Reviewers Krzysztof Siwek ( FoEE/ITEEMIS) Krzysztof Siwek, The Institute of the Theory of Electrical Engineering, Measurement and Information Systems ( FoEE/ITEEMIS) Faculty of Electrical Engineering ( FoEE)Ĭertifying unit Faculty of Electrical Engineering ( FoEE) Affiliation unit The Institute of the Theory of Electrical Engineering, Measurement and Information Systems ( FoEE/ITEEMIS) Study subject / specialization Automatyka i Robotyka Stosowana Language (pl) Polish Supervisor Krzysztof Siwek ( FoEE/ITEEMIS) Piotr Antoniak, Faculty of Electrical Engineering ( FoEE) Title in Polish Syjamskie sieci neuronowe w zadaniu weryfikacji tożsamości na podstawie zdjęć twarzy Prediction accuracy for data classes not included in the learning process. The proposed architecture achieved a 81.83% The given task and confirmed the Siamese Neural Network effectiveness in The results let draw conlusionsĪbout the significance of the parameters appropriate selection in relation to In the next part, theĬollected results are presented and analyzed. Rameters as well as the learning process parameters. Structure was tested, taking into account the variability of the network pa.
The prediction quality of the proposed Siamese Neural Network
The analytical part, the experiments conditions and initial assumptions wereįormulated. Ral Network fundamental structure and the idea of Siamese Neural NetworksĪnd the One-Shot Learning method are presented. The thesis intro-ĭuces the area of Artificial Neural Networks considering the learning processĪnd methods of improving generalization abilities. Matrix for the prediction quality assessment are explained. The diagnostic measures based on the confusion Theoretical part outlines the concept of machine learning and depicts the
ZADANIA NA ZBIORACH LICZBOWYCH VERIFICATION
The aim of this dissertation is the study of Siamese Neural NetworksĬapabilities in identity verification for a limited face images data set. Obstacle in biometric identification applications. The weakness of the classical approach to deep neural networksĪlso becomes apparent in lack of knowledge generalization on classes notīeing taken into consideration during training process making it a significiant Years despite the high accuracy in classification tasks are based on largeĭata sets. Siamese Neural Networks for identity verification based on face recognitionĪbstract Object identification on the basis of incomplete information about it isĪ human ability giving him an advantage over machine learning algorithms.ĭeep learning methods whose development has been intensified in recent
0 kommentar(er)
