Razpoznava umetno ustvarjenih slik z metodami strojnega učenja

magistrsko delo

Jan Premzl (Author), Milan Zorman (Mentor)

Abstract

V zaključnem delu smo izdelali modularno aplikacijo, ki je namenjena napovedovanju, ali je slika realna ali umetno ustvarjena. Za ta namen smo naučili enajst različnih modelov, vsakega s petnajstimi različnimi kombinacijami hiperparemetrov. Na podlagi tega smo dobili rezultate, kjer smo izračunali razne statistične mere in korelacije med rezultati. Poleg servisa za klasifikacijo oz. za učenje modelov smo izdelali tudi servis za ustvarjanje učne množice po lastnih željah, spletno aplikacijo, ki omogoča napovedi, in aplikacijski programski vmesnik, ki služi za komunikacijo med servisom za razpoznavo in spletno aplikacijo.

Keywords

strojno učenje;nevronske mreže;klasifikacija;PyTorch;

Data

Language:	Slovenian
Year of publishing:	2024
Typology:	2.09 - Master's Thesis
Organization:	UM FERI - Faculty of Electrical Engineering and Computer Science
Publisher:	[J. Premzl]
UDC:	004.932:004:8(043.2)
COBISS:	217286659
Views:	56
Downloads:	16
Average score:	0 (0 votes)
Metadata:

Other data

Secondary language:	English
Secondary title:	Detection of artificially generated images using methods of machine learning
Secondary abstract:	We have created a modular application, which is used to predict if an image is real or artificially generated. For this we trained eleven different models, each with fifteen different combinations of hyperparameters. On this basis we got results, where we calculated various statistical measures and correlations between results. In addition to the classification service, we developed a service to create a custom image dataset, a web application, which can predict an image and an application programming interface, which serves as a communication interface between the classification service and the web application.
Secondary keywords:	machine learning;neural networks;classification;PyTorch;
Type (COBISS):	Master's thesis/paper
Thesis comment:	Univ. v Mariboru, Fak. za elektrotehniko, računalništvo in informatiko, Računalništvo in informacijske tehnologije
Pages:	1 spletni vir (1 datoteka PDF (XV, 86 f.))
ID:	24411463