基于多任务级联卷积网络模型的人脸检测和识别Face Detection and Recognition Based on a Multitask Cascaded Convolution Network Model
刘其嘉,郭一娜,任晓文,李健宇
摘要(Abstract)：
深度学习的集成特征提取这一优点使得它广泛应用于人脸检测和识别。提出了一种多任务级联卷积网络模型(Multitask Cascaded Convolution Network,MTCNN)。基于Tensor Flow平台,基于改进的任务级联卷积网络模型检测到人脸,并且用Face Net算法对人脸进行特征提取,用KNN算法对人脸进行识别。实验结果表明,对不同光照下多人图像和遮挡图像的人脸进行检测和识别,具有良好的鲁棒性。
关键词(KeyWords)： 人脸检测和识别;深度学习;Tensor Flow;多任务级联卷积网络;Face Net
基金项目(Foundation): 国家自然科学基金(61301250);; 山西省高等学校优秀青年学术带头人支持计划(晋教科[2015]3号);; 山西省回国留学人员科研资助项目(2014-060)
作者(Author): 刘其嘉,郭一娜,任晓文,李健宇
参考文献(References)：
- [1] 严严,陈日伟,王菡子.基于深度学习的人脸分析研究进展[J].厦门大学学报,2017,56(1):13-24.
- [2] TURKTURK M M,PENTLAND A.Eigenfaces for recoginition[J]. Journa of cognitive Neuroscience,1991,3(1):71-86.
- [3] SWETS D L, WENG J J.Using discriminant eigenfeatures for image retrieval [J].IEEE Transactions on Pattern Analysis&Machine Intelligence, 1996(8):831-836.
- [4] HINTON G E, OSINDERO S,TEH Y. A fast learning algorithm for deep belief nets[J]. Neural Computation,2006,18(7):1527-1554.
- [5] VIOLA P, JONES M.Robust Real-Time Face Detection[J]. Internation J. Computer Vision,2004, 57(2):137-154.
- [6] YANIG TAIGMAN, MING YANG. MARC'AURELIO RANZATO.DeepFace: Closing the Gap to Human-Level Performance in Face Verification[C]// Computer Vision ard pattern Recognition,2014 IEEE conference on IEEE 2014:1701-1708.
- [7] SCHROFF F, KALENICHENKO D, PHILBIN J. Facenet: A unified embedding for face recognition and clustering[J]. arXiv preprint arXiv,2015:1503.03832.
- [8] 孙超利,郭一娜,谭瑛.径向基函数神经网络辅助的微粒群算法[J].太原科技大学学报,2017,38(3):178-184.
- [9] ZHANG KAIPENG, ZHANG ZHANPENG, LI ZHIFENG.Joint Face Detection and Alignment using Multi-task Cascaded Convolutional Networks[J].Computer Vision and Pattern Recognition,2016,23(10):1499-1503.
- [10] Y P L I. FU , MOHAMMED U. Elder J H. et al. Probabilistic models for inference about identity [J].TPAMI,2012,34(1):144-157.
- [11] CHEN D, CAO X, SUN F W J. Blessing of dimensionality: High-dimensional feature and its efficient compression for face verification[J]. 2013. DOI:10.1109/cvpr. 2013. 389.