Modeling and Simulation of Convolutional Neural Network (CNN) for Clothing Image Classification

Abstract

This research aims to design, implement, and evaluate a Convolutional Neural Network (CNN) model within the context of clothing image classification, with a core focus on the principles of Modeling and Simulation of computational systems. Accuracy in clothing image classification is an essential requirement to support automation in the e-commerce sector and digital management. The model is implemented using the MobileNetV2 Transfer Learning architecture, enhanced with GlobalAveragePooling2D, a Dense Layer, and Dropout for feature optimization. The dataset utilizes a total of 2000 clothing images evenly distributed across 4 clothing classes. Following preprocessing and data augmentation, the model was trained over 250 epochs. The simulation results indicate excellent model performance, substantiated by a Training Accuracy of 98.05% and a Validation Accuracy reaching 91.75%. This high accuracy validates the computational efficiency of the developed model and affirms the capability of CNN through the Transfer Learning approach as a viable solution for complex visual classification problems.