articleOct 19, 2017GREEN OA

NormFace

FWFeng WangXXXiang XiangJCJian ChengALAlan Loddon Yuille

University of Electronic Science and Technology of China · Johns Hopkins University

Indexed inarxivcrossref

Abstract

Thanks to the recent developments of Convolutional Neural Networks, the performance of face verification methods has increased rapidly. In a typical face verification method, feature normalization is a critical step for boosting performance. This motivates us to introduce and study the effect of normalization during training. But we find this is non-trivial, despite normalization being differentiable. We identify and study four issues related to normalization through mathematical analysis, which yields understanding and helps with parameter settings. Based on this analysis we propose two strategies for training using normalized features. The first is a modification of softmax loss, which optimizes cosine…

Citation impact

562
total citations
FWCI
15.62
Percentile
100%
References
13
Citations per year

Authors

4
  • FW
    Feng WangCorresponding

    University of Electronic Science and Technology of China

  • XX
    Xiang Xiang

    Johns Hopkins University

  • JC
    Jian Cheng

    University of Electronic Science and Technology of China

  • AL
    Alan Loddon Yuille

    Johns Hopkins University

Topics & keywords

Topics
Keywords
  • Normalization (sociology)
  • Softmax function
  • Boosting (machine learning)
  • Pattern recognition (psychology)
  • Cosine similarity
  • Convolutional neural network
No related works found for this paper.

Funding