Towards prior gap and representation gap for long-tailed recognition

作者：

Highlights：

• A unified theoretical framework for long-tailed recognition is established.

• Corresponding mitigation solutions for prior gap and representation gap are proposed.

• Theoretically analyzing the existing methods and the proposed methods in terms of the impact on two gaps.

• The proposed methods yield superior performance on five long-tailed benchmarks.

摘要

•A unified theoretical framework for long-tailed recognition is established.•Corresponding mitigation solutions for prior gap and representation gap are proposed.•Theoretically analyzing the existing methods and the proposed methods in terms of the impact on two gaps.•The proposed methods yield superior performance on five long-tailed benchmarks.

论文关键词：Long-tailed learning,Prior gap,Representation gap,Image recognition

论文评审过程：Received 15 December 2021, Revised 13 July 2022, Accepted 27 August 2022, Available online 29 August 2022, Version of Record 2 September 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2022.109012