Institutional Repository of Vilnius University
Pusiau prižiūrimo mokymosi tyrimas klasifikavimo modeliuose atpažįstant upinių bebrų veiklos žymes ortofoto nuotraukose /
Abstract
dc:descriptionThis thesis explores the application of semi-supervised learning methods in the automatic classification of orthophotos to detect Eurasian beaver (Castor fiber) activity. A self-training strategy was implemented: models were initially trained on a small set of labeled data, then used to classify a larger set of unlabeled images, expanding the training dataset automatically. Two classification tasks were addressed. The first – binary classification – aimed to distinguish between images with and without beaver activity. Here, the best performance was achieved by EfficientNetV2M and ConvNeXtBase models, with semi-supervised learning significantly improving accuracy, especially under limited data conditions. The second task involved three-class classification: distinguishing dams, trails, and neutral images. This task used smaller models (EfficientNetV2B0, ConvNeXtTiny, ResNet50V2) due to the severely smaller amount in data. 72 photos split across 3 classes proved to be not enough even with augmentations implemented, however, all models showed potential to learn even with a small number of photos.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Mažeikaitė, Ieva,
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:ELABAETD239031634&prefLang=en_US