Object Detection Method for Automated Classification of Distress in Rabat's Built Heritage

Abstract

Rabat, the capital city of Morocco, proudly boasts a rich and complex architecture-al legacy that beautifully blends historical influences ranging from Islamic to con-temporary designs. Conserving this unique heritage holds paramount importance in safeguarding the city's distinctiveness and cultural significance. Conventional approaches to cataloging and categorization have been time-consuming and susceptible to human errors. Hence, this study aims to overcome these obstacles by creating a sophisticated object detection model to streamline the classification process. In this study, we propose an innovative deep learning-driven approach to detect and classify various degradations of built heritage. The dataset used in this study comprises numerous captured images that display diverse types of degradation, including cracks, collapse, rising damp, spalling, delamination, and lichens. Manual annotation was conducted to label the various damages present in the dataset. These labeled images were then used to train and validate the model. Multiple performance metrics were employed to assess and evaluate the model's performance, including precision and recall. Based on the results, the developed model has demonstrated excellent performance in both detecting and classifying different types of damage. This model's effective use aids local authorities in urban planning, heritage preservation, education, and tourism promotion, yielding broad implications.