The prevalence of numerous Renaissance buildings displaying signs of damage underscores the urgent necessity for prompt detection and implementation of appropriate protective measures. However, interpreting a large number of damages in Renaissance walls manually is a time-intensive process, and the outcomes are highly reliant on the practitioner’s expertise. This paper presents an automatic segmentation algorithm designed to segment damage in images of Renaissance walls. Firstly, an imaging dataset containing seven damages to communal and Renaissance walls, including plants, micro-organisms, crack, spalling, efflorescence, erosion, and crush, is created. Secondly, a GridMask algorithm is employed for data augmentation to improve the robustness of the segmentation model. Finally, the established dataset, comprising 2,303 images (70%) for training, 658 images (20%) for validation, and 329 images (10%) for testing, is used to train the deep learning model and evaluate its precision in seg...

Automatic Detection of Damage in Renaissance Walls Using a Deep Learning Model

Dona Marco
;
Liu Xiaoyu;Secco Michele;da Porto Francesca
2025

Abstract

The prevalence of numerous Renaissance buildings displaying signs of damage underscores the urgent necessity for prompt detection and implementation of appropriate protective measures. However, interpreting a large number of damages in Renaissance walls manually is a time-intensive process, and the outcomes are highly reliant on the practitioner’s expertise. This paper presents an automatic segmentation algorithm designed to segment damage in images of Renaissance walls. Firstly, an imaging dataset containing seven damages to communal and Renaissance walls, including plants, micro-organisms, crack, spalling, efflorescence, erosion, and crush, is created. Secondly, a GridMask algorithm is employed for data augmentation to improve the robustness of the segmentation model. Finally, the established dataset, comprising 2,303 images (70%) for training, 658 images (20%) for validation, and 329 images (10%) for testing, is used to train the deep learning model and evaluate its precision in seg...
File in questo prodotto:
Non ci sono file associati a questo prodotto.
Pubblicazioni consigliate

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11577/3553637
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 0
  • ???jsp.display-item.citation.isi??? 0
  • OpenAlex ND
social impact