In regions prone to natural disasters, the buildings must follow specific construction standards to avoid demolition. One of the factors that predict the risk of damage is the roof material. This paper investigates the performance of various deep convolutional neural network architectures to classify buildings based on roof material from aerial drone imagery. We also propose a method that is an ensemble of ResNetResNet, ResNeXtResNeXt, and EfficientNetEfficientNet variants of convolutional neural networks, which performed the best in our experiments. We obtained a log loss value as low as 0.4373 using the proposed method. Therefore, the proposed method can be used to perform an accurate classification of roof material using aerial drone imagery.
Disaster Risk Mapping from Aerial Imagery Using Deep Learning Techniques
Lecture Notes in Civil Engineering
International Conference on Unmanned Aerial System in Geomatics ; 2021 ; Roorkee, India April 02, 2021 - April 04, 2021
2023-03-16
11 pages
Aufsatz/Kapitel (Buch)
Elektronische Ressource
Englisch
Object detection in synthetic aerial imagery using deep learning
British Library Conference Proceedings | 2023
|Deep Learning based Vehicle Detection in Aerial Imagery
GWLB - Gottfried Wilhelm Leibniz Bibliothek | 2022
|