An Empirical Study: ELM in Face Matching

Haidong Wang, Md. Fawwad Hussain, Himadri Mukherjee, Sk Md Obaidullah, Ravindra S. Hegadi, Kaushik Roy, Kc Santosh

Abstract

Face data happens everywhere and face matching/verification is the must, such as it helps track criminals; unlock your mobile phone; and pay your bill without credit cards (e.g. Apple Pay). More often, in real world, grayscale image data are used since the color images require more storage. Gray level faces can be studied through two different features: edges and texture since spatial properties could be preserved. Such features could be used to classify faces from one another. Instead of using distance-based feature matching concept, in this paper, a fast machine learning classifier, which we call extreme learning machine (ELM) is used, where we have taken several different activation functions, such as tanh, sigmoid, softlim, hardlim, gaussian, multiquadric and inv_multiquadric. In our tests, five different publicly available datasets, such as Caltech, AR, ColorFERET, IndianFaces and ORL are used. For all activation functions, we have tested with and without feature selection techniques, and compared with the state-of-the-art results.
BibTeX
@inproceedings{wang2018empirical,
  author = {Haidong Wang and Md. Fawwad Hussain and Himadri Mukherjee and Sk Md Obaidullah and Ravindra S. Hegadi and Kaushik Roy and KC Santosh},
  title = {{An Empirical Study: ELM in Face Matching}},
  booktitle = {RTIP2R},
  year = {2018},
  doi = {10.1007/978-981-13-9181-1\_25},
}

Loading related content...

Figure