Institutional Repository of Vilnius University
Veidų aptikimo ir atpažinimo algoritmai /
Abstract
dc:descriptionFace detection and recognition are processes that are difficult for humans and computers to understand and require a lot of research to solve. From the first attempts to match with full face templates or individual facial features, to sophisticated mathematical methods that look at the face as a collection of pixels or vectors, great progress have been made, but the development of this field may still be immeasurable. This progress was driven by the first faster algorithms, and subsequent development gained significant momentum. The paper examines algorithms used for face detection and recognition in photos and in real time. The generalized structure of these methods and individual algorithmic ways of solving the problem are discussed. Principal and Independent Component Analysis are reviewed, these methods are briefly compared, and related face recognition algorithms are presented. The first real-time Viola-Jones algorithm is also examined and the main steps that determined the speed of the algorithm's calculations are described. Convolutional neural networks, their main component layers and differences from the Viola-Jones algorithm are also reviewed for the comparison of real-time algorithms. To consolidate the theory, two algorithms have been implemented in the "Python" programming language: the eigenfaces algorithm for face recognition and the Viola-Jones algorithm for face detection. The eigenfaces algorithm averages the face of the training data and centers the data from which the eigenfaces are obtained. The test data is projected into the resulting face space and the Euclidean distance is calculated, which indicates whether the face was recognized. The Viola-Jones algorithm computes weak and strong classifiers from extracted quadratic features. Strong classifiers mark the features where the face is detected.
Degree
thesis:*- Grantor dc:publisher
- Institutional Repository of Vilnius University
- Year dc:date
- 2023
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Višniauskaitė, Akvilė,
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:ELABAETD192845601&prefLang=en_US