|
A Comparative Study of CNN Models for Crack Classification in Buildings
|
Nooruldeen Sameer Majeed   , Mohammed Mesgari * , Hayder Dibs |
|
|
|
Abstract: (6 Views) |
Cracks have become known as an important indicator of a building's state, and they can be categorized into several different types. Damage can occur due to various factors, including the building's age, design, the characteristics of the soil beneath the building, and environmental influences. For instance, Cracks resulting from seismic activity on structures, such as buildings, pose a serious risk and could potentially result in a structural collapse if not addressed. Based on the severity of their impact, the present study divided the cracks into four groups. In deep learning, we have used four models (VGG16, Alexnet, Resnet50, and CNN modification model) to compare their accuracy and completion times. The results demonstrated that the VGG16 model was more accurate. (98.32%),On the other hand, Resnet50 was the least accurate (75%). Nevertheless, the combination of the models offers an accuracy rate of 91%. The results demonstrated the efficacy of deep learning in rapidly and precisely detecting and classifying cracks.
|
Article number: 7 |
| Keywords: Convolutional neural network(CNN), crack classification, Deep learning, VGG16, Alexnet. |
|
|
Full-Text [PDF 625 kb]
(10 Downloads)
|
Type of Study: Research |
Subject:
Photo&RS
Received: 2024/08/12 | Accepted: 2025/12/23
|
|
|
|
|
|
|
| Send email to the article author |
|
|
|
| Add your comments about this article |
|
|
|
