Abstract

In the modern industries, surface defects in metal sheets significantly affect the quality, safety, usability and aes- thetics of the products. The steel is the most important building material in many manufacturing firms. Detecting quality issues of steel products and classifying steel defects is a challenging task and time-consuming manual effort. Recent progress in AI makes it possible to use advanced deep learning technologies for visual quality inspection for defect classification. The proposed system automates steel surface defect classification using VGG16 as a feature extractor to correctly classify the defect present based on the mapped features. The correctly trained neural network for the system achieves an accuracy of 97% ensuring higher precision of quality management system.