Concrete structures frequently experience cracks, and assessing their effects is crucial to maintaining long-term safety and halting structural deterioration. Automation has made it possible to identify and track cracks more effectively, which has made inspection procedures much easier. To detect surface damage in concrete materials, deep convolutional neural networks (CNNs) have been extensively investigated. In order to distinguish between crack and non-crack images, this study uses deep learning techniques, namely a convolution-based architecture in conjunction with transfer learning. The SDNET2018 database was utilized to supply large and varied samples for training and testing as these models require substantial amounts of data. This publicly available dataset contains thousands of images of concrete surfaces. It was used to refine two pre-trained architectures, VGG19 and ResNet50, which achieved test accuracy of 88.80% and 88.93%, respectively. The outcomes validate the potential of deep CNN models for structural monitoring applications and show how effective transfer learning is for automated crack detection.