This research introduces a novel approach that combines state-of-the-art convolutional neural networks (CNNs) with conventional image processing approaches to enable real-time defect detection and severity estimate in painted surfaces. Our method measures fracture severity and reliably finds faults. By utilizing a trained CNN model for defect detection and combining edge detection, contour analysis, and Euclidean distance computations for crack measurement, the system provides excellent precision and efficiency. The method's accuracy and robustness are validated by experiments in a variety of real-world applications, such as transportation safety, industrial quality control, and infrastructure maintenance. The instantaneous feedback provided by the suggested system's real-time capabilities enables preventive maintenance and raises industry safety requirements. In order to obtain better performance in real-world applications, this study emphasizes the value of combining deep learning with traditional techniques and emphasizes the promise of hybrid approaches in defect identification.