Institutional Repository of Vilnius University
Neuroninių tinklų naudojimas plaučių ligų analizei /
Abstract
dc:descriptionThis work analyses different architectural styles of convolutional neural networks that can be used to tackle the problem of lung nodule classification. The possibility of using these networks of different architectures together is also being examined. The theoretical part reviews 3 different networks that were proposed in scientific articles - modified VGG16-CNN network, Mask-R-CNN network and MGI-CNN network. The possibility of using different network architectures together is only possible when the networks have a similar structure - they need to be able to take the same input and produce classification output results in such format that could be used together with outputs of other neural networks. Based on the neural network models and ideas reviewed in the theoretical part, 3 experimental network models of different architectures were developed in the experimental part of the work, which are able to solve the classification task together based on majority and weighted consensus. This work compares the performance characteristics and experimental results of these networks, as well as examines the assumption whether consensus-based classification of images gives better results than the classification results of individual neural networks.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2020
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Bakštys, Vidmantas,
- Contributors dc:contributor
-
- Plankis, Tomas
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:ELABAETD81707172&prefLang=en_US