Towards open-set touchless palmprint recognition via weight-based meta metric learning
作者：
Highlights：
• Weight-based Meta Metric Learning (W2ML) is proposed for open-set palmprint recognition.
• Hard sample mining and weighting are adopted to improve the efficiency.
• Extensive experiments are conducted on four constrained and unconstrained benchmarks.
• By W2ML method, the identification accuracy is increased by up to 9.11%.
• By W2ML method, Equal Error Rates (EER) is reduced by up to 2.97%.
摘要
•Weight-based Meta Metric Learning (W2ML) is proposed for open-set palmprint recognition.•Hard sample mining and weighting are adopted to improve the efficiency.•Extensive experiments are conducted on four constrained and unconstrained benchmarks.•By W2ML method, the identification accuracy is increased by up to 9.11%.•By W2ML method, Equal Error Rates (EER) is reduced by up to 2.97%.
论文关键词：Biometrics,Palmprint recognition,Meta learning,Metric learning
论文评审过程：Received 29 September 2020, Revised 11 July 2021, Accepted 11 August 2021, Available online 12 August 2021, Version of Record 20 August 2021.
论文官网地址：https://doi.org/10.1016/j.patcog.2021.108247
