EpiSAM: Character Segmentation in Challenging Stone Inscriptions

2026-06-27

AI Summary

This work addresses the challenging problem of character segmentation in stone inscriptions, which is hindered by irregular surfaces, weathering, and low contrast. To tackle this, the authors propose EpiSAM, a framework that treats character detection as the core strategy, leveraging a prompt-guided Transformer architecture augmented with a novel neighborhood-aware mechanism. This mechanism explicitly models contextual neighboring characters during target character prediction, effectively mitigating boundary ambiguity. The study contributes the first fine-grained, densely annotated dataset dedicated to Southeast Asian stone inscriptions, facilitating epigraphic research. Experimental results demonstrate consistent performance gains over multiple baselines and highlight the model’s strong zero-shot generalization capability in challenging inscription scenarios.

Abstract

Stone inscriptions are invaluable sources of historical and linguistic knowledge, yet their automated analysis remains a major challenge due to surface irregularities, erosion, and low visual contrast. Conventional document and handwriting analysis techniques fail to perform well in these scenarios. In this work, we propose character detection as a core strategy for robust inscription analysis. We introduce EpiSAM, a prompt-guided transformer framework for character segmentation in stone inscriptions. Rather than treating characters in isolation, EpiSAM employs a novel neighbor-aware strategy, explicitly predicting adjacent characters alongside the target. These contextual cues resolve boundary ambiguities, improving mask generation and enabling more accurate character segmentation. Furthermore, we expand an existing stone inscription dataset by adding dense polygonal annotations for characters, thereby enabling comprehensive research on Southeast Asian epigraphy. Experimental results show that EpiSAM achieves consistent improvements over existing baselines, while also exhibiting strong zero-shot generalization in challenging epigraphic scenarios.

Problem

• stone inscriptions
• character segmentation
• surface irregularities
• low visual contrast
• epigraphy

Innovation

• character segmentation
• prompt-guided transformer
• neighbor-aware strategy
• stone inscriptions
• epigraphy