Deep coupled ResNet for low-resolution face recognition
Face images captured by surveillance cameras are often of low resolution (LR), which adversely affects the performance of their matching with high-resolution (HR) gallery images. Existing methods including super resolution, coupled mappings (CMs), multidimensional scaling, and convolutional neural n...
Saved in:
Main Authors: | , , |
---|---|
Other Authors: | |
Format: | Article |
Language: | English |
Published: |
2020
|
Subjects: | |
Online Access: | https://hdl.handle.net/10356/142565 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Nanyang Technological University |
Language: | English |
id |
sg-ntu-dr.10356-142565 |
---|---|
record_format |
dspace |
spelling |
sg-ntu-dr.10356-1425652020-06-24T07:03:33Z Deep coupled ResNet for low-resolution face recognition Lu, Ze Jiang, Xudong Kot, Alex School of Electrical and Electronic Engineering Engineering::Electrical and electronic engineering Convolutional Neural Network (CNN) Coupled Mappings (CMs) Face images captured by surveillance cameras are often of low resolution (LR), which adversely affects the performance of their matching with high-resolution (HR) gallery images. Existing methods including super resolution, coupled mappings (CMs), multidimensional scaling, and convolutional neural network yield only modest performance. In this letter, we propose the deep coupled ResNet (DCR) model. It consists of one trunk network and two branch networks. The trunk network, trained by face images of three significantly different resolutions, is used to extract discriminative features robust to the resolution change. Two branch networks, trained by HR images and images of the targeted LR, work as resolution-specific CMs to transform HR and corresponding LR features to a space where their difference is minimized. Model parameters of branch networks are optimized using our proposed CM loss function, which considers not only the discriminability of HR and LR features, but also the similarity between them. In order to deal with various possible resolutions of probe images, we train multiple pairs of small branch networks while using the same trunk network. Thorough evaluation on LFW and SCface databases shows that the proposed DCR model achieves consistently and considerably better performance than the state of the arts. MOE (Min. of Education, S’pore) 2020-06-24T07:03:33Z 2020-06-24T07:03:33Z 2018 Journal Article Lu, Z., Jiang, X., & Kot, A. (2018). Deep coupled ResNet for low-resolution face recognition. IEEE Signal Processing Letters, 25(4), 526-530. doi:10.1109/LSP.2018.2810121 1070-9908 https://hdl.handle.net/10356/142565 10.1109/LSP.2018.2810121 2-s2.0-85042866380 4 25 526 530 en IEEE Signal Processing Letters © 2018 IEEE. All rights reserved. |
institution |
Nanyang Technological University |
building |
NTU Library |
country |
Singapore |
collection |
DR-NTU |
language |
English |
topic |
Engineering::Electrical and electronic engineering Convolutional Neural Network (CNN) Coupled Mappings (CMs) |
spellingShingle |
Engineering::Electrical and electronic engineering Convolutional Neural Network (CNN) Coupled Mappings (CMs) Lu, Ze Jiang, Xudong Kot, Alex Deep coupled ResNet for low-resolution face recognition |
description |
Face images captured by surveillance cameras are often of low resolution (LR), which adversely affects the performance of their matching with high-resolution (HR) gallery images. Existing methods including super resolution, coupled mappings (CMs), multidimensional scaling, and convolutional neural network yield only modest performance. In this letter, we propose the deep coupled ResNet (DCR) model. It consists of one trunk network and two branch networks. The trunk network, trained by face images of three significantly different resolutions, is used to extract discriminative features robust to the resolution change. Two branch networks, trained by HR images and images of the targeted LR, work as resolution-specific CMs to transform HR and corresponding LR features to a space where their difference is minimized. Model parameters of branch networks are optimized using our proposed CM loss function, which considers not only the discriminability of HR and LR features, but also the similarity between them. In order to deal with various possible resolutions of probe images, we train multiple pairs of small branch networks while using the same trunk network. Thorough evaluation on LFW and SCface databases shows that the proposed DCR model achieves consistently and considerably better performance than the state of the arts. |
author2 |
School of Electrical and Electronic Engineering |
author_facet |
School of Electrical and Electronic Engineering Lu, Ze Jiang, Xudong Kot, Alex |
format |
Article |
author |
Lu, Ze Jiang, Xudong Kot, Alex |
author_sort |
Lu, Ze |
title |
Deep coupled ResNet for low-resolution face recognition |
title_short |
Deep coupled ResNet for low-resolution face recognition |
title_full |
Deep coupled ResNet for low-resolution face recognition |
title_fullStr |
Deep coupled ResNet for low-resolution face recognition |
title_full_unstemmed |
Deep coupled ResNet for low-resolution face recognition |
title_sort |
deep coupled resnet for low-resolution face recognition |
publishDate |
2020 |
url |
https://hdl.handle.net/10356/142565 |
_version_ |
1681056927237275648 |