Deep embedding learning with adaptive large margin N-pair loss for image retrieval and clustering
作者：
Highlights：
• Learn a discriminative deep embedding via adaptive large margin con- straint.
• Introduce large angular margin optimization concept instead of exploring hard example mining strategy.
• Expanding easy training samples to hard boundary samples via Virtual Point Generating method.
• The exerted margin constraint is local-adaptive and befitting for fine-grained datasets.
• Experiments on several popular databases achieve competitive clustering and retrieval results.
摘要
•Learn a discriminative deep embedding via adaptive large margin con- straint.•Introduce large angular margin optimization concept instead of exploring hard example mining strategy.•Expanding easy training samples to hard boundary samples via Virtual Point Generating method.•The exerted margin constraint is local-adaptive and befitting for fine-grained datasets.•Experiments on several popular databases achieve competitive clustering and retrieval results.
论文关键词：Embedding learning,Adaptive margin,Virtual point generating,Discriminative feature
论文评审过程：Received 18 September 2018, Revised 18 April 2019, Accepted 1 May 2019, Available online 2 May 2019, Version of Record 6 May 2019.
论文官网地址：https://doi.org/10.1016/j.patcog.2019.05.011
