CarDreamer - Diseños de automóviles generados por una red neuronal
Fecha
2023-10-26
Autores
Título de la revista
ISSN de la revista
Título del volumen
Editor
Resumen
En este TFG, se aborda el desafiante problema de generar imágenes realistas de automóviles utilizando Redes generativas adversariales (GAN). Tras un breve repaso del Machine Learning, se profundiza en el funcionamiento de
las GAN en los siguientes capítulos. Las GAN son redes neuronales altamente efectivas en la generación de imágenes
realistas a partir de conjuntos de entrenamiento. El objetivo del TFG es diseñar, desarrollar y evaluar un sistema de
generación de imágenes de alta calidad de automóviles mediante GAN. Para ello, se busca un conjunto de datos de
entrenamiento que contenga imágenes reales de automóviles y se optimizará el modelo ajustando parámetros y
arquitecturas. Además, se explorarán alternativas de modelos generativos a las GAN para obtener una comparativa de
rendimiento Este TFG se motiva por el interés en inteligencia artificial, aprendizaje profundo y minería de datos, al
abordar un problema complejo que implica algoritmos y modelos avanzados.
This Bachelor's Thesis addresses the challenging problem of generating realistic images of cars using Generative Adversarial Networks (GAN). After a brief overview of Machine Learning, the functioning of GANs is further explored in the following chapters. GANs are highly effective neural networks for generating realistic images from training sets. The objective of this Bachelor's Thesis is to design, develop and evaluate a high-quality image generation system for cars using GANs. To accomplish this, a training dataset containing real car images will be sought and the model will be explored in order to obtain a performance comparison. This Bachelor's Thesis is motivated by the interest in artificial intelligence, deep learning and data mining, by tackling a complex problem that involves advanced algorithms and models
This Bachelor's Thesis addresses the challenging problem of generating realistic images of cars using Generative Adversarial Networks (GAN). After a brief overview of Machine Learning, the functioning of GANs is further explored in the following chapters. GANs are highly effective neural networks for generating realistic images from training sets. The objective of this Bachelor's Thesis is to design, develop and evaluate a high-quality image generation system for cars using GANs. To accomplish this, a training dataset containing real car images will be sought and the model will be explored in order to obtain a performance comparison. This Bachelor's Thesis is motivated by the interest in artificial intelligence, deep learning and data mining, by tackling a complex problem that involves advanced algorithms and models
Descripción
Palabras clave
Tratamiento inteligente de la Información