Stacking PCANet +: An Overly Simplified ConvNets Baseline for Face Recognition

Cheng Yaw Low; Andrew Beng Jin Teoh; Kar Ann Toh

doi:10.1109/LSP.2017.2749763

Stacking PCANet +: An Overly Simplified ConvNets Baseline for Face Recognition

Cheng Yaw Low, Andrew Beng Jin Teoh, Kar Ann Toh

Department of Electrical and Electronic Engineering

Research output: Contribution to journal › Article › peer-review

34 Citations (Scopus)

Abstract

The principal component analysis network (PCANet) is asserted as a parsimonious stacking-based convolutional neural networks (CNNs) instance for generic object recognition including face. However, to be regarded a CNN resemblance, PCANet lacks a nonlinearity in between two successive convolutional layers. The multilayer PCANet (by neglecting the nonlinearity pre-requisite) is also deemed far-fetched for the network depth beyond two, due to feature dimensionality explosion. We thus devise a PCANet alternative, dubbed PCANet+ in this letter, to untangle these constraints. To be more precise, conforming to the CNN essentials, PCANet+ conveys a mean-pooling unit manipulating each feature map. On top of that, we streamline the PCANet topology to permit a deep construction with an expanded PCA filter ensemble. We scrutinize the PCANet+ performance using face recognition technology and other two faces in the wild datasets, namely, labeled faces in the wild and YouTube faces. The experimental results reveal that the PCANet+ descriptor prevails over its predecessor and other stacking-based descriptors in face identification and verification, serving a baseline for ConvNets.

Original language	English
Article number	8027103
Pages (from-to)	1581-1585
Number of pages	5
Journal	IEEE Signal Processing Letters
Volume	24
Issue number	11
DOIs	https://doi.org/10.1109/LSP.2017.2749763
Publication status	Published - 2017 Nov

Bibliographical note

Funding Information:
Manuscript received July 13, 2017; revised August 23, 2017; accepted August 31, 2017. Date of publication September 7, 2017; date of current version September 21, 2017. This work was supported by the National Research Foundation of Korea funded by the Korea government (Ministry of Science, ICT and Future Planning) under Grant 2016R1A2B4011656. The associate editor coordinating the review of this manuscript and approving it for publication was Dr. Vishal Monga. (Corresponding author: Andrew Beng-Jin Teoh.) The authors are with the School of Electrical and Electronic Engineering, College of Engineering, Yonsei University, Seoul 120749, South Korea (e-mail: engyawlow@yonsei.a.kr; bjteoh@yonsei.a.kr; katoh@yonsei.a.kr).

Publisher Copyright:
© 1994-2012 IEEE.

All Science Journal Classification (ASJC) codes

Signal Processing
Electrical and Electronic Engineering
Applied Mathematics

Access to Document

10.1109/LSP.2017.2749763

Cite this

@article{c23ff668612f4bbeaef9980a47960c2b,

title = "Stacking PCANet +: An Overly Simplified ConvNets Baseline for Face Recognition",

abstract = "The principal component analysis network (PCANet) is asserted as a parsimonious stacking-based convolutional neural networks (CNNs) instance for generic object recognition including face. However, to be regarded a CNN resemblance, PCANet lacks a nonlinearity in between two successive convolutional layers. The multilayer PCANet (by neglecting the nonlinearity pre-requisite) is also deemed far-fetched for the network depth beyond two, due to feature dimensionality explosion. We thus devise a PCANet alternative, dubbed PCANet+ in this letter, to untangle these constraints. To be more precise, conforming to the CNN essentials, PCANet+ conveys a mean-pooling unit manipulating each feature map. On top of that, we streamline the PCANet topology to permit a deep construction with an expanded PCA filter ensemble. We scrutinize the PCANet+ performance using face recognition technology and other two faces in the wild datasets, namely, labeled faces in the wild and YouTube faces. The experimental results reveal that the PCANet+ descriptor prevails over its predecessor and other stacking-based descriptors in face identification and verification, serving a baseline for ConvNets.",

author = "Low, {Cheng Yaw} and Teoh, {Andrew Beng Jin} and Toh, {Kar Ann}",

note = "Funding Information: Manuscript received July 13, 2017; revised August 23, 2017; accepted August 31, 2017. Date of publication September 7, 2017; date of current version September 21, 2017. This work was supported by the National Research Foundation of Korea funded by the Korea government (Ministry of Science, ICT and Future Planning) under Grant 2016R1A2B4011656. The associate editor coordinating the review of this manuscript and approving it for publication was Dr. Vishal Monga. (Corresponding author: Andrew Beng-Jin Teoh.) The authors are with the School of Electrical and Electronic Engineering, College of Engineering, Yonsei University, Seoul 120749, South Korea (e-mail: engyawlow@yonsei.a.kr; bjteoh@yonsei.a.kr; katoh@yonsei.a.kr). Publisher Copyright: {\textcopyright} 1994-2012 IEEE.",

