Self-restrained triplet loss for accurate masked face recognition

作者：

Highlights：

• A solution to improve masked face verification performance.

• A novel loss function, the self-restrained triplet loss.

• Detailed evaluation on two real masked face datasets and two synthetically generated masked face datasets.

• Demonstration of the proposed solution on top of three face recognition models.

摘要

•A solution to improve masked face verification performance.•A novel loss function, the self-restrained triplet loss.•Detailed evaluation on two real masked face datasets and two synthetically generated masked face datasets.•Demonstration of the proposed solution on top of three face recognition models.

论文关键词：COVID-19,Biometric recognition,Identity verification,Masked face recognition

论文评审过程：Received 15 February 2021, Revised 21 October 2021, Accepted 29 November 2021, Available online 1 December 2021, Version of Record 11 December 2021.

论文官网地址：https://doi.org/10.1016/j.patcog.2021.108473