Generativna suparnička mreža za izradu umjetničkih slika

master's thesis

• Title (english): A generative adversarial network for artistic images creation
• Author: Marko Mlakić
• Mentor: Tomislav Stipančić (mentor)
• Granter: University of Zagreb; Faculty of Mechanical Engineering and Naval Architecture; Zagreb
• Defense date and country: 2023-12-07, Croatia
• Scientific / art field, discipline and subdiscipline: TECHNICAL SCIENCES; Mechanical Engineering

Abstract

Rad se bavi razvojem računalnog modela zasnovanog na metodologiji generativnih suparničkih mreža (poznato i kao GAN) za stvaranje umjetničkih slika. Model je treniran na određenom skupu podataka te je uspješan u generiranju slika stilova različitih umjetnika. U radu su detaljno opisani tehnički aspekti modela, uključujući arhitektura generatora i diskriminatora. Također je opisana strategija učenja GAN-a te izazovi (kao i rješenja) koji su se događali tijekom procesa. Dodatno je model testiran s pomoću maksimalne prosječne razlike, koja je služila kao evaluacijska metrika radi provjere učinkovitosti i preciznosti generativnog modela. Dio projekta uključuje i razvoj „web“ stranice koja služi kao platforma za interakciju s modelom. „Web“ stranica sadrži korisničko sučelje za učitavanje slika i pružanje povratnih informacija o generiranim slikama.

Abstract (english)

The paper focuses on developing a computational model based on the methodology of generative adversarial networks (also known as GAN) for creating artistic paintings. The model was trained on a specific dataset and is successful in generating images in the style of various artists. The paper details the technical aspects of the model, including the architecture of the generator and discriminator. It also describes the GAN learning strategy for the assigned task, as well as the challenges and solutions encountered during the process. Additionally, the model was tested using the maximum mean discrepancy, which served as an evaluation metric to check the effectiveness and precision of the generative model. Part of the project also includes the development of a website that serves as a platform for interacting with the model. The website contains a user interface for uploading images and providing feedback on the generated images.

• Keywords: generativna suparnička mreža; generator; diskriminator; strojno učenje
• Keywords (english): generative adversarial network; generator; discriminator; machine learning
• Language: croatian
• URN:NBN: urn:nbn:hr:235:350086
• Study programme: Title: Mechanical Engineering; Study programme type: university; Study level: graduate; Academic / professional title: magistar/magistra inženjer/inženjerka strojarstva (magistar/magistra inženjer/inženjerka strojarstva)
• Type of resource: Text
• File origin: Born digital
• Access conditions: Open access
• Created on: 2023-11-30 15:21:26