Publication: Técnicas inteligentes con imágenes para dispositivos móviles
Loading...
Official URL
Full text at PDC
Publication Date
2021
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citations
Exportar
Abstract
Los dispositivos móviles inteligentes tales como smartphones o tablets han evolucionado notablemente tanto en su forma física como en sus funcionalidades y capacidades de procesamiento. Esta evolución ha permitido que las personas puedan realizar más y más tareas con un dispositivo que cabe en la palma de sus manos y que antes sólo se podía llevar a cabo con dispositivos de gran tamaño, como por ejemplo un ordenador personal, o una videoconsola.
En el presente trabajo se propone el diseño de una aplicación para dispositivos móviles iOS que permite el reconocimiento de cinco actividades físicas realizadas por personas que portan el dispositivo.
Para ello, se propone un modelo de Red Neuronal Residual (ResNet) dentro del Aprendizaje Profundo. La red recibe como entradas secuencias de imágenes o vídeos, captados con el dispositivo móvil con los que se realiza el proceso de entrenamiento y decisión. Con tal propósito, se ha diseñado una estrategia que integra el modelo de red indicado, bajo la cobertura de una aplicación informática, para realizar el procesamiento a nivel local, lo que representa un avance importante desde el punto de vista de la privacidad de los datos, a la vez que se mantiene la efectividad de procesamiento.
Los resultados obtenidos respecto de la toma de decisiones corroboran esta circunstancia.
Smartphones or tablets have notably evolved in their physical form, in their functionalities and processing capabilities. This evolution has allowed people to perform more and more tasks on a device that fits in the palm of their hands that previously could only be done on large devices such as a personal computer or a videogame console. This work, propose the design of an application for iOS mobile devices that allows the recognition of five physical activities carried out by people who carry the device. For this, a Residual Neural Network (ResNet) model is proposed within Deep Learning. The network receives as inputs sequences of images or videos, captured with the mobile device with which the training and decision process is carried out. For this purpose, a strategy has been designed that integrates the indicated network model, under the cover of a computer application, to perform the processing at the local level, which represents an important advance from the point of view of data privacy. , while maintaining processing effectiveness. The results obtained regarding decision-making corroborate this circumstance.
Smartphones or tablets have notably evolved in their physical form, in their functionalities and processing capabilities. This evolution has allowed people to perform more and more tasks on a device that fits in the palm of their hands that previously could only be done on large devices such as a personal computer or a videogame console. This work, propose the design of an application for iOS mobile devices that allows the recognition of five physical activities carried out by people who carry the device. For this, a Residual Neural Network (ResNet) model is proposed within Deep Learning. The network receives as inputs sequences of images or videos, captured with the mobile device with which the training and decision process is carried out. For this purpose, a strategy has been designed that integrates the indicated network model, under the cover of a computer application, to perform the processing at the local level, which represents an important advance from the point of view of data privacy. , while maintaining processing effectiveness. The results obtained regarding decision-making corroborate this circumstance.
Description
Trabajo de Grado en Ingeniería del Software, Facultad de Informática UCM, Departamento de Ingeniería del Software e Inteligencia Artificial, Curso 2020/2021.