A Design and Training Application for Deep Convolutional Neural Networks and Support Vector Machines Developed on MATLAB

This paper presents a user-friendly design application development environment based on MATLAB that facilitates two applications using convolutional neural networks (CNNs) and support vector machines (SVMs). Firstly, an application of deep CNN (DCNN) for visual inspection is developed and is trained using a large number of images to inspect undesirable defects such as crack, burr, protrusion, chipping, spot and fracture phenomena which appear in the manufacturing process of resin molded articles. The DCNN is named sssNet. Then, two kinds of SVMs are respectively incorporated with two trained DCNNs, i.e., our designed sssNet and well-known AlexNet, to classify test images with high recognition rate into accept as OK or reject as NG categories, in which compressed features obtained from the DCNNs are used as inputs for the SVMs. The usability and operability of the developed design and training application for DCNNs and SVMs are demonstrated and evaluated through training and classification experiments.