These results reveal complex polygon generation as an unresolved challenge and motivate topology-aware benchmarks and model designs. Code and data are available at https://github.com/seai-lab/PolyTopoBench.

Problem

• vector polygon generation
• remote sensing imagery
• complex polygons
• topology preservation

Innovation

• Vector Polygon Generation
• Remote Sensing Imagery
• Complex Polygons
• Topology-aware Benchmark
• Evaluation Framework