YOLO: Roof Material Detection Using Aerial Imagery

Roofing materials are useful in assessing poverty, suitability of a building after a natural disaster and urban development. Planning urbanization has made it difficult and cumbersome to collect data on roofing materials. The earth information provided through remote sensing, are used by the researchers in detecting the roofing materials through various methods preferred over pixel-based methods. In Machine Learning (ML) based material classification feature extraction and choice of features are labour intensive. Hence, in this work, an attempt is made to categorize the buildings into metallic, and concrete roofs using Deep Learning (YOLOv7) modelThe result analysis inferred that the CIs_loss in YOLOv7 reduced by 99.157% compared to the average YOLOv5 and YOLOv8, and demonstrated an mAP enhancement of 32.52% over YOLOv5 and 6.06% over YOLOv8. Additionally, YOLOv7 exhibits an average decrease of 62.44% in time per iteration compared to both YOLOv5 and YOLOv8. Thus, the results prove that the YOLOv7 model can be used for material identification of roofs.