year = "2017",

month = nov,

doi = "10.1109/LSP.2017.2749763",

language = "English",

volume = "24",

pages = "1581--1585",

journal = "IEEE Signal Processing Letters",

issn = "1070-9908",

publisher = "Institute of Electrical and Electronics Engineers Inc.",

number = "11",

}

TY - JOUR

T1 - Stacking PCANet +

T2 - An Overly Simplified ConvNets Baseline for Face Recognition

AU - Low, Cheng Yaw

AU - Teoh, Andrew Beng Jin

AU - Toh, Kar Ann

N1 - Funding Information: Manuscript received July 13, 2017; revised August 23, 2017; accepted August 31, 2017. Date of publication September 7, 2017; date of current version September 21, 2017. This work was supported by the National Research Foundation of Korea funded by the Korea government (Ministry of Science, ICT and Future Planning) under Grant 2016R1A2B4011656. The associate editor coordinating the review of this manuscript and approving it for publication was Dr. Vishal Monga. (Corresponding author: Andrew Beng-Jin Teoh.) The authors are with the School of Electrical and Electronic Engineering, College of Engineering, Yonsei University, Seoul 120749, South Korea (e-mail: engyawlow@yonsei.a.kr; bjteoh@yonsei.a.kr; katoh@yonsei.a.kr). Publisher Copyright: © 1994-2012 IEEE.

PY - 2017/11

Y1 - 2017/11

N2 - The principal component analysis network (PCANet) is asserted as a parsimonious stacking-based convolutional neural networks (CNNs) instance for generic object recognition including face. However, to be regarded a CNN resemblance, PCANet lacks a nonlinearity in between two successive convolutional layers. The multilayer PCANet (by neglecting the nonlinearity pre-requisite) is also deemed far-fetched for the network depth beyond two, due to feature dimensionality explosion. We thus devise a PCANet alternative, dubbed PCANet+ in this letter, to untangle these constraints. To be more precise, conforming to the CNN essentials, PCANet+ conveys a mean-pooling unit manipulating each feature map. On top of that, we streamline the PCANet topology to permit a deep construction with an expanded PCA filter ensemble. We scrutinize the PCANet+ performance using face recognition technology and other two faces in the wild datasets, namely, labeled faces in the wild and YouTube faces. The experimental results reveal that the PCANet+ descriptor prevails over its predecessor and other stacking-based descriptors in face identification and verification, serving a baseline for ConvNets.

AB - The principal component analysis network (PCANet) is asserted as a parsimonious stacking-based convolutional neural networks (CNNs) instance for generic object recognition including face. However, to be regarded a CNN resemblance, PCANet lacks a nonlinearity in between two successive convolutional layers. The multilayer PCANet (by neglecting the nonlinearity pre-requisite) is also deemed far-fetched for the network depth beyond two, due to feature dimensionality explosion. We thus devise a PCANet alternative, dubbed PCANet+ in this letter, to untangle these constraints. To be more precise, conforming to the CNN essentials, PCANet+ conveys a mean-pooling unit manipulating each feature map. On top of that, we streamline the PCANet topology to permit a deep construction with an expanded PCA filter ensemble. We scrutinize the PCANet+ performance using face recognition technology and other two faces in the wild datasets, namely, labeled faces in the wild and YouTube faces. The experimental results reveal that the PCANet+ descriptor prevails over its predecessor and other stacking-based descriptors in face identification and verification, serving a baseline for ConvNets.

UR - http://www.scopus.com/inward/record.url?scp=85029160419&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85029160419&partnerID=8YFLogxK

U2 - 10.1109/LSP.2017.2749763

DO - 10.1109/LSP.2017.2749763

M3 - Article

AN - SCOPUS:85029160419

SN - 1070-9908

VL - 24

SP - 1581

EP - 1585

JO - IEEE Signal Processing Letters

JF - IEEE Signal Processing Letters

IS - 11

M1 - 8027103

ER -

Stacking PCANet +: An Overly Simplified ConvNets Baseline for Face Recognition

Abstract

Bibliographical note

All Science Journal Classification (ASJC) codes

Access to Document

Other files and links

Fingerprint

Cite this