In order to evaluate long-term structural performance and reduce the risk of collapse, cracks in concrete buildings must be examined. The process of inspecting concrete structures has become much simpler as a result of the development of automation, which has made it possible to find and see structural cracks. Deep convolutional neural networks (CNNs) have been particularly investigated by researchers for the detection of surface degradation in concrete. This article proposes using deep learning technology, specifically the convolution-based neural architectures and Transfer Learning, to identify images of cracks. For training and testing, these models need a sizable database. However, gathering photos by hand can be difficult. As a result, this study proposes using the combination of two publicly accessible databases to categorize cracks in concrete structures. An accuracy level of roughly 98.17% was attained by using a pre-trained model, VGG16, indicating the model's superior performance.