HAM: Hybrid attention module in deep convolutional neural networks for image classification

作者：

Highlights：

• Proposing an attention module: Hybrid Attention Module (HAM).

• HAM can be embedded into any state-of-the-art CNN architectures.

• HAM improve networks performance without significantly increasing parameters.

• Compared with other state-of-the-art attention modules, HAM achieve better performance on the standard datasets.

• On STL-10 datasets, HAM can further reduce the negative impact of less data on the performance as networks go deeper.

摘要

•Proposing an attention module: Hybrid Attention Module (HAM).•HAM can be embedded into any state-of-the-art CNN architectures.•HAM improve networks performance without significantly increasing parameters.•Compared with other state-of-the-art attention modules, HAM achieve better performance on the standard datasets.•On STL-10 datasets, HAM can further reduce the negative impact of less data on the performance as networks go deeper.

论文关键词：Hybrid attention module,Channel attention map,Spatial feature descriptor,HAM-integrated networks

论文评审过程：Received 17 August 2021, Revised 13 March 2022, Accepted 7 May 2022, Available online 10 May 2022, Version of Record 13 May 2022.

论文官网地址：https://doi.org/10.1016/j.patcog.2022.108785