Institutional Repository of Vilnius University
Fotokameros tipo nustatymas naudojant gilųjį mokymąsi /
Abstract
dc:descriptionThe work consists of theoretical and practical parts. The theoretical part describes deep learning, also components, and working principle of artificial neural networks. Later, the structure and functioning of the convolutional neural networks used in the practical part are explained in detail. In addition, practically used parts are described: activation functions, optimization functions, loss functions, the pre-trained model, the role of hyperparameters in the training of the networks, and how the appropriate hyperparameters can be found. The paper then analyses the work related to the camera type identification problem and describes what the dataset provided by the ”Kaggle” challenge is composed of. In the practical part, each training model used during experiments structure and detailed description are given. In addition, it explains how the experiments are performed and summarise the results. At the end of the paper, the results are listed and 3 conclusions are drawn. The ”Inception-ResNet-v2” pre-trained model can be successfully applied to a problem requiring low-level feature recognition and the preparation of the dataset has a significant impact on the further model training. Through experiments, it is also shown that the classification accuracy of the model can be significantly improved without major changes to the model architecture.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2022
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Klimovič, Erika,
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:ELABAETD146241411&prefLang=en_US