20.500.12556/DKUM-74834 Uporabniška programska oprema za napredno organiziranje in urejanje digitalnih slik na osnovi nevronskih mrež diplomsko delo User software for advanced organizing and editing of digital images based on neural networks V diplomskem delu smo razvili razširljivo uporabniško programsko opremo za organiziranje in urejanje digitalnih slik. Program se deli na tri dele, in sicer organizacijski, urejevalni ter napredni način. Organizacijski način dodatno omogoča razširljive dinamične bližnjice na tipkovnici. Poleg tega smo implementirali dve konvolucijski nevronski mreži. Prva je namenjena skaliranju in ostrenju slik, druga pa za ločevanje nočnih in dnevnih slik. Obe naučeni mreži sta vključeni v napredni način naše aplikacije kot vtičnika. Program smo implementirali v ogrodju .NET, s pomočjo WPF-ja ter programskega jezika C#. Nevronski mreži smo implementirali v jeziku Python in z uporabo aplikacijskega programskega vmesnika Keras. Našo aplikacijo in rezultate smo ovrednotili tako kvalitativno kot tudi kvantitativno na manjši testni množici slik. Dobljeni rezultati so bili primerljivi s sorodnimi rešitvami. Po opravljeni fazi testiranja smo našo programsko opremo posredovali v ocenitev še dvema profesionalnima fotografoma. Dobili smo pozitivno povratno informacijo. Na osnovi vseh rezultatov in povratnih informacij lahko trdimo, da naša programska oprema deluje učinkovito in ponuja primerno uporabniško izkušnjo. In the thesis we developed a scalable user software for organizing and editing of digital images. The program is separated into three parts, organizing, editing and advanced mode. Additionally, the organizing mode offers extensible and customizable dynamic key bindings. Moreover, we also implemented two convolutional neural networks. The first was aimed at scaling and sharpening images, and the second for separating night and day photos. Both trained neural networks are included in the advanced mode of our application as extensions. The program was implemented in in the .NET framework with the help of WPF and the programming language C#. The neural networks were implemented in Python and with the use of the Keras API. Our application, and the results were evaluated, both qualitatively and quantitatively on a smaller test set of images. The results were comparable with other related solutions. After the completed testing phase, we sent our software to two professional photographers for evaluation. We received positive feedback. Based on all the results and feedback we can say that our software works efficiently and offers a proper user experience. fotografije windows aplikacija programski jezik C# konvolucijske nevronske mreže photographs windows application programming language C# convolutional neural networks true true false [K. Žagar] Slovenski jezik Angleški jezik Diplomsko delo/naloga 2019-09-09 13:57:56 2019-11-23 07:54:17 2022-08-09 09:37:02 2020-01-11 03:12:20 2019 0 Maribor 0 Maribor IX, 49 str. 0000-00-00 NiDoloceno NiDoloceno NiDoloceno 0000-00-00 0000-00-00 0000-00-00 004.8:004.932.4(043.2) 22912534 URN:SI:UM:DK:TJBUTLB2 UN_Zagar_Kristjan_2019.pdf UN_Zagar_Kristjan_2019.pdf 1 E4E481E579B88EEEDCCEC51594738356 a3cd547fedf81ab67901413ce7e519100157cba1c599d97ba10225b6284cbc7d fde4e8a5-7c10-11eb-bb7a-00155d0001ca 20.500.12556/dkum/c4a859ea-3b7d-46d4-81ab-ba2a122e7364 https://dk.um.si/Dokument.php?lang=slv&id=138164 Fakulteta za elektrotehniko, računalništvo in informatiko 0 0 0