Automated Crack Detection via Semantic Segmentation Approaches Using Advanced U-Net Architecture

Cracks affect the robustness and adaptability of various infrastructures, including buildings, bridge piers, pavement, pipelines. Therefore, reliability automated crack detection are essential. In this study, we conducted image segmentation using datasets by applying advanced architecture U-Net. First, collected integrated from prior studies, cracks in buildings pavements. For effective localization cracks, used U-Net-based neural networks, ResU-Net, VGGU-Net, EfficientU-Net. The models were evaluated five-fold cross-validation several evaluation metrics mean pixel accuracy (MPA), intersection over union (MIoU), confusion matrix. results dataset showed that ResU-Net (68.47%) achieves highest MIoU with a relatively low number parameters compared to VGGU-Net (67.71%) EfficientU-Net (68.07%). addition performance, lowest test runtime, 40 milliseconds per single image, true positive rate 45.00% pixel-wise recognition test. As trained validated diverse surfaces, proposed approach can be as pre-trained model task few data sources. Furthermore, both practical managerial implications discussed herein.