Face alignment method based on scale self-adaption and incremental learning

doi:10.11772/j.issn.1001-9081.2017122928

Abstract

Abstract: To solve the problems of traditional regression-based methods such as the loss of texture caused by face scale normalization, costing more time for retraining expanded data set to improve the generalization ability of the original model, and even potential non-convergence and incomputability, a face alignment method based on scale self-adaption and Incremental Learning (IL) was proposed. Firstly, the mapping relationship between the initial face points and the standard face points was established. Secondly, the extraction of the texture features on the original image and the normalization of the face scales were achieved via the mapping relationship. At last, incremental learning was applied to the new data set by using the existing models, which improved the generalization ability of the original model quickly. The experimental results show that the proposed method performs higher alignment accuracy than traditional regression-based methods. On the AFW (Annotated Faces in the Wild) dataset (68 feature points), the accuracy is increased by 2 to 4 percentage points; and on a 100000-level large dataset (5 feature points), the robustness is increased by 1 to 2 percentage points compared to the methods based on Deep Learning (DL). In addition, the proposed method is not only suitable for face alignment regression model, but also applicable for solving other regression models.

Key words: face alignment, Incremental Learning (IL), scale self-adaption, cascaded regression, robustness

摘要： 针对传统基于回归的人脸对齐算法在人脸尺度归一化时会造成纹理的损失，以及为了提升算法模型的泛化能力必须扩充数据集重新训练而导致训练时间增加，甚至出现不收敛、不可计算等问题，提出一种基于尺度自适应与增量式学习（IL）的人脸对齐方法来提高定位精度。首先，建立初始人脸形状与标准人脸形状的映射关系；然后，通过映射关系实现纹理特征在原图上的提取和人脸尺度的归一化；最后，利用算法模型在新的数据集上进行增量式的学习，快速提高原模型的泛化能力。实验结果表明，与传统回归方法相比，所提方法有更高的对齐精度，特别是在AFW数据集（68个特征点）上提高了2~4个百分点；在10万级别的大数据集（5个特征点）上，所提方法的鲁棒性比基于深度学习的方法高1~2个百分点。同时，所提的增量式学习方法不仅适用于人脸对齐场景下的回归模型求解，还适用于其他应用场景下回归模型的求解。

关键词: 人脸对齐, 增量式学习, 尺度自适应, 级联回归, 鲁棒性

CLC Number:

TP391.413

CHEN Ping, GONG Xun. Face alignment method based on scale self-adaption and incremental learning[J]. Journal of Computer Applications, 2018, 38(7): 2064-2069.

陈平, 龚勋. 基于尺度自适应与增量式学习的人脸对齐方法[J]. 计算机应用, 2018, 38(7): 2064-2069.

References

[1] COOTES T F, TAYLOR C J, LANITIS A. Active shape models:evaluation of a multi-resolution method for improving image search[C]//Proceedings of the 1994 British Machine Vision Conference.[S.l.]:BMVC Press, 1994:327-338.
[2] HOTELLING H. Analysis of a complex of statistical variables into principal components[J]. British Journal of Educational Psychology, 1933, 24(6):417-520.
[3] COOTES T F, EDWARDS G J, TAYLOR C J. Active appearance models[C]//Proceedings of the 1998 European Conference on Computer Vision. Berlin:Springer, 1998:484-498.
[4] CRISTINACCE D, COOTES T F. Feature detection and tracking with constrained local models[C]//Proceedings of the 2006 British Machine Vision Conference.[S.l.]:BMVC Press, 2006:929-938.
[5] DOLLÁR P, WELINDER P, PERONA P. Cascaded pose regression[C]//CVPR 2010:Proceedings of the 2010 IEEE Conference on Computer Vision and Pattern Recognition. Washington, DC:IEEE Computer Society, 2010:1078-1085.
[6] BURGOS-ARTIZZU X P, PERONA P, DOLLAR P. Robust face landmark estimation under occlusion[C]//Proceedings of the 2013 IEEE International Conference on Computer Vision. Washington, DC:IEEE Computer Society, 2013:1513-1520.
[7] XIONG X, TORRE F D L. Supervised descent method and its applications to face alignment[C]//Proceedings of the 2013 IEEE Conference on Computer Vision and Pattern Recognition. Washington, DC:IEEE Computer Society, 2013:532-539.
[8] LOWE D G. Distinctive image features from scale-invariant key-points[J]. International Journal of Computer Vision, 2004, 60(2):91-110.
[9] REN S, CAO X, WEI Y, et al. Face alignment at 3000 FPS via regressing local binary features[C]//CVPR 2014:Proceedings of the 2014 IEEE Conference on Computer Vision and Pattern Recognition. Washington, DC:IEEE Computer Society, 2014:1685-1692.
[10] 刘袁缘,谢忠,周顺平,等.基于条件迭代更新随机森林的非约束人脸特征点精确定位[J].计算机辅助设计与图形学学报,2017,29(10):1881-1890.(LIU Y Y, XIE Z, ZHOU S P, et al. Conditional iteration updated random forests for unconstrained facial feature location[J]. Journal of Computer-Aided Design and Computer Graphics, 2017, 29(10):1881-1890.)
[11] 张海艳,高尚兵,姜明新.应用姿态估计的人脸特征点定位算法[J].计算机应用,2017,37(11):3256-3260.(ZHANG H Y, GAO S B, JIANG M X. Facial feature localization algorithm using pose estimation[J]. Journal of Computer Applications, 2017, 37(11):3256-3260.)
[12] 贾项南,于凤芹,陈莹.改进的显式形状回归人脸特征点定位算法[J].计算机应用,2018,38(5):1289-1293.(JIA X N, YU F Q, CHEN Y. Improved explicit shape regression for face alignment[J]. Journal of Computer Applications, 2018,38(5):1289-1293.)
[13] ZHANG K, ZHANG Z, LI Z, et al. Joint face detection and alignment using multitask cascaded convolutional networks[J]. IEEE Signal Processing Letters, 2016, 23(10):1499-1503.
[14] SLICE D E. Landmark coordinates aligned by Procrustes analysis do not lie in Kendall's shape space[J]. Systematic Biology, 2001, 50(1):141-149.
[15] DALAL N, TRIGGS B. Histograms of oriented gradients for human detection[C]//Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision & Pattern Recognition. Washington, DC:IEEE Computer Society, 2005:886-893.
[16] KAZEMI V, SULLIVAN J. One millisecond face alignment with an ensemble of regression trees[C]//CVPR 2014:Proceedings of the 2014 Computer Vision and Pattern Recognition. Washington, DC:IEEE Computer Society, 2014:1867-1874.
[17] ZHANG J, SHAN S, KAN M, et al. Coarse-to-Fine Auto-Encoder Networks (CFAN) for real-time face alignment[C]//Proceedings of the 2014 European Conference on Computer Vision. Berlin:Springer, 2014:1-16.
[18] SUN Y, WANG X, TANG X. Deep convolutional network cascade for facial point detection[C]//CVPR 2013:Proceedings of the 2013 IEEE Conference on Computer Vision and Pattern Recognition. Washington, DC:IEEE Computer Society, 2013:3476-3483.