Aristotle University Of Thessaloniki (AUTH)
Μέθοδοι μηχανικής μάθησης για σημασιολογική ανάλυση βίντεο
Abstract
dc:descriptionThis dissertation studied machine learning and computer vision methods for semantic image and video analysis. More specifically, the problems of face recognition, human action recognition, object recognition and video summarization were studied as classification problems, along with the problems of privacy protection and robustness to adversarial attacks. Supervised, semi-supervised and deep learning algorithms were proposed, in the context of one-class classification, multiclass classification, adversarial attacks and defenses. Novel methodologies that provide increased classification performance were developed, attributed to the introduction of additional regularization terms to the optimization problems of well-known learning algorithms, in order to exploit geometric training data relationships. Finally, a novel adversarial attack methodology that introduces privacy protection criteria was developed, along with a mechanism that increases the classification model robustness against adversarial attacks, by incorporating a novel neural network layer based on one-class classifiers.
Degree
thesis:*- Grantor dc:publisher
- Aristotle University Of Thessaloniki (AUTH)
- Year dc:date
- 2019
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Mygdalis, Vasileios
- Μυγδάλης, Βασίλειος
Subjects
dc:subject × 12Rights
dc:rights- Statement dc:rights
-
- BY
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/46942