Diseño de una Arquitectura de Red Neuronal Convolucional para la clasificación de objetos

Abstract

One of the most important and fundamental problems in the area of Computer Vision is object detection. There are a large number of applications that require seek objects in a scene and then classifying them, considering the complexity that exists when there are several categories. The deep learning techniques have emerged as a very powerful strategy in the automatic feature extraction from images, causing significant improvements in the general problem of object detection. The goal of this article is to present the design of a Convolutional Neural Network architecture suitable for classifying 6 different categories of common objects: bed, stairs, table, door, chair and sofa. The results obtained indicate a precision greater than 90% in the experiments carried out.