Institutional Repository of Vilnius University
Gestų kalbos vienetų atpažinimas iš video srauto /
Abstract
dc:descriptionIn this study the recognition of Sign language units from a video stream is being analyzed. A dataset of 22 sign units with 20 videos for each class and 3 units with 50 videos for each class was created. Classes which had 20 videos was augmented to make 50 videos for each one of them. Based on this data, the convolutional neural network using Inception v3 model was trained. Further, based on convolutional neural network results, recurrent neural network was trained. Finally, three different models was created which are able to detect 3, 10 and 25 Lithuanian Sign language classes. The best results was presented by 10 classes model (87,34%), but 25 signs model has a higher usability within 79,18% accuracy. To sum up, these results give a possibility to adapt Lithuanian Sign language models in a real world. The usability of this model can be increased by further studies and trainings.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2018
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Ambrazas, Pranciškus,
- Contributors dc:contributor
-
- Petkevičius, Linas
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:ELABAETD29548677&prefLang=en_US