Método para la clasificación de imágenes de moda con TensorFlow

Guillermina Muñoz Zamora; Fernanda Teresa Paredes Miranda; Sigifredo García Alva; Jesús Raúl Cruz Rentería

doi:10.46588/invurnus.v18i1.88

Authors

Guillermina Muñoz Zamora Tecnológico Nacional de México Campus Nogales, Posgrado e Investigación.
Fernanda Teresa Paredes Miranda Tecnológico Nacional de México Campus Nogales, Posgrado e Investigación.
Sigifredo García Alva Tecnológico Nacional de México Campus Nogales, Posgrado e Investigación.
Jesús Raúl Cruz Rentería Tecnológico Nacional de México Campus Nogales, Posgrado e Investigación.

DOI:

https://doi.org/10.46588/invurnus.v18i1.88

Keywords:

TensorFlow, Fashion classification, clothes classification

Abstract

In the fashion business there is a wide variety of clothing, each with its own name to identify what type of garment it is. This article describes a method to generate a model that identifies the type of clothing based on 15 categories, applying the technology of image identification and machine learning such as convolutional neural networks under the use of the Google TensorFlow library to develop a fashion clothing classifier. Concluding after the testing phase 95% accuracy was achieved. While with an external test set 70% was achieved, considering the quality of the images as one of the possible reasons for failure.