Learning Pose-Aware Models for Pose-Invariant Face Recognition in the Wild

被引:57
|
作者
Masi, Iacopo [1 ]
Chang, Feng-Ju [1 ]
Choi, Jongmoo [1 ]
Harel, Shai [3 ]
Kim, Jungyeon [1 ]
Kim, Kanggeon [1 ]
Leksut, Jatuporn [1 ]
Rawls, Stephen [2 ]
Wu, Yue [2 ]
Hassner, Tal [2 ,3 ]
AbdAlmageed, Wael [2 ]
Medioni, Gerard [1 ]
Morency, Louis-Philippe [4 ]
Natarajan, Prem [2 ]
Nevatia, Ram [1 ]
机构
[1] Univ Southern Calif, Inst Robot & Intelligent Syst, Los Angeles, CA 90007 USA
[2] Univ Southern Calif, Inst Informat Sci, Los Angeles, CA 90007 USA
[3] Open Univ Israel, IL-4353701 Raanana, Israel
[4] Carnegie Mellon Univ, Sch Comp Sci, Pittsburgh, PA 15213 USA
关键词
Face recognition; CNN; pose-aware;
D O I
10.1109/TPAMI.2018.2792452
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
We propose a method designed to push the frontiers of unconstrained face recognition in the wild with an emphasis on extreme out-of-plane pose variations. Existing methods either expect a single model to learn pose invariance by training on massive amounts of data or else normalize images by aligning faces to a single frontal pose. Contrary to these, our method is designed to explicitly tackle pose variations. Our proposed Pose-Aware Models (PAM) process a face image using several pose-specific, deep convolutional neural networks (CNN). 3D rendering is used to synthesize multiple face poses from input images to both train these models and to provide additional robustness to pose variations at test time. Our paper presents an extensive analysis of the IARPA Janus Benchmark A (IJB-A), evaluating the effects that landmark detection accuracy, CNN layer selection, and pose model selection all have on the performance of the recognition pipeline. It further provides comparative evaluations on IJB-A and the PIPA dataset. These tests show that our approach outperforms existing methods, even surprisingly matching the accuracy of methods that were specifically fine-tuned to the target dataset. Parts of this work previously appeared in [1] and [2].
引用
收藏
页码:379 / 393
页数:15
相关论文
共 50 条
  • [1] Pose-Aware Face Recognition in the Wild
    Masi, Iacopo
    Rawls, Stephen
    Medioni, Gerard
    Natarajan, Prem
    [J]. 2016 IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR), 2016, : 4838 - 4846
  • [2] Practical Pose Normalizaiton for Pose-Invariant Face Recognition
    Wu, Zhongjun
    Li, Shan
    Deng, Weihong
    [J]. PROCEEDINGS 3RD IAPR ASIAN CONFERENCE ON PATTERN RECOGNITION ACPR 2015, 2015, : 116 - 120
  • [3] APA: Adaptive Pose Alignment for Pose-Invariant Face Recognition
    An, Zhanfu
    Deng, Weihong
    Hu, Jiani
    Zhong, Yaoyao
    Zhao, Yuying
    [J]. IEEE ACCESS, 2019, 7 : 14653 - 14670
  • [4] Pose-invariant features and personalized correspondence learning for face recognition
    Yongbin Gao
    Hyo Jong Lee
    [J]. Neural Computing and Applications, 2019, 31 : 607 - 616
  • [5] Pose-invariant features and personalized correspondence learning for face recognition
    Gao, Yongbin
    Lee, Hyo Jong
    [J]. NEURAL COMPUTING & APPLICATIONS, 2019, 31 (Suppl 1): : 607 - 616
  • [6] Disentangled Representation Learning GAN for Pose-Invariant Face Recognition
    Tran, Luan
    Yin, Xi
    Liu, Xiaoming
    [J]. 30TH IEEE CONFERENCE ON COMPUTER VISION AND PATTERN RECOGNITION (CVPR 2017), 2017, : 1283 - 1292
  • [7] Unsupervised face Frontalization for pose-invariant face recognition
    Liu, Yanfei
    Chen, Junhua
    [J]. IMAGE AND VISION COMPUTING, 2021, 106
  • [8] Biological Inspired Pose-Invariant Face Recognition
    Wi, Noel Tay Nuo
    Kiong, Loo Chu
    [J]. 2012 WORLD AUTOMATION CONGRESS (WAC), 2012,
  • [9] A Comprehensive Survey on Pose-Invariant Face Recognition
    Ding, Changxing
    Tao, Dacheng
    [J]. ACM TRANSACTIONS ON INTELLIGENT SYSTEMS AND TECHNOLOGY, 2016, 7 (03)
  • [10] Pose-invariant face recognition using AAMs
    Corcoran, P.
    Ionita, M.
    Costache, G.
    [J]. PROCEEDINGS OF THE 10TH INTERNATIONAL CONFERENCE ON OPTIMIZATION OF ELECTRICAL AND ELECTRONIC EQUIPMENT, VOL IV, 2006, : 181 - 184