摘 要：人脸关键点检测是计算机视觉任务中研究的一个重要话题。但目前的人脸关键点检测算法只能有效地提取人脸的表观信息，未能充分挖掘人脸的结构信息。为了解决上述问题，提出了循环推理网络用于人脸关键点的检测，通过分批次循环递归地学习人脸结构信息和人脸表观信息，使得神经网络能有效地提取人脸的结构信息。通过在 AFLW2000-3D 数据集的实验表明，文章的算法优于其他经典的人脸关键点检测算法。
关键词：人脸关键点检测；人脸表观信息；人脸结构信息；神经网络
DOI:10.19850/j.cnki.2096-4706.2023.04.023
中图分类号：TP391.4 文献标识码：A 文章编号：2096-4706（2023）04-0091-04
Face Landmark Detection Algorithm Based on Recurrent Inference Network
WANG Xing
(Beijing China-Power Information Technology Co., Ltd., Beijing 102208, China)
Abstract: Face landmark detection is an important topic in computer vision tasks. However, the current face landmark detection algorithm can only effectively extract the face apparent information, but not fully explore the structural information of the face. In order to solve the above problems, a recurrent inference network is proposed to detect the landmark of the face. The neural network can effectively extract the structural information of the face by learning the structural information and the apparent information of the face recursively in batches. Experiments on AFLW2000-3D datasets show that the proposed algorithm outperforms other classical face landmark detection algorithms in this paper.
Keywords: face landmark detection; face apparent information; face structure information; neural network
参考文献：
[1] JIN Y，LI Q，JIANG D，et al. High-fidelity 3D face reconstruction with multi-scale details [J].Pattern Recognition Letters， 2022，153：51-58.
[2] ZHAO D，QI Y. Generative landmarks guided eyeglasses removal 3D face reconstruction [C]//International Conference on Multimedia Modeling. Cham：Springer ，2022：109-120.
[3] SADIQ M，SHI D. Attentive occlusion-adaptive deep network for facial landmark detection [J/OL].Pattern Recognition，2022，125：[2022-09-20].https://doi.org/10.1016/j.patcog.2021.108510.
[4] FENG S，NONG X，HU H. Cascaded structure-learning network with using adversarial training for robust facial landmark detection [J].ACM Transactions on Multimedia Computing， Communications，and Applications （TOMM），2022，18（2）：1-20.
[5] BLANZ V，VETTER T. Face recognition based on fitting a 3d morphable model [J].IEEE Transactions on pattern analysis and machine intelligence，2003，25（9）：1063-1074.
[6] LEE Y J，LEE S J，PARK K R，et al. Single view-based 3D face reconstruction robust to self-occlusion [J].EURASIP Journal on Advances in Signal Processing，2012（1）：1-20.
[7] QU C，MONARI E，SCHUCHERT T，et al. Fast，robust and automatic 3D face model reconstruction from videos [C]//2014 11th IEEE International Conference on Advanced Video and Signal Based Surveillance （AVSS）.Seoul：IEEE，2014：113-118.
[8] LIU F，ZENG D，LI J，et al. Cascaded regressor based 3D face reconstruction from a single arbitrary view image[J/OL].arXiv： 1509.06161 [cs.CV].（2015-08-21）.https://arxiv.org/abs/1509.06161v1.
[9] FENG L，ZENG D，ZHAO Q，et al. Joint face alignment and 3d face reconstruction [C]//European Conference on Computer Vision.Cham：Springer，2016：545-560.
[10] WANG X，LI X，WU S. Graph structure reasoning network for face alignment and reconstruction [C]//International Conference on Multimedia Modeling. Cham：Springer，2021：493-505.
[11] 王兴 . 基于人脸关键点和三维重建的算法研究与实现[D]. 宁夏：宁夏大学，2020.
[12] JIANG L，WU X J，Kittler J. Dual Attention MobDenseNet （DAMDNet） for Robust 3D Face Alignment [C]//2019 IEEE/CVF International Conference on Computer Vision Workshop （ICCVW）. Seoul：IEEE，2019：504-513.
[13] FENG Z H，KITTLER J，AWAIS M，et al. Wing loss for robust facial landmarklocalisation with convolutional neural networks [C]//In 2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.IEEE，2018：2235-2245.
[14] CHENG X，WANG P，YANG R. Learning depth with convolutional spatial propagation network [J].arXiv：1810.02695 [cs. CV].（2018-10-04）.https://arxiv.org/abs/1810.02695v3.
[15] BELHUMEUR P N，JACOBS D W，KRIEGMAN D J， et al. Localizing parts of faces using a consensus of exemplars [J].IEEE transactions on pattern analysis and machine intelligence，2013，35（12）： 2930-2940.
[16] YU X，HUANG J，ZHANG S，et al. Face landmark fitting via optimized part mixtures and cascaded deformable model [J]. Transactions on Pattern Analysis & Machine Intelligence，2016（11）： 2212-2226.
[17] BURGOS-ARTIZZU X P，PERONA P，DOLLAR P. Robust face landmark estimation under occlusion [C]//2013 IEEE International Conference on Computer Vision.Sydney：IEEE，2013：1513-1520.
[18] CAO X，WEI Y，WEN F，et al. Face alignment by explicit shape regression [J].International Journal of Computer Vision，2014， 107（2）：177-190.
[19] ZHU X，LEI Z，LIU X，et al. Face alignment across large poses：A 3d solution [C]//Proceedings of the IEEE conference on computer vision and pattern recognition.IEEE，2016：146-155.
作者简介：王兴（1996—），男，汉族，宁夏中卫人，研发工程师，中级职称，硕士研究生，研究方向：计算机视觉（人脸关键点检测、人脸重建、年龄估计、OCR 识别技术）。