Year: 2016
Type: dissertation
Abstract: Deep learning has become part of many state-of-the-art systems in multiple disciplines (specially in computer vision and speech processing). In this thesis Convolutional Neural Networks are used to solve the problem of recognizing people in images, both for verification and identification. Two different architectures, AlexNet and VGG19, both winners of the ILSVRC, have been fine-tuned and tested with four datasets: Labeled Faces in the Wild, FaceScrub, YouTubeFaces and Google UPC, a dataset generated at the UPC. Finally, with the features extracted from these fine-tuned networks, some verifications algorithms have been tested including Support Vector Machines, Joint Bayesian and Advanced Joint Bayesian formulation. The results of this work show that an Area Under the Receiver Operating Characteristic curve of 99.6% can be obtained, close to the state-of-the-art performance.
Author: Carlos Roig Marí
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W2573170414