Fusion loss and inter-class data augmentation for deep finger vein feature learning
作者：
Highlights：
• Classification loss and metric learning loss have different discrimination ability.
• Fusion loss improves the generalization of learned features.
• Inter-class data augmentation enhances diversity with new finger vein classes.
• Intra-class and inter-class data augmentation resolve data shortage problem.
• Real-time verification system shows high efficiency and reliability.
摘要
•Classification loss and metric learning loss have different discrimination ability.•Fusion loss improves the generalization of learned features.•Inter-class data augmentation enhances diversity with new finger vein classes.•Intra-class and inter-class data augmentation resolve data shortage problem.•Real-time verification system shows high efficiency and reliability.
论文关键词：Finger vein recognition,Deep learning,Fusion loss,Data augmentation,Open-set
论文评审过程：Received 29 August 2020, Revised 4 December 2020, Accepted 6 January 2021, Available online 12 January 2021, Version of Record 5 February 2021.
论文官网地址：https://doi.org/10.1016/j.eswa.2021.114584
