magistrsko delo
Nejc Vesel (Author), Peter Peer (Mentor), Vitomir Štruc (Co-mentor), Blaž Meden (Co-mentor)

Abstract

Staranje obrazov je področje, ki se ukvarja z modeliranjem staranja osebe iz ene same referenčne slike. Želimo ustvariti generativni model, ki nam s pomočjo nevronskih mrež ustvari slike referenčne osebe pri različnih starostnih skupinah. Pri našem pristopu smo želeli cilj doseči z uporabo različnih generativnih arhitektur. Preizkusili smo nekaj uveljavljenih pristopov ter implementirali nekaj lastnih idej, ki se niso izkazale za najuspešnejše. Dobljeni končni rezultati so bili pod pričakovanji, vendar naloga naredi pregled nad preizkušenimi pristopi in njihovo implementacijo. Naloga predstavlja dobro podlago za nadaljnje raziskovanje na tem področju, saj naredi pregled nad uspešnimi in neuspešnimi pristopi ter težavami, ki se pojavljajo pri raziskovanju tega področja.

Keywords

stanje obrazov;variacijski avtoenkoder;generativne mreže;nevronske mreže;generativne nasprotniške mreže;nasprotniški avtoenkoder;

Data

Language: Slovenian
Year of publishing:
Typology: 2.09 - Master's Thesis
Organization: UL FRI - Faculty of Computer and Information Science
Publisher: [N. Vesel]
UDC: 004
COBISS: 18607449 Link will open in a new window
Views: 1480
Downloads: 663
Average score: 0 (0 votes)
Metadata: JSON JSON-RDF JSON-LD TURTLE N-TRIPLES XML RDFA MICRODATA DC-XML DC-RDF RDF

Other data

Secondary language: English
Secondary title: Face aging using deep generative neural networks
Secondary abstract: Face aging as a research topic is dealing with modelling human aging from a reference photo. We want a generative model that, using generative neural networks, generates images of a reference person at a different age. We implemented some existing approaches and developed some of our own, however, they didn't return results that we wished for. The final results were below expectations, however, the thesis makes a good overview over the implemented approaches and their implementation. The thesis creates a good foundation for further research. It gives a good overview over successful and non successful approaches and the difficulties that arise when doing research on this topic.
Secondary keywords: face aging;variational autoencoder;generative networks;neural networks;generative adverserial networks;adverserial autoencoders;
Type (COBISS): Master's thesis/paper
Study programme: 0
Embargo end date (OpenAIRE): 1970-01-01
Thesis comment: Univ. v Ljubljani, Fak. za matematiko in fiziko, Oddelek za matematiko, Računalništvo in matematika - 2. stopnja
Pages: XII, 80 str.
ID: 11008321