Garment classifier model based on Neural Networks (#1293)

Authors

Arroyo-Paz, Antonio

Aleman-Gonzales, Leonid

Ingaluque-Arapa, Marga

Tapia-Catacora, Pablo

Jimenez-Chura, Adolfo

Marca-Maquera, Hugo

Rodriguez-Aburto, Cesar

Abstract

Clothing sorting is a process that revolutionizes the organization of closets and provides a better shopping experience in e-commerce. One of the technological alternatives to address this process are Convolutional Neural Networks (CNN) due to their optimal understanding of particular characteristics through a visual representation, such as an image. This paper aims to implement a CNN model that processes information from clothing images through hidden layers to identify distinctive patterns that allow their efficient categorization. A methodology based on the construction of a classificatory model using the Python programming language, the "TensorFlow" and "TensorFlow Datasets" libraries, the Google Colaboraty virtual machine and images taken from the public repository "Fashion-MNIST", belonging to the online clothing store "Zalando", was used. As a result, a functional CNN with an accuracy of 88.57% was obtained. Finally, this work is considered as a reference article for future works that focus on the practical utilities of a CNN in different fields.