Unit:
Department of Informatics and TelecommunicationsΠληροφορική
Author:
KANELLIS PANTELEIMON
Supervisors info:
Θεοχάρης Θεοχάρης, Καθηγητής , Τμήμα Πληροφορικής και Τηλεπικοινωνιών, ΕΚΠΑ
Αντώνιος Δανελάκης, Δρ, Τμήμα Πληροφορικής και Τηλεπικοινωνιών, ΕΚΠΑ
Original Title:
Assessing the URxD 3D Face Recognition Algorithm on Synthetic Facial Data
Translated title:
Assessing the URxD 3D Face Recognition Algorithm on Synthetic Facial Data
Summary:
Real-time face recognition on real faces has been an extensive research topic in computer science in the last three decades. One of the most successful face recognition techniques, that has dominated performance evaluations for almost 15 years, is URxD. This thesis examines the performance of URxD in computer generated faces, and compares the results with real human faces. More specifically, a 3D database containing 100 synthetic faces is generated. Each face has a neutral and a random expression with random intensity. The source code of URxD is recompiled and edited in order to run in modern machines. Later on, the synthetic 3D database is used as input for URxD in order to compute the success rate, that is how many faces were recognised correctly. The result of URxD is considered successful for one face, if the program correlates the neutral expression of the face with the random expression of the same face. Finally, URxD was tested on 100 real faces, solely for comparison purposes.
Main subject category:
Technology - Computer science
Keywords:
face recognition, synthetic faces, 3D database, facial expressions, face recognition algorithm, recognition rate
