Institutional Repository of Vilnius University
Veido išraiškų atpažinimas naudojant gilujį mokymąsi ir Principinių Komponentų analizę /
Abstract
dc:descriptionThis work examines the ability to classify facial expressions by using three methods: Convolutional neural networks (CNN), Support Vector Machines (SVM) and XGBoost. For training, FER-2013 dataset is used. This experiment tries to answer whether or not Principal Component Analysis is a valid technique while building a facial expression recognition model. PCA initialized data were used on an SVM algorithm in order to create the baseline for this experiment. CNN proved to have the highest precision when guessing the facial expression. which is 65%. XGBoost and SVM reached the rates of 50% and 46% respectively. The testing data of FER-2013 dataset were not the only data that the precision is evaluated upon. The models were also evaluated using manually collected data from the internet. The data was cropped and resized using the OpenCV library. In this case, the CNN model classified the data with 78% precision. Meanwhile, XGBoost and SVM did the same thing with 41% and 52% accuracy.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2021
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Bagdonas, Mantas,
- 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:ELABAETD107116523&prefLang=en_US