Plan2Map: A Multimodal Benchmark for Document-Grounded Geospatial Boundary Reconstruction from Planning Records

2026-06-01

AI Summary

Reconstructing precise geographic boundaries from planning documents containing only unstructured text and images is highly challenging due to the absence of machine-readable annotations. This work addresses this problem by introducing Plan2Map, the first multimodal benchmark comprising 208 real-world cases, and proposes GeoPlanAgent—a tool-augmented, multi-stage reasoning framework that integrates vision-language models, geospatial tool invocation, supervised boundary segmentation, and geometric validation. Evaluated on Plan2Map, the proposed approach substantially outperforms end-to-end vision-language baselines, achieving an average Intersection over Union (IoU) of 0.736 and a median IoU of 0.904, with 67.8% of predicted boundaries attaining an IoU of at least 0.8.

Abstract

Planning records define restrictions over geographic areas, but their source documents often provide only indirect spatial evidence rather than machine-readable boundaries. We introduce Plan2Map, a 208-case multimodal benchmark for document-grounded geospatial boundary reconstruction from UK planning records. Given only a source planning document, systems must reconstruct a valid geospatial boundary from notice text, schedules, map plates, map labels, and boundary annotations; the reference GeoJSON is held out for scoring. We propose GeoPlanAgent, a document-grounded, geospatial-tool-in-the-loop system that decomposes the task into evidence extraction, localisation, map registration, boundary segmentation, projection, and verification. On Plan2Map, GeoPlanAgent achieves 0.736 mean IoU and 0.904 median IoU, with 67.8\% of predictions at or above 0.8 IoU, substantially outperforming direct VLM-to-GeoJSON baselines. Diagnostic analysis shows that direct VLM prediction remains unreliable, while remaining errors are concentrated in localisation and map registration, and supervised boundary segmentation substantially improves pixel-level mask quality. Plan2Map provides a concrete testbed for multimodal geospatial reconstruction from public planning records. Project page: https://odeb1.github.io/Plan2Map_Project_Page/.

Problem

• geospatial boundary reconstruction
• planning records
• document-grounded
• multimodal benchmark
• GeoJSON

Innovation

• multimodal geospatial reconstruction
• document-grounded reasoning
• geospatial boundary segmentation
• tool-in-the-loop agent
• planning records interpretation