Virtual sample-based deep metric learning using discriminant analysis

作者：

Highlights：

• A novel method to define sample similarity for deep metric learning that is based on linear discriminant analysis for analyzing characteristic of embedding space.

• Applying to existing deep metric learning scheme and proving better similarity than previous works.

• Demonstrating outstanding performance via tests on two different environments of general and few-shot retrieval settings.

摘要

•A novel method to define sample similarity for deep metric learning that is based on linear discriminant analysis for analyzing characteristic of embedding space.•Applying to existing deep metric learning scheme and proving better similarity than previous works.•Demonstrating outstanding performance via tests on two different environments of general and few-shot retrieval settings.

论文关键词：Linear discriminant analysis,Deep metric learning,Retrieval task

论文评审过程：Received 19 March 2020, Revised 23 July 2020, Accepted 6 September 2020, Available online 10 September 2020, Version of Record 14 September 2020.

论文官网地址：https://doi.org/10.1016/j.patcog.2020.107643