mlsanityData to 14 Oct 2026

PolyWorld: Polygonal Building Extraction with Graph Neural Networks in Satellite Images

arXiv 2111.15491 ↗First posted 30 Nov 2021cs.CVcited by 43

Not on the map. Why?

Abstract (arXiv)

While most state-of-the-art instance segmentation methods produce binary segmentation masks, geographic and cartographic applications typically require precise vector polygons of extracted objects instead of rasterized output. This paper introduces PolyWorld, a neural network that directly extracts building vertices from an image and connects them correctly to create precise polygons. The model predicts the connection strength between each pair of vertices using a graph neural network and estimates the assignments by solving a differentiable optimal transport problem. Moreover, the vertex positions are optimized by minimizing a combined segmentation and polygonal angle difference loss. PolyWorld significantly outperforms the state of the art in building polygonization and achieves not only notable quantitative results, but also produces visually pleasing building polygons. Code and trained weights are publicly available at https://github.com/zorzi-s/PolyWorldPretrainedNetwork.

Key facts. Cited by 43 papers in the map (data through 14 Oct 2026). Builds on 22 of its 43 references that are in the map. First posted on arXiv 30 Nov 2021.

Builds on

22 of 43 references.

References · 22

  1. 2017Attention Is All You Needcited by 60,361
  2. 2014Microsoft COCO: Common Objects in Contextcited by 17,608
  3. 2017Mask R-CNNcited by 7,277
  4. 2013Sinkhorn Distances: Lightspeed Computation of Optimal Transportation Distancescited by 2,316
  5. 2019SuperGlue: Learning Feature Matching with Graph Neural Networkscited by 1,443
  6. 2018Path Aggregation Network for Instance Segmentationcited by 798
  7. 2017Generalised Dice overlap as a deep learning loss function for highly unbalanced segmentationscited by 620
  8. 2018On Regularized Losses for Weakly-supervised CNN Segmentationcited by 121
  9. 2018Efficient Interactive Annotation of Segmentation Datasets with Polygon-RNN++cited by 120
  10. 2018Normalized Cut Loss for Weakly-supervised CNN Segmentationcited by 112
  11. 2019Fast Interactive Object Annotation with Curve-GCNcited by 95
  12. 2019PolyTransform: Deep Polygon Transformer for Instance Segmentationcited by 92
10 more
  1. 2017Annotating Object Instances with a Polygon-RNNcited by 88
  2. 2018TernausNet: U-Net with VGG11 Encoder Pre-Trained on ImageNet for Image Segmentationcited by 76
  3. 2018Topological Map Extraction from Overhead Imagescited by 58
  4. 2018Learning deep structured active contours end-to-endcited by 38
  5. 2019DARNet: Deep Active Ray Network for Building Segmentationcited by 30
  6. 2021MonteFloor: Extending MCTS for Reconstructing Accurate Large-Scale Floor Planscited by 26
  7. 2018TernausNetV2: Fully Convolutional Network for Instance Segmentationcited by 25
  8. 2020Machine-learned Regularization and Polygonization of Building Segmentation Maskscited by 22
  9. 2020Regularization of Building Boundaries in Satellite Images using Adversarial and Regularized Lossescited by 4
  10. 2020Quantization in Relative Gradient Angle Domain For Building Polygon Estimationcited by 3

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2022: 5; 2023: 11; 2024: 10; 2025: 8; 2026: 9.

43 papers.

Sort:

By area:

