Papers › SpaGBOL: Spatial-Graph-Based Orientated Localisation

SpaGBOL: Spatial-Graph-Based Orientated Localisation

23 Sep 2024arXiv:2409.15514archive 2025-07-28

Tavis Shore, Oscar Mendez, Simon Hadfield

Cross-View Geo-Localisation within urban regions is challenging in part due to the lack of geo-spatial structuring within current datasets and techniques. We propose utilising graph representations to model sequences of local observations and the connectivity of the target location. Modelling as a graph enables generating previously unseen sequences by sampling with new parameter configurations. To leverage this newly available information, we propose a GNN-based architecture, producing spatially strong embeddings and improving discriminability over isolated image embeddings. We outline SpaGBOL, introducing three novel contributions. 1) The first graph-structured dataset for Cross-View Geo-Localisation, containing multiple streetview images per node to improve generalisation. 2) Introducing GNNs to the problem, we develop the first system that exploits the correlation between node proximity and feature similarity. 3) Leveraging the unique properties of the graph representation - we demonstrate a novel retrieval filtering approach based on neighbourhood bearings. SpaGBOL achieves state-of-the-art accuracies on the unseen test graph - with relative Top-1 retrieval improvements on previous techniques of 11%, and 50% when filtering with Bearing Vector Matching on the SpaGBOL dataset.

PaperPDFCode

Code

tavisshore/SpaGBOL officialmentioned in papermentioned on GitHubpytorch report

Repository list and official/mentioned flags are the archive's, frozen 2025-07-28. Reachability, where shown, is from one Syntology probe window (2026-09-16 to 2026-09-18); repositories not probed show nothing. GitHub stars are not tracked.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Camera LocalizationCross-View Geo-LocalisationImage RetrievalImage-Based LocalizationNavigateOutdoor LocalizationRetrievalVisual Localizationgeo-localization

Datasets

Introduced by this paper, per the archive.

SpaGBOL

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Cross-View Geo-Localisation SpaGBOL SpaGBOL Top-1 56.48 #1 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL SpaGBOL Top-1% 87.24 #1 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL SpaGBOL Top-10 83.85 #1 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL SpaGBOL Top-5 77.47 #1 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL Sample4Geo Top-1 50.80 #2 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL Sample4Geo Top-1% 82.32 #2 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL Sample4Geo Top-10 79.96 #2 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL Sample4Geo Top-5 74.22 #2 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL SAIG-D Top-1 25.65 #4 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL SAIG-D Top-1% 68.22 #4 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL SAIG-D Top-10 62.29 #4 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL SAIG-D Top-5 51.44 #4 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL GeoDTR+ Top-1 17.49 #5 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL GeoDTR+ Top-1% 59.41 #5 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL GeoDTR+ Top-10 52.01 #5 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL GeoDTR+ Top-5 40.27 #5 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL L2LTR Top-1 11.23 #6 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL L2LTR Top-1% 49.52 #6 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL L2LTR Top-10 42.5 #6 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL L2LTR Top-5 31.27 #6 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL DSM Top-1 5.82 #7 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL DSM Top-1% 18.62 #7 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL DSM Top-10 14.13 #7 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL DSM Top-5 10.21 #7 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL CVFT Top-1 4.02 #8 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL CVFT Top-1% 27.19 #8 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL CVFT Top-10 20.29 #8 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL CVM-Net Top-1 2.87 #9 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL CVM-Net Top-1% 28.33 #9 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL CVM-Net Top-10 21.51 #9 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL CVM-Net Top-5 13.02 #9 of 9 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL 180° SpaGBOL Top-1 40.88 #1 of 1 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL 180° SpaGBOL Top-5 63.79 #1 of 1 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL 90° SpaGBOL Top-1 18.63 #1 of 1 Archive leaderboard report
Cross-View Geo-Localisation SpaGBOL 90° SpaGBOL Top-5 43.2 #1 of 1 Archive leaderboard report
Cross-View Geo-Localisation VIGOR-Graph SpaGBOL Accuracy (Top-1) 31.88 #1 of 1 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

ConvNeXtConvolutionGraph Neural NetworkGraphSAGE

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections