Deep Learning Models for Metal Surface Defect Detection

This paper presents the initial results of the research on the application of neural network architectures for the prediction of metal surface defects, namely a classical convolutional neural network, a manually crafted VGG architecture, and a pre-trained deep neural network (InceptionV3). The investigation involves six classes of defect: Crazing, Inclusion, Patches, Pitted, Rolled and Scratches. The importance of this research lies in addressing the critical need for efficient and accurate detection of metal surface defects in manufacturing industries, while also considering the computational cost of training of different Deep Learning architectures. Detecting these defects early is crucial for ensuring product quality, reducing waste, and minimizing potential safety hazards. Traditional inspection methods are often time-consuming and prone to human error. This research not only provides practical insights for industrial applications but also contributes to the ongoing discourse on the suitability of different neural network architectures for image-based classification tasks in manufacturing industries.