A novel SCD11 CNN model Performance Evaluation with Inception V3, VGG16 and ResNet50 using Surface Crack Dataset

Image processing techniques can be used to detect potential defects in images of infrastructure elements that have been collected or scanned. Aside from image processing, machine learning technologies are increasingly being applied to increase crack detection performance and resilience. A novel Surface Crack Detection Convolutional Neural Network (CNN) named the SCD11 CNN model is proposed. The main aim of this research paper is to improve efficiency and minimize the loss rate. The proposed CNN model is compared with the CNNs, including Inception V3, VGG16, and ResNet50, using a Surface Cracks Dataset. The accuracy and loss of the proposed model are compared. The results show that the proposed SCD11 CNN model performs better in test accuracy than Inception V3 and VGG16 and has a slightly lower accuracy than ResNet50.