|
|
Diffusion Models and their Impact on Cybersecurity
Dvorščák, Patrik ; Homoliak, Ivan (oponent) ; Lapšanský, Tomáš (vedoucí práce)
This thesis explores the performance of diffusion models (DMs) and generative adversarial networks (GANs) in creating AI-generated visual content across multiple applications, including face synthesis, text-to-image generation, artistic rendering, image-to-image translation, video synthesis, and super-resolution. Through comparative experiments, this research evaluates the models' ability to generate detailed, realistic, and artistically compelling visuals from textual and image prompts. The results reveal that DMs excel in producing highly detailed images that closely follow text prompts, particularly effective in face synthesis and text-to-image tasks. In contrast, GANs are more adept at rendering realistic environmental scenes, suitable for applications requiring immersive visuals. Both model types are competent in artistic rendering, though they differ in style adaptation and creativity. The thesis concludes with future research directions aimed at enhancing model efficacy and integrating these technologies more effectively into practical applications.
|
|
|
Doplnění chybějící části obrazu pomocí hlubokého učení
Zobaník, Radek ; Kubík, Tibor (oponent) ; Šilling, Petr (vedoucí práce)
V této práci vznikla aplikace pro testování a porovnávání metod pro doplnění chybějící části obrazu za využití hlubokého učení a byly natrénovány dvě metody, pconv s konvoluční architekturou, respektive AOT-GAN s GAN architekturou. Práce popisuje návrh výsledné aplikace, její funkcionalitu a důležité body implementace. Byla zvolena datová sada, na které byly vybrané modely optimálně natrénovány. Proběhly experimenty na AOT-GAN modelu, kdy se zkoumal vliv počtu AOT bloků v generátoru na výsledný doplněný obraz. Všechny experimenty byly kvalitativně a kvantitativně porovnány. Výsledky ukázaly úctyhodné výsledky při práci s přírodní scenérií.
|
|
|
Brute Force Attack on Fingerprint Access System Using Genetic Algorithms
Keszi, Marián ; Rydlo, Štěpán (oponent) ; Kanich, Ondřej (vedoucí práce)
This work deals with a brute force attack on a fingerprint-based access system using genetic algorithms. It includes the design of a genetic algorithm as a mean to perform the brute force attack. In the study, a generative adversarial network was used as a generator of synthetic fingerprints trained on the SOCOFing dataset. Experiments were performed focusing on the possibility of inserting known information about a fingerprint fragment into the input vector, as well as experiments using advanced methods of the genetic algorithm to modify the input vector with the goal of overcoming the VeriFinger algorithm. The experiments led us to the conclusion that even with the help of a generative adversarial network and genetic algorithm, we were unable to surpass the VeriFinger algorithm.
|
|
|
Porovnání metod pro doplnění chybějící části obrazu založených na hlubokém učení
Rajsigl, Tomáš ; Herout, Adam (oponent) ; Španěl, Michal (vedoucí práce)
Cílem této bakalářské práce je porovnat metody hlubokého učení pro doplnění chybějící části obrazu pomocí kvantitativních metrik jako jsou PSNR, SSIM a LPIPS. Pro dodatečné subjektivní ohodnocení byla taktéž provedena uživatelská studie. K porovnání byly použity celkem čtyři neuronové sítě založené na architektuře GAN. Navrhovaná architektura neuronové sítě a její modifikované verze byly porovnávány oproti síti AOT-GAN. Experimenty ukázaly, že v obrazech s malou chybějící částí dosáhla varianta navržené metody 29% zlepšení oproti již zmiňované metodě AOT-GAN. Toto tvrzení podporují i výsledky uživatelské studie, kde byla tato metoda vyhodnocena jako nejlepší. V rámci této práce vznikla malá datová sada určená pro vyhodnocení metod retušování obrazu při úloze odstraňování objektů. Reálné využití těchto metod je demonstrováno prostřednictvím webové aplikace.
|
|
|
Antropometrická evaluace generovaných snímků obličeje
Mikyšek, Jakub ; Rydlo, Štěpán (oponent) ; Goldmann, Tomáš (vedoucí práce)
Tato práce se zabývá porovnáním uměle vygenerovaných obličejů s reálnými snímky pomocí analýzy klíčových bodů obličeje a měřením proporcí mezi těmito body. Zkoumá také oblasti týkající se umělých neuronových sítí, přičemž se zaměřuje na sítě GAN, které dokážou generovat uměle vytvořené snímky obličeje. Zkoumá jejich proces, architekturu a dostupné modely. Cílem práce je zhodnotit, jak se uměle vygenerované snímky liší od reálných, a zjistit, v jakých proporcích je rozdíl největší.
