TY - GEN
T1 - ConDense
T2 - 3rd International Conference on Pattern Recognition and Artificial Intelligence, ICPRAI 2022
AU - Lang, Dane
AU - van der Haar, Dustin
N1 - Publisher Copyright:
© 2022, Springer Nature Switzerland AG.
PY - 2022
Y1 - 2022
N2 - Fingerprint recognition is now a common, well known and generally accepted form of biometric authentication. The popularity of fingerprint recognition also makes it the focus of many studies which aim to constantly improve the technology in terms of factors such as accuracy and speed. This study sets out to create fingerprint recognition architectures which improve upon pre-trained architectures - named ConDense - that provide stronger if not comparable accuracy in comparison to related works on the authentication/identification task. Each of these ConDense architectures are tested against databases 1A, 2A, 3A provided by FVC 2006. The ConDense architectures presented in this study performed well across the varying image qualities in the given databases, with the lowest EERs achieved by this study’s architectures being 1.385% (DB1A), 0.041% (DB2A) and 0.871% (DB3A). In comparison to related works, the architectures presented in this study performed the best in terms of EER against DB1A, and DB3A. The lowest EER for DB2A reported by a related work was 0.00%.
AB - Fingerprint recognition is now a common, well known and generally accepted form of biometric authentication. The popularity of fingerprint recognition also makes it the focus of many studies which aim to constantly improve the technology in terms of factors such as accuracy and speed. This study sets out to create fingerprint recognition architectures which improve upon pre-trained architectures - named ConDense - that provide stronger if not comparable accuracy in comparison to related works on the authentication/identification task. Each of these ConDense architectures are tested against databases 1A, 2A, 3A provided by FVC 2006. The ConDense architectures presented in this study performed well across the varying image qualities in the given databases, with the lowest EERs achieved by this study’s architectures being 1.385% (DB1A), 0.041% (DB2A) and 0.871% (DB3A). In comparison to related works, the architectures presented in this study performed the best in terms of EER against DB1A, and DB3A. The lowest EER for DB2A reported by a related work was 0.00%.
KW - Convolutional Neural Networks
KW - Fingerprint recognition
KW - Transfer learning
UR - http://www.scopus.com/inward/record.url?scp=85132118725&partnerID=8YFLogxK
U2 - 10.1007/978-3-031-09282-4_2
DO - 10.1007/978-3-031-09282-4_2
M3 - Conference contribution
AN - SCOPUS:85132118725
SN - 9783031092817
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 15
EP - 27
BT - Pattern Recognition and Artificial Intelligence - 3rd International Conference, ICPRAI 2022, Proceedings
A2 - El Yacoubi, Mounîm
A2 - Granger, Eric
A2 - Yuen, Pong Chi
A2 - Pal, Umapada
A2 - Vincent, Nicole
PB - Springer Science and Business Media Deutschland GmbH
Y2 - 1 June 2022 through 3 June 2022
ER -