MapSAM2: Adapting SAM2 for Automatic Segmentation of Historical Map Images and Time Series

2025-10-31

AI Summary

This paper addresses the challenge of automatic segmentation of historical map images and associated time-series data. We propose a unified spatiotemporal modeling framework that treats single or multi-temporal maps as pseudo-video inputs and performs few-shot fine-tuning of the SAM2 vision foundation model. Our contributions include: (1) a memory-augmented spatiotemporal attention mechanism enabling cross-frame feature alignment and instance association; (2) the Siegfried Building Time-Series dataset, along with a pseudo-time-series generation strategy using single-year maps to reduce annotation cost; and (3) a geometrically aware sliding-window tiling scheme for robust inference. Experiments demonstrate substantial improvements in both polygonal object segmentation accuracy (e.g., buildings) and inter-temporal instance linking under limited supervision, outperforming state-of-the-art methods. The code and dataset will be publicly released.

Technology Category

• Computer Vision: Segmentation
• Knowledge Representation and Reasoning: Geometric, Spatial, and Temporal Reasoning
• Planning, Routing, and Scheduling: Optimization of Spatio-temporal Systems

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Graph Algorithms and Modeling for the Web: Algorithms and analysis for heterogeneous, signed, attributed, multi-relational, temporal, higher-order, and annotated Web-related graphs

Abstract

Historical maps are unique and valuable archives that document geographic features across different time periods. However, automated analysis of historical map images remains a significant challenge due to their wide stylistic variability and the scarcity of annotated training data. Constructing linked spatio-temporal datasets from historical map time series is even more time-consuming and labor-intensive, as it requires synthesizing information from multiple maps. Such datasets are essential for applications such as dating buildings, analyzing the development of road networks and settlements, studying environmental changes etc. We present MapSAM2, a unified framework for automatically segmenting both historical map images and time series. Built on a visual foundation model, MapSAM2 adapts to diverse segmentation tasks with few-shot fine-tuning. Our key innovation is to treat both historical map images and time series as videos. For images, we process a set of tiles as a video, enabling the memory attention mechanism to incorporate contextual cues from similar tiles, leading to improved geometric accuracy, particularly for areal features. For time series, we introduce the annotated Siegfried Building Time Series Dataset and, to reduce annotation costs, propose generating pseudo time series from single-year maps by simulating common temporal transformations. Experimental results show that MapSAM2 learns temporal associations effectively and can accurately segment and link buildings in time series under limited supervision or using pseudo videos. We will release both our dataset and code to support future research.

Problem

• Automating historical map segmentation despite stylistic variability and data scarcity
• Constructing linked spatio-temporal datasets from historical map time series efficiently
• Reducing annotation costs for temporal analysis of geographic features in maps

Innovation

• Adapts visual foundation model for map segmentation
• Treats map images and time series as videos
• Generates pseudo time series from single-year maps