Institutional Repository of Vilnius University
Detalus vaizdų panašumas naudojant trejetų tinklus /
Abstract
dc:descriptionFine-grained Images Similarity using Triplet-network This work describes person re-identification by using Triplet-network model implementation. It was trained by using several datasets - CUHK01 and Market-1501. The final results were compared with other competitive image recognition networks in a market. In order to have correct comparison results some metrics were provided in this document. Triplet-network was achieved by customizing one of the most popular pre-trained convolutional neural networks - VGG16. Instead of using traditional loss function - triplet loss function was used. After lots of comparison of final produced results with other networks data results, the conclusion was reached that triplet networks can be considered as strongly competitive network because it can reach great identification accuracy after it is trained with datasets mentioned previously.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2020
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Butkevičius, Andrius,
- Contributors dc:contributor
-
- Valaitis, Vytautas
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:ELABAETD81705613&prefLang=en_US