Institutional Repository of Vilnius University
Konvoliucinių neuroninių tinklų ir transformerių palyginimas klasifikuojant palydovinius vaizdus /
Abstract
dc:descriptionAdvances in modern technology, especially in transformer neural networks, have opened up new possibilities for solving image classification tasks and challenged traditional - convolution neurons networks. Satellite photo analysis requires highly accurate classification results, as minor errors also affect different places: climate monitoring, agricultural and urban planning, security. For these reasons, a comparison has been made between transformers and convolutional networks to determine which of them ensures greater accuracy and efficiency in the analysis of satellite images. This study sought to analyse the advantages and disadvantages of both networks. The results showed that VIT transformers are more resistant to hyperparameter change and more stable with larger resolution photos when CNN. However, convolution neural networks have shown their merit - they have been more stable for reducing data and when pictures are smaller resolutions.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Popkovaitė, Diana,
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:ELABAETD210647815&prefLang=en_US