Institutional Repository of Vilnius University
Generaciniu priešiniu neuroniniu tinklu sugeneruotų paveiksliukų tikroviškumo vertinimas /
Abstract
dc:descriptionThis paper analyzes possibilities of evaluating realism in images, generated by generative adversarial networks. Firstly, some of many GAN models are described. Then GAN evaluation metrics and their suitability to evaluate images are analyzed. After that, criteria for image evaluation metrics are established and three metrics, that fits the criteria, are selected. After evaluating selected metrics, it is decided, that they are not enough to evaluate realism and new metric is suggested – intended to evaluate generated images and not GAN models. Experiment is conducted and it indicates, that suggested metric is valid and also shows strong and weak sides of this metric.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2020
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Žilionis, Martynas,
- Contributors dc:contributor
-
- Valaitis, Vytautas
Rights
dc:rights- Statement dc:rights
-
- info:eu-repo/semantics/openAccess
- Language dc:language
- lit
Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.vu.lt/VU:ELABAETD81704685&prefLang=en_US