Avcı, AdemKocakulak, MustafaAcır, Nurettin2021-03-202021-03-202019https://hdl.handle.net/20.500.12885/69011th International Conference on Electrical and Electronics Engineering (ELECO) -- NOV 28-30, 2019 -- Bursa, TURKEYWith the increase in the number of publicly available finger-vein datasets, most of the recent studies on finger-vein biometrics have started to use Convolutional Neural Networks (CNNs). Since it is not an easy task to create a biometric dataset with a large number of users due to several privacy reasons, this study uses 4 publicly available finger-vein datasets. However, these datasets have a limited number of samples per user. From this point of view, in order to avoid possible overfitting problems that occur due to limited training samples, this study provides 4 empirical convolutional neural network designs without using any preprocessing operation for each biometric dataset after systematic comparisons.eninfo:eu-repo/semantics/closedAccess[No Keywords]Conference Object580584WOS:000552654100114N/AN/A