二进制特征与联合层叠结构的人脸识别研究
|
摘要
针对人脸识别阶段计算时间长的问题,提出一种基于二进制特征与联合层叠结构的人脸识别方法。在卷积神经网络中构建哈希层,将哈希层得到的编码作为分类器输入,同时加入Softmax分类损失函数和哈希损失函数作为优化目标之一;在学习特征表示的同时,学习它对应的哈希函数,使得提取到的特征从浮点型转换为二进制的特征,并使哈希函数满足独立性和量化误差最小的约束;针对哈希算法精度轻微下降的问题,通过联合级联结构将深度特征变换与深度二进制人脸哈希相结合,通过多种特征与多种度量的多次选择,最终匹配出最佳的目标作为结果。经实验验证,该算法在保证识别率的情况下,能缩短计算时间。
In order to solve the long time consumption of face identification,this paper proposed an algorithm of face recognition based on binary feature and joint layered structure.We constructed the hash layer in convolutional neural network,and the code obtained from the hash layer was input as a classifier.Softmax classification loss function and hash loss function were added as one of the optimization objectives.When learning feature representation,the corresponding hash function was also learned,which made the extracted feature transform from floating point to binary,and made the hash function satisfy the constraint of independence and minimum quantization error.Aiming at the slight precision reduction of hashing algorithm,the depth feature transformation was combined with the depth binary human face hash through the joint cascade structure.Through multiple selections of multiple features and multiple measurements,the best target was finally matched as the result.The experimental results show that the algorithm can shorten the computational time when the recognition rate is guaranteed.
In order to solve the long time consumption of face identification,this paper proposed an algorithm of face recognition based on binary feature and joint layered structure.We constructed the hash layer in convolutional neural network,and the code obtained from the hash layer was input as a classifier.Softmax classification loss function and hash loss function were added as one of the optimization objectives.When learning feature representation,the corresponding hash function was also learned,which made the extracted feature transform from floating point to binary,and made the hash function satisfy the constraint of independence and minimum quantization error.Aiming at the slight precision reduction of hashing algorithm,the depth feature transformation was combined with the depth binary human face hash through the joint cascade structure.Through multiple selections of multiple features and multiple measurements,the best target was finally matched as the result.The experimental results show that the algorithm can shorten the computational time when the recognition rate is guaranteed.
引文
[1]Kim K I,Jung K,Kim H J.Face recognition using kernel principal component analysis[J].IEEE signal processing letters,2002,9(2):40-42.
[2]PaulevéL,Jégou H,Amsaleg L.Locality sensitive hashing:A comparison of hash function types and querying mechanisms[J].Pattern Recognition Letters,2010,31(11):1348-1358.
[3]Gong Y,Lazebnik S,Gordo A,et al.Iterative quantization:a procrustean approach to learning binary codes for largescale image retrieval[C]//Computer Vision and Pattern Recognition.IEEE,2011:817-824.
[4]Liu H,Wang R,Shan S,et al.Deep supervised hashing for fast image retrieval[C]//IEEEConference on Computer Vision and Pattern Recognition.IEEE,2016:2064-2072.
[5]Strecha C,Bronstein A,Bronstein M,et al.LDAHash:Improved matching with smallerdescriptors[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2012,34(1):66-78.
[6]Chang J R,Chen Y S.Batch-normalized maxout network in network[EB].ar Xiv:1511.02583,2015.
[7]Szegedy C,Liu W,Jia Y,et al.Going deeper with convolutions[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.IEEE,2015:1-9.
[8]He K,Zhang X,Ren S,et al.Deep residual learning for image recognition[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.IEEE,2016:770-778.
[9]Weiss Y,Torralba A,Fergus R.Spectral hashing[C]//Conference on Neural Information Processing Systems,Vancouver,British Columbia,Canada,December.DBLP,2008:1753-1760.
[10]Zhang K,Zhang Z,Li Z,et al.Joint face detection and alignment using multi-task cascaded convolutional networks[J].IEEE Journal of Solid-State Circuits,2016,23(99):1161-1173.
[11]Chen D,Ren S,Wei Y,et al.Joint cascade face detection and alignment[M]//Computer Vision-ECCV 2014.Springer International Publishing,2014:109-122.
[12]Li J,Zhang Y.Learning SURF cascade for fast and accurate object detection[C]//IEEEConference on Computer Vision and Pattern Recognition.IEEE Computer Society,2013:3468-3475.
