Institutional Repository of Vilnius University
Nuotraukų panašumu grįstas vaizdo žaidimų paieškos filtras naudojant konvoliucinius neuroninius tinklus /
Abstract
dc:descriptionThis paper examined the implementation of video game search filters and recommendation systems in digital video game distribution platforms and their shortcomings. Similar problems related to video game photos that have been solved so far were also analyzed, the methods used to solve the problem were examined, and what limitations were encountered. In order to create a dataset, different ways to extract video games were examined and data was prepared for further model training. In order to analyze the most suitable proposed solution for finding similar video games based on photos, different convolutional neural network architectures were trained and various methods such as transfer learning, fine-tuning, or contrast learning were applied. The trained models were evaluated using similarity measurement methods both on a single photo sample and on a video game album. The best model for this problem was selected - EfficientNetV2S, which achieved 77.19% recall and presented video games with similar visual art styles. The shortcomings of this task were analyzed, such as lack of data, class imbalance, and difficulties in collecting data.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Nenartavičiūtė, Laura,
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:ELABAETD308338217&prefLang=en_US