Institutional Repository of Vilnius University
Veidų atpažinimas. Algoritmų palyginimas per veidų panašumo atpažinimą /
Abstract
dc:descriptionEven though a lot of algorithms are developed for facial detection and recognition, but it’s complicated to evaluate objectively which algorithms are better than the others. This research introduces simple 7 stages process, which enables algorithm comparison and works both for traditional and CNN algorithms. Initial data taken from Google Search engine are processed using MTCNN facial detection. As a result, some invalid images are discarded, as search engine is not totally reliable. Selected images are reviewed manually to prevent CNN poisoning and 160x160 sized cutouts are collected for further learning of FaceNet and Eigenfaces algorithms. Following the training 2000 pairs of same person faces and 2000 pairs of different person faces are randomly selected and compared using FaceNet and Eigenfaces algorithms. The result is evaluated statistically via confusion matrix on different thresholds to show practical possibility to use this method for statistical evaluations. It was noticed in this research that FaceNet is considerably slower and is trained slower than Eigenfaces, but it’s considerably more precise.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2024
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Murinaitė, Geda,
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:ELABAETD210646734&prefLang=en_US