Clothing Image Classification Using VGG-19 Deep Learning Model For E-commerce Web Application

Abstract

Artificial Intelligence applications become one of the most important tools that help to increase the profits of e-commerce stores. It is expected that AI methods will push this aspiration to an even higher level. This study aims at classifying images of clothing products using deep learning (DL) techniques while embedding them in an e-commerce web application. We utilize deep learning techniques to determine the type of clothing products in the image, such as shirts, dresses, pants, shoes, etc. This study performs several steps, including requirements collecting and modeling, DL model training using two deep learning models, and then testing the models and the system�s accuracy on a set of images. We have used two deep learning models improved from classic Convolutional Neural Networks (CNN). The Convolutional Neural Network models are achieving high accuracy on image classification tasks. Therefore, the literature proposes many suggested improvements to CNN architecture, such as Xception and VGG-19 architectures. In this study, we have selected a VGG-19-based clothing image classification. We found out that the VGG-19 model outperforms the Xception model. Therefore, the trained VGG-19 model is incorporated into a clothing store web application for classifying clothing image products. Testing accuracy is found, and then a manual test of the system accuracy is achieved using in-lab sample images.