Synthetic Training Datasets for Architectural Conservation: A Deep Learning Approach for Decay Detection

Architectural heritage conservation increasingly relies on innovative tools for detecting and monitoring degradation. The study presented in the current paper explores the use of synthetic datasets—namely, rendered images derived from photogrammetric models—to train convolutional neural networks (CN...

Full description

Saved in:
Bibliographic Details
Main Authors: Giacomo Patrucco, Francesco Setragno, Antonia Spanò
Format: Article
Language:English
Published: MDPI AG 2025-05-01
Series:Remote Sensing
Subjects:
Online Access:https://www.mdpi.com/2072-4292/17/10/1714
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850255960790007808
author Giacomo Patrucco
Francesco Setragno
Antonia Spanò
author_facet Giacomo Patrucco
Francesco Setragno
Antonia Spanò
author_sort Giacomo Patrucco
collection DOAJ
description Architectural heritage conservation increasingly relies on innovative tools for detecting and monitoring degradation. The study presented in the current paper explores the use of synthetic datasets—namely, rendered images derived from photogrammetric models—to train convolutional neural networks (CNNs) for the automated detection of deterioration in historical reinforced concrete structures. The primary objective is to assess the effectiveness of synthetic images for deep learning training, comparing their performance with models trained on traditional datasets. The research focuses on a significant case study: the parabolic concrete arch of Morano sul Po. Two classification scenarios were tested: a single-class model for structure recognition and a multi-class model for identifying degradation patterns, such as exposed reinforcement bars. The findings indicate that synthetic datasets can effectively support structure identification, achieving results comparable to those obtained with real-world imagery. However, challenges arise in multi-class classification, particularly in distinguishing fine-grained degradation features. This study highlights the potential of artificial datasets in overcoming the limitations of annotated data availability in heritage conservation. The proposed approach represents a promising step toward automating documentation and damage assessment, ultimately contributing to more efficient and scalable heritage monitoring strategies.
format Article
id doaj-art-7f7b7ad586e04ed2a57782fd42992384
institution OA Journals
issn 2072-4292
language English
publishDate 2025-05-01
publisher MDPI AG
record_format Article
series Remote Sensing
spelling doaj-art-7f7b7ad586e04ed2a57782fd429923842025-08-20T01:56:45ZengMDPI AGRemote Sensing2072-42922025-05-011710171410.3390/rs17101714Synthetic Training Datasets for Architectural Conservation: A Deep Learning Approach for Decay DetectionGiacomo Patrucco0Francesco Setragno1Antonia Spanò2Lab G4CH, Department of Architecture and Design (DAD), Politecnico di Torino, Viale Mattioli, 39, 10125 Torino, ItalyVolta Robots srl, Via Roberto Lepetit, 34, 21040 Gerenzano, ItalyLab G4CH, Department of Architecture and Design (DAD), Politecnico di Torino, Viale Mattioli, 39, 10125 Torino, ItalyArchitectural heritage conservation increasingly relies on innovative tools for detecting and monitoring degradation. The study presented in the current paper explores the use of synthetic datasets—namely, rendered images derived from photogrammetric models—to train convolutional neural networks (CNNs) for the automated detection of deterioration in historical reinforced concrete structures. The primary objective is to assess the effectiveness of synthetic images for deep learning training, comparing their performance with models trained on traditional datasets. The research focuses on a significant case study: the parabolic concrete arch of Morano sul Po. Two classification scenarios were tested: a single-class model for structure recognition and a multi-class model for identifying degradation patterns, such as exposed reinforcement bars. The findings indicate that synthetic datasets can effectively support structure identification, achieving results comparable to those obtained with real-world imagery. However, challenges arise in multi-class classification, particularly in distinguishing fine-grained degradation features. This study highlights the potential of artificial datasets in overcoming the limitations of annotated data availability in heritage conservation. The proposed approach represents a promising step toward automating documentation and damage assessment, ultimately contributing to more efficient and scalable heritage monitoring strategies.https://www.mdpi.com/2072-4292/17/10/1714deep learningartificial training datasets generationautomatic classificationdecay detectionconcrete heritage
spellingShingle Giacomo Patrucco
Francesco Setragno
Antonia Spanò
Synthetic Training Datasets for Architectural Conservation: A Deep Learning Approach for Decay Detection
Remote Sensing
deep learning
artificial training datasets generation
automatic classification
decay detection
concrete heritage
title Synthetic Training Datasets for Architectural Conservation: A Deep Learning Approach for Decay Detection
title_full Synthetic Training Datasets for Architectural Conservation: A Deep Learning Approach for Decay Detection
title_fullStr Synthetic Training Datasets for Architectural Conservation: A Deep Learning Approach for Decay Detection
title_full_unstemmed Synthetic Training Datasets for Architectural Conservation: A Deep Learning Approach for Decay Detection
title_short Synthetic Training Datasets for Architectural Conservation: A Deep Learning Approach for Decay Detection
title_sort synthetic training datasets for architectural conservation a deep learning approach for decay detection
topic deep learning
artificial training datasets generation
automatic classification
decay detection
concrete heritage
url https://www.mdpi.com/2072-4292/17/10/1714
work_keys_str_mv AT giacomopatrucco synthetictrainingdatasetsforarchitecturalconservationadeeplearningapproachfordecaydetection
AT francescosetragno synthetictrainingdatasetsforarchitecturalconservationadeeplearningapproachfordecaydetection
AT antoniaspano synthetictrainingdatasetsforarchitecturalconservationadeeplearningapproachfordecaydetection