López González, PaulaArjona, RosarioBaturone Castillo, María Iluminada2025-02-212025-02-212024-12-11López González, P., Arjona, R. y Baturone Castillo, M.I. (2024). Exploring Vein Biometrics on Ordinary Smartphones Using CNNs and Transfer Learning with Open and Closed Sets. En International Conference of the Biometrics Special Interest Group (BIOSIG) (1-5), Darmstadt, Germany: Institute of Electrical and Electronics Engineers.979-8-3503-7371-41617-5468https://hdl.handle.net/11441/169205This paper explores the feasibility of using vein recognition for biometric authentication on ordinary smartphones. A new dataset, USE-V2, consisting of 6600 vein images from dorsal hands and wrists, was collected under three different ambient conditions. Instead of designing a Convolutional Neural Network (CNN) from scratch, we used transfer learning with a pre-trained FaceNet model for extracting vein features and evaluated its performance in both open-set and closed-set scenarios with USE-V2 dataset. The recognition performance is acceptable when an open-set scenario with selection of ambient condition is considered (average EER of 5.18% for dorsal hands and 5.88% for wrists), and quite competitive for a closed-set scenario (average EER of 0.6% for dorsal hands and 0.75% for wrists). This approach paves the way for an efficient multimodal system integrating facial and vein recognition on smartphones, sharing most of the CNN layers.application/pdf6 p.engAttribution-NonCommercial-NoDerivatives 4.0 Internationalhttp://creativecommons.org/licenses/by-nc-nd/4.0/Vein RecognitionVascular BiometricsTransfer LearningCNNFaceNetSmartphoneinfo:eu-repo/semantics/conferenceObjectinfo:eu-repo/semantics/embargoedAccesshttps://doi.org/10.1109/biosig61931.2024.10786727