Institutional Repository of Vilnius University
Žymių pasaulio objektų atpažinimas naudojant vaizdų panašumo vertinimą /
Abstract
dc:descriptionImage similarity and recognition play a key role in computer vision applications. With the increasing amount of image data available in different domains, reliable methodologies for extracting meaningful features and determining image similarities are needed. The aim of this work is to develop image recognition models using deep learning techniques. In this work, models based on a triplet loss function and ResNet50V2 or ViT backbones were developed and evaluated against each other and against pretrained ResNet50V2 and ViT on a variety of metrics: recall, precision, accuracy, Recall@K, area under the recall-precision curve. Compared to the pretrained models, all models showed decline in results. The main conclusion is that the GLDv2 dataset is too complex for conventional convolutional neural network and transformer neural network models, and that additional deep learning methods are needed to exploit these models.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Štaupas, Edvinas,
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:ELABAETD210647452&prefLang=en_US