|
|
|
Defect localization and analysis in GaN
Gazdík, Richard ; Šik, Ondřej (oponent) ; Bábor, Petr (vedoucí práce)
This master’s thesis is concerned with the localisation and analysis of threading dislocations in GaN epitaxial layers. The thesis is divided into two parts – theoretical and experimental. The theoretical part explains the origin and nature of threading dislocations. Additionally, it lays foundations for a better understanding of perhaps less known techniques, which can be used to study them – electron channeling contrast imaging and defect-selective etching. The experimental part describes the procedures done to carry these techniques, in addition to TEM diffraction-contrast imaging and its associated FIB sample preparation, out. We show that each of the techniques can be used independently to characterize threading dislocations, but that there is a possibility to gain complementary information by combining them.
|
|
|
Deposition of GaN nanocrystals with Ag nanoparticles
Michalko, Matej ; Dvořák, Petr (oponent) ; Čalkovský, Vojtěch (vedoucí práce)
The presented bachelor thesis deals with the preparation of GaN nanocrystals with Ag nanoparticles. In the theoretical part GaN, its properties and applications are introduced. Furthermore, possible substrates for GaN growth and preparation methods are presented, with a main focus on MBE. The thesis also discusses the basic principles of photoluminescence and its enhancement using metal nanoparticles. The experimental part deals with the preparation of GaN nanocrystals with Ag nanoparticles. First, Ga islands are deposited on Si(111) substrate, the second step is Ga nitridation. Subsequently, Ag nanoparticles of different sizes are deposited. The different steps were optimized and measured by analytical techniques such as SEM and photoluminescence. Finally, the dependence of the photoluminescence enhancement on the size of Ag nanoparticles was established.
|
|
|
Photo Livening Application
Bobola, Adrián ; Šalko, Milan (oponent) ; Malinka, Kamil (vedoucí práce)
The goal of this work is to create a web application for animating static photographs. The application allows users to animate their portraits and group photos. Users can upload their own motion, which they want to use, and the application will use it to animate a selected part of the uploaded photo. The faces in the photos are automatically detected, while users have the option of manually marking faces in a given photo. The application supports recording custom motion from a video file or directly from a webcam. The server-side of the application is implemented in Python using the Django framework. The client-side of the application utilizes JavaScript and the React framework. Communication between the client and the server is facilitated via REST API. The thesis also provides an overview of existing tools of similar types, compares them, and discusses identified shortcomings. Additionally, the thesis explains the techniques and principles used to animate static photographs.
|
|
|
Reconstruction of Damaged Parts of Fingerprint Image Using Neural Nets
Bobocký, Boris ; Dyk, Tomáš (oponent) ; Kanich, Ondřej (vedoucí práce)
In this paper, I propose a method for reconstructing damaged fingerprints using generative adversarial networks (GANs), implemented with Python and the PyThorch library. I have trained a specific GAN model on a dataset of approximately twenty thousand prints, created with Anguli and other damage simulation tools. This approach produced excellent results and could have wide application in biometric systems. This work highlights the potential of deep learning in the fields of image reconstruction and biometrics.
|
|
|
Preparation and characterization of nanostructured III-V semiconductor materials
Maniš, Jaroslav ; Kostelník,, Petr (oponent) ; Hospodková,, Alice (oponent) ; Šikola, Tomáš (vedoucí práce)
The aim of the presented PhD thesis was to develop and analyze gallium nitride (GaN) nanostructures in three different forms. Firstly, three dimensional GaN nanocrystals prepared on graphene were studied from the perspective of the intrinsic crystal properties as well as growth statistics. Adopting the method of droplet epitaxy allowed the formation of such nanostructures at a low substrate temperature (T = 200°C). In order to demonstrate possible applications, the proof of concept of an UV sensitive device was designed and tested with the successful results and the great promise to the future work. Secondly, two dimensional GaN nanostructures were prepared on a pristine silicon surface also at low temperature (T = 200°C). Following experiments were focused on a study of a crystal structure and an elemental analysis as these structures have been observed for the first time. Two dimensional structures are promising candidates into the high power applications which are emerging in these days. Thus, preparation of such 2D GaN nanostructures serves as a solid foundation for the further research. Thirdly, one dimensional GaN horizontal nanowires were fabricated on different sapphire planes. The prepared nanowires provided adequate dataset for the subsequent data analysis related to the growth kinetics. Collected dataset was used for verification of the developed theoretical model of the nanowire growth. It has been shown that the theoretical model describes the growth of nanowires with great precision and, thus, provide a useful insight into the growth mechanisms.
|
host ::
RSS.