End-to-End Asbestos Roof Detection on Orthophotos Using Transformer-Based YOLO Deep Neural Network

Cesare Davide Pace, Alessandro Bria, Mariano Focareta, Gabriele Lozupone, Claudio Marrocco, Giuseppe Meoli, Mario Molinara

Abstract: Asbestos, a hazardous material associated with severe health issues, requires accurate identification for safe management and removal. This study presents a novel end-to-end deep learning approach using a transformer-based YOLOv5 network for detecting asbestos roofs in high-resolution orthophotos, filling a gap in the scientific literature where end-to-end solutions are lacking. The model is trained on a dataset containing orthophotos with various roof types and conditions around Pisa in Italy. The transformer-based YOLO architecture enhances the detection capabilities compared to traditional CNNs. The proposed method demonstrates high accuracy in asbestos roof detection, outperforming traditional remote sensing techniques, and offers an effective, automated solution for targeting removal efforts and mitigating associated health risks. This end-to-end approach fills a gap in the existing literature and presents a promising direction for future research in asbestos roof detection.