Por favor, use este identificador para citar o enlazar a este item:
http://hdl.handle.net/10261/304911
COMPARTIR / EXPORTAR:
SHARE BASE | |
Visualizar otros formatos: MARC | Dublin Core | RDF | ORE | MODS | METS | DIDL | DATACITE | |
Título: | Image Segmentation on Wind Turbine Blades using Generative Adversarial Networks |
Autor: | Ortega Rodríguez, Carles | Director: | Agudo, Antonio CSIC ORCID | Fecha de publicación: | 14-sep-2022 | Editor: | CSIC-UPC - Instituto de Robótica e Informática Industrial (IRII) Universidad Politécnica de Cataluña |
Resumen: | Image segmentation is a method in computer vision consisting in labeling different objects in an image. For this purpose, numerous algorithms have been developed, and, in the last years, especially based in convolutional neural networks. In this work we will analyze a relatively novel model known as Generative Adversarial Networks and use it to solve a real problem. The first part of the work will introduce the reader from scratch to machine learning and specially to this type of networks, GANs. Besides that, we will deepen in the challenges of training these models and how we can solve this issues. Secondly, we will use a GAN to segment blades from backgrounds in wind turbine pictures, in order to do the first step to develop a tool which detects defects in blades structure automatically. Then, we will evaluate this segmentations produced by our model in comparison with U-Net, a well-known neural network which is used to the same purpose. | Descripción: | Trabajo fin de grado presentado en la Universidad Politécnica de Cataluña, Grado en Matemáticas | Versión del editor: | http://hdl.handle.net/2117/373845 | URI: | http://hdl.handle.net/10261/304911 |
Aparece en las colecciones: | (IRII) Tesis |
Ficheros en este ítem:
Fichero | Descripción | Tamaño | Formato | |
---|---|---|---|---|
accesoRestringido.pdf | 15,38 kB | Adobe PDF | Visualizar/Abrir |
CORE Recommender
NOTA: Los ítems de Digital.CSIC están protegidos por copyright, con todos los derechos reservados, a menos que se indique lo contrario.