Automated Graffiti Detection: A Novel Approach to Maintaining Historical Architecture in Communities

Graffiti is common in many communities and even affects our historical heritage structures. This leads to a decrease the revenue associated with commercial activities or services (e.g., shops, restaurants, residences), potentially reduces tourism region. Visual data, form of photographs, becoming an efficient mechanism record information. Photographs can be quickly captured, are already frequently posted online by ordinary citizens tourists, residents, visitors). Exploiting image data through automation computer vision provides new opportunity simplify current manual graffiti-monitoring processes, enabling automated detection, localization, quantification such markings. In this study, we developed vision-based graffiti-detection technique using convolutional neural network. Images collected from structures interest within community utilized automatically inspect for graffiti case which collect contribute there high degree duplication repetition, lack GPS These hinder direct use images automating process. To address these challenges, built high-resolution, single-view façade (orthophotos) before applying robust detector. The detector was database 1022 damaged contaminated gathered during recent European Union project, entitled “Safeguarding Cultural Heritage Technical Organisational Resources Management” (STORM). A total 818 were used training (10% set randomly chosen validation set), achieving 88% accuracy among remaining 204 samples testing. Using trained detector, demonstrated Church Agios Nikolaos (Leontariou), Kantza, Greece.