Answers are grounded in this paper and cite their sources.
Systematic review identifies AI methods, datasets, and challenges in crack detection for cultural heritage sites.
Cracks are a critical defect to be identified during inspections of cultural heritage sites. In this context, several recent studies have explored Artificial Intelligence (AI) techniques for crack detection in historic structures. However, the literature still lacks a structured bibliographic review on the state of the art in this research field. Therefore, the objective of this paper is to present a systematic literature review on deep learning for crack detection in cultural heritage. A total of 26 papers, published between 2020 and 2025 and indexed in the Scopus database, were analyzed. The proposed methodology addressed five research questions, aiming to identify the most cited papers and the journals with the highest number of publications, the availability of public datasets, the most commonly used deep learning methods, the types of monuments and countries studied, and future research perspectives. Key findings include the identification of six publicly available datasets containing images for training AI models in crack detection for cultural heritage. Additionally, YOLO emerged as the most frequently adopted deep learning method in this domain. The review also showed that case studies were conducted on monuments or objects from 12 different countries, with several studies focusing on UNESCO World Heritage Sites. Finally, seven main future challenges were identified, with the most prominent being: improving the generalization capabilities of deep learning methods, enhancing the quality and availability of datasets, and advancing the use of autonomous aerial vehicles in heritage inspection tasks.
No takes yet. Share an insight, caveat, or question.
Ottoni et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: