“Construcción de una aplicación móvil con redes neuronales convolucionales para reconocer imágenes de prendas de vestir de tiendas en línea en Ecuador”

Abstract

The objective of this degree project is to build a mobile application with convolutional neural networks to recognize images of clothing from online stores in Ecuador, using the agile development methodology SCRUM. For this reason, an application was developed with neural networks of the FASHION MNIST dataset, with the services of Flask and Nest Js, the model of a non-relational base in Mongo DB and the framework in Ionic. The final product are two APK of the Android and iOS operating system, which allows the registration of an administrator to make the entry of urls that you want to scrape the images found on the web pages of the stores in Ecuador and save them in a database. And for other users to take and upload photo to be able to predict similar garments of the ones they want to search. In the black box testing section it is verified that the functional requirements were met in each of the planned phases, with the load tests the amount of requests that the application can support was identified from a minimum of 100 which indicates 0% errors and a maximum of 10000 with a 4.13% error rate and the stress test consists of testing the limits that the application supports by testing a maximum of 100000 requests the application does not work properly the requests sent.