  1. 2022VectorMapNet: End-to-end Vectorized HD Map Learningcited by 245
  2. 2023SkySense: A Multi-Modal Remote Sensing Foundation Model Towards Universal Interpretation for Earth Observation Imagerycited by 205
  3. 2022SatlasPretrain: A Large-Scale Dataset for Remote Sensing Image Understandingcited by 167
  4. 2023InstaGraM: Instance-level Graph Modeling for Vectorized HD Map Learningcited by 40
  5. 2023ScalableMap: Scalable Map Learning for Online Long-Range Vectorized HD Map Constructioncited by 16
  6. 2023GlobalMapper: Arbitrary-Shaped Urban Layout Generationcited by 15
  7. 20243D Building Reconstruction from Monocular Remote Sensing Images with Multi-level Supervisionscited by 13
  8. 2024PolygonGNN: Representation Learning for Polygonal Geometries with Heterogeneous Visibility Graphcited by 10
  9. 2024Pix2Poly: A Sequence Prediction Method for End-to-end Polygonal Building Footprint Extraction from Remote Sensing Imagerycited by 8
  10. 2022Accurate Polygonal Mapping of Buildings in Satellite Imagerycited by 7
  11. 2023HiT: Building Mapping with Hierarchical Transformerscited by 6
  12. 2024PolyR-CNN: R-CNN for end-to-end polygonal building outline extractioncited by 6
31 more
  1. 2024RoIPoly: Vectorized Building Outline Extraction Using Vertex and Logit Embeddingscited by 6
  2. 2023Patched Line Segment Learning for Vector Road Mappingcited by 5
  3. 2024P2PFormer: A Primitive-to-polygon Method for Regular Building Contour Extraction from Remote Sensing Imagescited by 5
  4. 2023Data Leakage Detection and De-duplication in Large Scale Geospatial Image Datasetscited by 4
  5. 2025Global Collinearity-aware Polygonizer for Polygonal Building Mapping in Remote Sensingcited by 4
  6. 2025HoliTracer: Holistic Vectorization of Geographic Objects from Large-Size Remote Sensing Imagerycited by 4
  7. 2022PolyBuilding: Polygon Transformer for End-to-End Building Extractioncited by 3
  8. 2024Beyond Grid Data: Exploring Graph Neural Networks for Earth Observationcited by 3
  9. 2025LDPoly: Latent Diffusion for Polygonal Road Outline Extraction in Large-Scale Topographic Mappingcited by 3
  10. 2026ACPV-Net: All-Class Polygonal Vectorization for Seamless Vector Map Generation from Aerial Imagerycited by 3
  11. 2023BiSVP: Building Footprint Extraction via Bidirectional Serialized Vertex Predictioncited by 1
  12. 20233D detection of roof sections from a single satellite image and application to LOD2-building reconstructioncited by 1
  13. 2024GeoFormer: A Multi-Polygon Segmentation Transformercited by 1
  14. 2025SAR Object Detection with Self-Supervised Pretraining and Curriculum-Aware Samplingcited by 1
  15. 2025The P3 Dataset: Pixels, Points and Polygons for Multimodal Building Vectorizationcited by 1
  16. 2025Bridging the Gap Between Sparsity and Redundancy: A Dual-Decoding Framework with Global Context for Map Inferencecited by 1
  17. 2025UniVector: Unified Vector Extraction via Instance-Geometry Interactioncited by 1
  18. 2022Primitive Graph Learning for Unified Vector Mappingcited by 0
  19. 2023SuperpixelGraph: Semi-automatic generation of building footprint through semantic-sensitive superpixel and neural graph networkscited by 0
  20. 2023Context-Enhanced Detector For Building Detection From Remote Sensing Imagescited by 0
  21. 2024SISP: A Benchmark Dataset for Fine-grained Ship Instance Segmentation in Panchromatic Satellite Imagescited by 0
  22. 2024Enhancing Polygonal Building Segmentation via Oriented Cornerscited by 0
  23. 2025Polygonizing Roof Segments from High-Resolution Aerial Images Using Yolov8-Based Edge Detectioncited by 0
  24. 2026PolyBuild: An End-to-End Method for Polygonal Building Contour Extraction from High-Resolution Remote Sensing Imagescited by 0
  25. 2026Vector Map as Language: Toward Unified Remote Sensing Vector Mappingcited by 0
  26. 2026GeoAI-based post-segmentation quality validation of building footprints via spatial feature engineeringcited by 0
  27. 2026SiZeUp: Fast 3D Proxy from Aerial Images via Depth Ordinal Losscited by 0
  28. 2026Diffusion Transformers for Roof Graph Synthesis and Reconstructioncited by 0
  29. 2026Topologically Consistent Agricultural Parcel Vectorization with Semantic-Guided Diffusion and Topology-Aware Polygonizationcited by 0
  30. 2026PolyTopoBench: A Benchmark for Complex Vector Polygon Generation from Remote Sensing Imagerycited by 0
  31. 2026LoDEOT: Low-Dimensional and Efficient Offset Tokens for Building Footprint Extraction from Off-Nadir Imagerycited by 0