BR-NPA: A non-parametric high-resolution attention model to improve the interpretability of attention

作者：

Highlights：

• We propose a novel interpretable spatial attention method called BR-NPA.

• Its attention module is a non-parametric layer that separates the parts of the object.

• We increase resolution which yields detailed attention maps, focused on the object.

• BR-NPA improves attention maps’ interpretability compared to previous SOTA methods.

摘要

•We propose a novel interpretable spatial attention method called BR-NPA.•Its attention module is a non-parametric layer that separates the parts of the object.•We increase resolution which yields detailed attention maps, focused on the object.•BR-NPA improves attention maps’ interpretability compared to previous SOTA methods.

论文关键词：Deep learning,Interpretability,Spatial attention,Resolution,Non-parametric

论文评审过程：Received 21 December 2021, Revised 28 June 2022, Accepted 21 July 2022, Available online 29 July 2022, Version of Record 2 August 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2022.108927