Institutional Repository of Vilnius University
Duomenų papildymo efektyvumas apmokant iš plaučių echoskopijų Covid-19 atpažįstančius neuroninius tinklus /
Abstract
dc:descriptionUnfortunately, there are too few lung ultrasound videos or pictures of Covid-19 patients to be able to fully train a highly accurate neural network that would also be big and complicated enough for diagnosis. That is why data augmentation is used for getting more data. Usually, simple geometric augmentations are used for lung ultrasound datasets, but it is uncertain which techniques are the most suitable and how each of them impact the model's performance. Therefore, the goal of this work was to find the best data augmentation techniques for lung ultrasound pictures. By comparing multiple kinds of data augmentation methods it was aimed to find the combination with the best model accuracy on test data. The most accurate model, reaching 81,5% accuracy, was trained with RICAP and multiple geometrical augmentation techniques applied during training.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2022
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Fetingytė, Elina,
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:ELABAETD146241202&prefLang=en_US