Reconocimiento de imágenes de monumentos del patrimonio cultural de Madrid mediante técnicas de aprendizaje profundo
Loading...
Official URL
Full text at PDC
Publication date
2024
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citation
Abstract
En el presente trabajo se desarrolla una aplicación de reconocimiento de monumentos en la ciudad de Madrid mediante el uso de técnicas de reconocimiento de imágenes basadas en aprendizaje profundo.
Para ello se utilizan Redes Neuronales Convolucionales, concretamente AlexNet, GoogLeNet, SqueezeNet e InceptionV3. Dichas redes poseen la capacidad de ser entrenadas con una serie de parámetros para que aprendan características de imágenes de monumentos de la ciudad de Madrid, como parte de su patrimonio cultural. El objetivo es identificar un monumento, mediante el reconocimiento de imágenes basado en las mencionadas técnicas.
Antes de comenzar el entrenamiento el usuario puede configurar una serie de parámetros tales como el algoritmo de optimización, el tamaño del mini lote, el número de épocas o la tasa de aprendizaje. Tras el entrenamiento también se puede realizar una clasificación de imágenes mediante el modelo entrenado para observar los resultados.
Por último, la principal funcionalidad consiste en que una vez que el modelo ha sido entrenado por el usuario según la arquitectura de red y los parámetros elegidos, el modelo puede identificar un monumento a partir de una imagen de este, que constituye la entrada al modelo entrenado.
Adicionalmente, cuando el modelo reconoce un monumento, se le ofrece al usuario un contexto cultural e histórico para que pueda obtener información adicional acerca del monumento reconocido.
The objective of this project has been the development of an application for recognizing monuments in Madrid using Deep Learning techniques. Within the context of Deep Learning, the application makes use of Convolutional Neural Networks (CNN) such as AlexNet, GoogLeNet, SqueezeNet and InceptionV3. These models have the capability to be trained with a set of parameters so that they can learn characteristics from a series of images of monuments of Madrid as part of Madrid’s cultural heritage and identify a monument through image recognition. Before starting the training, users can configure several parameters such as the solver method for optimization, the mini-batch size, the number of epochs, or the learning rate. After training, image classification can also be performed by the model to observe the results. Finally, the primary use of the application is that once the model has been trained by the user according to the net architecture and the chosen parameters, the model can identify a specific monument by analyzing an input image of it. Additionally, when a model recognizes a monument, the user is provided with cultural and historical context, allowing them to obtain additional information about the recognized monument.
The objective of this project has been the development of an application for recognizing monuments in Madrid using Deep Learning techniques. Within the context of Deep Learning, the application makes use of Convolutional Neural Networks (CNN) such as AlexNet, GoogLeNet, SqueezeNet and InceptionV3. These models have the capability to be trained with a set of parameters so that they can learn characteristics from a series of images of monuments of Madrid as part of Madrid’s cultural heritage and identify a monument through image recognition. Before starting the training, users can configure several parameters such as the solver method for optimization, the mini-batch size, the number of epochs, or the learning rate. After training, image classification can also be performed by the model to observe the results. Finally, the primary use of the application is that once the model has been trained by the user according to the net architecture and the chosen parameters, the model can identify a specific monument by analyzing an input image of it. Additionally, when a model recognizes a monument, the user is provided with cultural and historical context, allowing them to obtain additional information about the recognized monument.
Description
Trabajo de Fin de Grado en Ingeniería del Software, Facultad de Informática UCM, Departamento de Ingeniería del Software e Inteligencia Artificial, Curso 2023/2024