Landmark perturbation-based data augmentation for unconstrained face recognition
详细信息 查看全文
- 作者：Jiang-Jing Lv ; lvjiangjing@cigit.ac.cn" class="auth_mail" title="E-mail the corresponding author ; Cheng Cheng chengcheng@cigit.ac.cn" class="auth_mail" title="E-mail the corresponding author ; Guo-Dong Tian tianguodong@cigit.ac.cn" class="auth_mail" title="E-mail the corresponding author ; Xiang-Dong Zhou zhouxiangdong@cigit.ac.cn" class="auth_mail" title="E-mail the corresponding author ; Xi Zhou zhouxi@cigit.ac.cn" class="auth_mail" title="E-mail the corresponding author
- 关键词：Feature representation ; Face recognition ; Landmark perturbation ; Misalignment ; Deep convolutional neural networks
- 刊名：Signal Processing: Image Communication
- 出版年：2016
- 出版时间：September 2016
- 年：2016
- 卷：47
- 期：Complete
- 页码：465-475
- 全文大小：2539 K
文摘
- •
Landmark perturbation-based data augmentation method is able to generate different kinds of artificial face images automatically.
- •
The trained DCNN model using landmark perturbation-based data augmentation method is robust to misalignment.
- •
The proposed data augmentation method improves face recognition rates, meanwhile it provides faster convergence of DCNN training at early stage.