Klasifikasi model pakaian menggunakan Convolational Neural Network

Abstract

As the e-commerce industry continues to grow rapidly, the need for automated clothing classification systems is becoming increasingly urgent to improve user experience and efficiency in product search. Classification carried out manually or using traditional methods is often less efficient and accurate, especially considering the wide diversity of clothing models. This research aims to increase the level of accuracy and efficiency in clothing model classification through the use of the Convolutional Neural Network (CNN) method, which is known to be effective in identifying complex visual patterns. This research involves a series of steps, starting from collecting datasets, preprocessing data, designing the CNN architecture, training the model using several convolution layers and adding regularizers to prevent overfitting, to evaluating model performance with accuracy and precision metrics. The addition of regularizers to the CNN model succeeded in increasing accuracy up to 89.59% on the test set, and improving performance stability in clothing image classification. CNN models show great potential in automating clothing classification in the e-commerce sector, supporting better user experience and increasing operational efficiency. It is hoped that the development of a CNN-based automatic classification system can improve services on e-commerce platforms, speed up the product search process, and inspire further innovation in image recognition and classification in the digital industry.