Numerous buildings in Bangladesh were constructed without following standard building codes. As a result, those are vulnerable to increased earthquake frequency and variable loads. To address this issue, old buildings need to be monitored frequently by non-destructive testing (NDT) to avoid any failures. The identification of cracks and dampness is one of the most important parts of this test. Generally, this detection part is costly and time-consuming to conduct it manually. To resolve this, the current study has been performed for intelligent structural damage identification based on deep learning techniques. A state-of-the-art Convolutional Neural Network (CNN)-based object detection model YOLOv4-tiny has been used to detect cracks and dampness and repair cost analysis of damages. The study outcome suggests that using our proposed deep model, it is possible to detect building cracks with a mean Average Precision (%mAP) of 72.46%. In addition to traditional structural health monitori