Face image classification by pooling raw features

作者：

Highlights：

• We propose the first face feature extraction method by directly pooling raw patches.

• The method can capture strong structural information of individual faces.

• The method achieves state-of-the-art results on several face recognition datasets.

• Multi-level spatial pooling and dense extraction of local patches play critical roles.

摘要

Highlights•We propose the first face feature extraction method by directly pooling raw patches.•The method can capture strong structural information of individual faces.•The method achieves state-of-the-art results on several face recognition datasets.•Multi-level spatial pooling and dense extraction of local patches play critical roles.

论文关键词：Face recognition,Image classification,Feature pooling

论文评审过程：Received 23 December 2014, Revised 16 August 2015, Accepted 11 January 2016, Available online 22 January 2016, Version of Record 27 February 2016.

论文官网地址：https://doi.org/10.1016/j.patcog.2016.01.010