Fusion Loss and Inter-Class Data Augmentation for Deep Finger Vein Feature Learning

Finger vein recognition (FVR) based on deep learning (DL) has gained rising attention in recent years. However, the performance of FVR is limited by the insufficient amount of finger vein training data and the weak generalization of learned features. To address these limitations and improve the performance, we propose a simple framework by jointly considering intensive data augmentation, loss function design and network architecture selection. Firstly, we propose a simple inter-class data augmentation technique that can double the number of finger vein training classes with new vein patterns via vertical flipping. Then, we combine it with conventional intra-class data augmentation methods to achieve highly diversified expansion, thereby effectively resolving the data shortage problem. In order to enhance the discrimination of deep features, we design a fusion loss by incorporating the classification loss and the metric learning loss. We find that the fusion of these two penalty signals will lead to a good trade-off between the intra-class similarity and inter-class separability, thereby greatly improving the generalization ability of learned features. We also investigate various network architectures for FVR application in terms of performances and model complexities. To examine the reliability and efficiency of our proposed framework, we implement a real-time FVR system to perform end-to-end verification in a near-realworld working condition. In challenging open-set evaluation protocol, extensive experiments conducted on three public finger vein databases and an in-house database confirm the effectiveness of the proposed method.