Surface Damage Identification for Heritage Site Protection: A Mobile Crowd-sensing Solution Based on Deep Learning

A deep learning-based crowd-sensing solution demonstrates high accuracy (86.8% precision, 84% recall) in automatically identifying surface damage on heritage walls, addressing the inefficiencies of manual inspection methods. The study, conducted at Algiers' Kasbah, trained a MobileNetV2 CNN to detect efflorescence, spall, cracks, and mold using annotated images, with results showing effectiveness even with low-resolution inputs. This approach enables real-time, participatory damage detection via a mobile app, offering a scalable tool for heritage preservation.

Meklati, S., Boussora, K., Abdi, M. E. H., & Berrani, S.-A. (2023). Surface Damage Identification for Heritage Site Protection: A Mobile Crowd-sensing Solution Based on Deep Learning. Journal on Computing and Cultural Heritage, 16(2), 1–24. https://doi.org/10.1145/3569093