Object-based Surface Defect Detection Using Deep Learning

Abstract

Surface defect detection is a critical task in quality assurance across various industrial sectors, manufacturing, automotive, and electronics. Traditional manual inspection methods are often time-consuming, inconsistent, and prone to human error. This project presents a real-time, object-based surface defect detection system leveraging deep learning and embedded hardware to automate and enhance the inspection process. The model is evaluated based on precision, recall, and F1-score, demonstrating high performance in identifying and localizing defects in real-time. Experimental results show that our approach outperforms traditional computer vision methods, achieving over 95% accuracy in defect classification. The system is scalable for industrial applications, enabling efficient quality assessment with minimal manual intervention. This work highlights the potential of deep learning in automating defect detection, reducing production costs, and improving manufacturing standards. Future research will focus on enhancing generalization across diverse materials and defect types.