LGCN: Learnable Gabor Convolution Network for Human Gender Recognition in the Wild

Peng CHEN; Weijun LI; Linjun SUN; Xin NING; Lina YU; Liping ZHANG

doi:10.1587/transinf.2018EDL8239

Open Access
LGCN: Learnable Gabor Convolution Network for Human Gender Recognition in the Wild

Peng CHEN, Weijun LI, Linjun SUN, Xin NING, Lina YU, Liping ZHANG

Full Text Views

90

Cite this

Free PDF (1.4MB)

Summary :

Human gender recognition in the wild is a challenging task due to complex face variations, such as poses, lighting, occlusions, etc. In this letter, learnable Gabor convolutional network (LGCN), a new neural network computing framework for gender recognition was proposed. In LGCN, a learnable Gabor filter (LGF) is introduced and combined with the convolutional neural network (CNN). Specifically, the proposed framework is constructed by replacing some first layer convolutional kernels of a standard CNN with LGFs. Here, LGFs learn intrinsic parameters by using standard back propagation method, so that the values of those parameters are no longer fixed by experience as traditional methods, but can be modified by self-learning automatically. In addition, the performance of LGCN in gender recognition is further improved by applying a proposed feature combination strategy. The experimental results demonstrate that, compared to the standard CNNs with identical network architecture, our approach achieves better performance on three challenging public datasets without introducing any sacrifice in parameter size.

Publication: IEICE TRANSACTIONS on Information Vol.E102-D No.10 pp.2067-2071

Publication Date: 2019/10/01

Publicized: 2019/06/13

Online ISSN: 1745-1361

DOI: 10.1587/transinf.2018EDL8239

Type of Manuscript: LETTER

Category: Image Recognition, Computer Vision

Authors

Peng CHEN
  CAS,University of Chinese Academy of Sciences
Weijun LI
  CAS,University of Chinese Academy of Sciences
Linjun SUN
  CAS,University of Chinese Academy of Sciences
Xin NING
  CAS,University of Chinese Academy of Sciences
Lina YU
  CAS,University of Chinese Academy of Sciences
Liping ZHANG
  CAS,University of Chinese Academy of Sciences

Keyword

gender recognition, learnable Gabor convolutional neural network, learnable Gabor filter, back propagation

Cite this

Copy

Peng CHEN, Weijun LI, Linjun SUN, Xin NING, Lina YU, Liping ZHANG, "LGCN: Learnable Gabor Convolution Network for Human Gender Recognition in the Wild" in IEICE TRANSACTIONS on Information, vol. E102-D, no. 10, pp. 2067-2071, October 2019, doi: 10.1587/transinf.2018EDL8239.
Abstract: Human gender recognition in the wild is a challenging task due to complex face variations, such as poses, lighting, occlusions, etc. In this letter, learnable Gabor convolutional network (LGCN), a new neural network computing framework for gender recognition was proposed. In LGCN, a learnable Gabor filter (LGF) is introduced and combined with the convolutional neural network (CNN). Specifically, the proposed framework is constructed by replacing some first layer convolutional kernels of a standard CNN with LGFs. Here, LGFs learn intrinsic parameters by using standard back propagation method, so that the values of those parameters are no longer fixed by experience as traditional methods, but can be modified by self-learning automatically. In addition, the performance of LGCN in gender recognition is further improved by applying a proposed feature combination strategy. The experimental results demonstrate that, compared to the standard CNNs with identical network architecture, our approach achieves better performance on three challenging public datasets without introducing any sacrifice in parameter size.
URL: https://global.ieice.org/en_transactions/information/10.1587/transinf.2018EDL8239/_p

Copy

@ARTICLE{e102-d_10_2067,
author={Peng CHEN, Weijun LI, Linjun SUN, Xin NING, Lina YU, Liping ZHANG, },
journal={IEICE TRANSACTIONS on Information},
title={LGCN: Learnable Gabor Convolution Network for Human Gender Recognition in the Wild},
year={2019},
volume={E102-D},
number={10},
pages={2067-2071},
abstract={Human gender recognition in the wild is a challenging task due to complex face variations, such as poses, lighting, occlusions, etc. In this letter, learnable Gabor convolutional network (LGCN), a new neural network computing framework for gender recognition was proposed. In LGCN, a learnable Gabor filter (LGF) is introduced and combined with the convolutional neural network (CNN). Specifically, the proposed framework is constructed by replacing some first layer convolutional kernels of a standard CNN with LGFs. Here, LGFs learn intrinsic parameters by using standard back propagation method, so that the values of those parameters are no longer fixed by experience as traditional methods, but can be modified by self-learning automatically. In addition, the performance of LGCN in gender recognition is further improved by applying a proposed feature combination strategy. The experimental results demonstrate that, compared to the standard CNNs with identical network architecture, our approach achieves better performance on three challenging public datasets without introducing any sacrifice in parameter size.},
keywords={},
doi={10.1587/transinf.2018EDL8239},
ISSN={1745-1361},
month={October},}

Copy

TY - JOUR
TI - LGCN: Learnable Gabor Convolution Network for Human Gender Recognition in the Wild
T2 - IEICE TRANSACTIONS on Information
SP - 2067
EP - 2071
AU - Peng CHEN
AU - Weijun LI
AU - Linjun SUN
AU - Xin NING
AU - Lina YU
AU - Liping ZHANG
PY - 2019
DO - 10.1587/transinf.2018EDL8239
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E102-D
IS - 10
JA - IEICE TRANSACTIONS on Information
Y1 - October 2019
AB - Human gender recognition in the wild is a challenging task due to complex face variations, such as poses, lighting, occlusions, etc. In this letter, learnable Gabor convolutional network (LGCN), a new neural network computing framework for gender recognition was proposed. In LGCN, a learnable Gabor filter (LGF) is introduced and combined with the convolutional neural network (CNN). Specifically, the proposed framework is constructed by replacing some first layer convolutional kernels of a standard CNN with LGFs. Here, LGFs learn intrinsic parameters by using standard back propagation method, so that the values of those parameters are no longer fixed by experience as traditional methods, but can be modified by self-learning automatically. In addition, the performance of LGCN in gender recognition is further improved by applying a proposed feature combination strategy. The experimental results demonstrate that, compared to the standard CNNs with identical network architecture, our approach achieves better performance on three challenging public datasets without introducing any sacrifice in parameter size.
ER -