Research on 3D Face Alignment Method based on Convolutional Neural Network

Yuding Ke

《International Core Journal of Engineering》 8卷9期 (2022/09) Pp. 45-52

https://doi.org/10.6919/ICJE.202209_8(9).0009

摘要

With the decline of hardware device price and the development of image recognition analysis technology, face recognition technology has been widely used in various industries. As an important part of face recognition, face alignment bridges the two important steps of face detection and feature extraction, and plays the role of the top and bottom. Relevant research results show that if face alignment is effective, face recognition performance is effectively improved, and vice versa. Most of the traditional face alignment methods use two-dimensional faces for alignment, and two-dimensional face pictures usually cannot represent the change of face depth, which leads to the problem of inconsistent shape when the face is rotated in three-dimensional space. Because faces are usually rotatable in three dimensions, the use of two-dimensional alignment methods is prone to large errors. To this end, this paper proposes a 3D face alignment method based on convolutional neural networks. The method uses the end-to-end idea to complete the face alignment by inputting only one pair of face images and using the PRN (Position Map Regression Network) structure to locate the key points of the face, so as to obtain the 3D information of the face. In order to further optimize the network model parameters, the same weight ratio of the face area not is set and the loss function is improved.

關鍵字

3D Face Alignment ； Convolutional Neural Network ； 3D Deformation Model

參考文獻

Cootes T F, Edwards G J, Taylor C J. Active appearance models[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2001, 23(6):681-685.

Cootes T F, Taylor C J. Active Shape Model Search using Local Grey-Level Models: A Quantitative Evaluation[C]//Machine Vision Conference, British, 1993:639-648.

Cristinacce D, Cootes T F. Feature detection and tracking with constrained local models[C]//Proceedings of the British Machine Vision Conference,Edinburgh, 2006:1-10.

Cristinacce D, Cootes T F. Feature detection and tracking with constrained local models[C]//Proceedings of the British Machine Vision Conference. Edinburgh,2006:929-938.

Dollár P, Welinder P, Perona P. Cascaded pose regression[C]//Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition,San Francisco,2010,1078-1085.