[13]Chen G,Shao Y,Tang C,et al.Deep transformation learning for face recognition in the unconstrained scene[J].Machine Vision&Applications,2018,29(10):1-11.
[14]Maaten L V D,Weinberger K.Stochastic triplet embedding[C]//IEEE International Workshop on Machine Learning for Signal Processing.IEEE,2012:1-6.
[15]Ng H W,Winkler S.A data-driven approach to cleaning large face datasets[C]//IEEEInternational Conference on Image Processing.IEEE,2014:343-347.
[16]Jin Z,Li C,Lin Y,et al.Density sensitive hashing[J].IEEE Transactions On Cybernetics,2014,44(8):1362-1371.
[17]Yu F X,Kumar S,Gong Y,et al.Circulant binary embedding[C]//International conference on machine learning.2014:7.
[18]Koutaki G,Shirai K,Ambai M.Fast Supervised Discrete Hashing and its Analysis[EB].ar Xiv:1611.10017,2016.
[19]Heo J P,Lee Y,He J,et al.Spherical hashing:binary code embedding with hyperspheres[J].IEEE Trans Pattern Anal Mach Intell,2015,37(11):2304.
[20]Yi D,Lei Z,Liao S,et al.Learning face representation from scratch[EB].ar Xiv:1411.7923,2014.
[2]PaulevéL,Jégou H,Amsaleg L.Locality sensitive hashing:A comparison of hash function types and querying mechanisms[J].Pattern Recognition Letters,2010,31(11):1348-1358.
[3]Gong Y,Lazebnik S,Gordo A,et al.Iterative quantization:a procrustean approach to learning binary codes for largescale image retrieval[C]//Computer Vision and Pattern Recognition.IEEE,2011:817-824.
[4]Liu H,Wang R,Shan S,et al.Deep supervised hashing for fast image retrieval[C]//IEEEConference on Computer Vision and Pattern Recognition.IEEE,2016:2064-2072.
[5]Strecha C,Bronstein A,Bronstein M,et al.LDAHash:Improved matching with smallerdescriptors[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2012,34(1):66-78.
[6]Chang J R,Chen Y S.Batch-normalized maxout network in network[EB].ar Xiv:1511.02583,2015.
[7]Szegedy C,Liu W,Jia Y,et al.Going deeper with convolutions[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.IEEE,2015:1-9.
[8]He K,Zhang X,Ren S,et al.Deep residual learning for image recognition[C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition.IEEE,2016:770-778.
[9]Weiss Y,Torralba A,Fergus R.Spectral hashing[C]//Conference on Neural Information Processing Systems,Vancouver,British Columbia,Canada,December.DBLP,2008:1753-1760.
[10]Zhang K,Zhang Z,Li Z,et al.Joint face detection and alignment using multi-task cascaded convolutional networks[J].IEEE Journal of Solid-State Circuits,2016,23(99):1161-1173.
[11]Chen D,Ren S,Wei Y,et al.Joint cascade face detection and alignment[M]//Computer Vision-ECCV 2014.Springer International Publishing,2014:109-122.
[12]Li J,Zhang Y.Learning SURF cascade for fast and accurate object detection[C]//IEEEConference on Computer Vision and Pattern Recognition.IEEE Computer Society,2013:3468-3475.
[13]Chen G,Shao Y,Tang C,et al.Deep transformation learning for face recognition in the unconstrained scene[J].Machine Vision&Applications,2018,29(10):1-11.
[14]Maaten L V D,Weinberger K.Stochastic triplet embedding[C]//IEEE International Workshop on Machine Learning for Signal Processing.IEEE,2012:1-6.
[15]Ng H W,Winkler S.A data-driven approach to cleaning large face datasets[C]//IEEEInternational Conference on Image Processing.IEEE,2014:343-347.
[16]Jin Z,Li C,Lin Y,et al.Density sensitive hashing[J].IEEE Transactions On Cybernetics,2014,44(8):1362-1371.
[17]Yu F X,Kumar S,Gong Y,et al.Circulant binary embedding[C]//International conference on machine learning.2014:7.
[18]Koutaki G,Shirai K,Ambai M.Fast Supervised Discrete Hashing and its Analysis[EB].ar Xiv:1611.10017,2016.
[19]Heo J P,Lee Y,He J,et al.Spherical hashing:binary code embedding with hyperspheres[J].IEEE Trans Pattern Anal Mach Intell,2015,37(11):2304.
[20]Yi D,Lei Z,Liao S,et al.Learning face representation from scratch[EB].ar Xiv:1411.7923,2014.