Universal pooling – A new pooling method for convolutional neural networks
作者：
Highlights：
• Pooling is one of the key elements in a convolutional neural network.
• In this paper, we propose a new pooling method named universal pooling (UP).
• UP can actually be considered as a channel-wise local spatial attention module.
• UP includes previous simple poolings such as MP, AP, and SP as special cases.
• UP achieves better performance than the previous pooling methods.
摘要
•Pooling is one of the key elements in a convolutional neural network.•In this paper, we propose a new pooling method named universal pooling (UP).•UP can actually be considered as a channel-wise local spatial attention module.•UP includes previous simple poolings such as MP, AP, and SP as special cases.•UP achieves better performance than the previous pooling methods.
论文关键词：Attention methods,Convolutional neural network,Average pooling,Max pooling,Stride pooling,Universal pooling
论文评审过程：Received 3 November 2020, Revised 21 March 2021, Accepted 17 April 2021, Available online 23 April 2021, Version of Record 12 May 2021.
论文官网地址：https://doi.org/10.1016/j.eswa.2021.115084
