DEEP LEARNING FOR FACE

RECOGNITION

HAZIRLAYAN: Firas ALJANABİ (198146001008)

SORUMLU: Prof.Dr. Adem Alpaslan ALTUN
Introduction:

Face is the most common characteristic used by

humans for recognition. Face recognition (FR) is a
classical problem and is still very active in computer
vision and image understanding. Fig. 1 shows the
whole pipeline of an automatic face recognition
system. A face image is fed into the system, and face
detection and face alignment are processed. And
then a feature extractor is used to extract features.
Finally, the system compares the extracted features
with the gallery faces to do face matching.
 In face matching, there are two different tasks: face
verification (FV) and face identification (FI). FV is to
determine whether a given pair of face images or videos
belongs to the same subject. FI is a one-to-many
matching, recognizing the person from a set of gallery
face images or videos of different subjects. Face
identification usually assumes the query person has
already enrolled in the gallery, which is a closed-set
problem. Watch-list is similar to face identification but it
does not guarantee all query subjects are registered in the
gallery, which is an open-set problem. In the real world, it
is normal to treat FI as an open-set problem.
 Since the seminal work of (Turk et al., 1991) introduced
eigenfaces to the computer vision community, and (Viola et al.,
2001) proposed the use of Haar features for face detection, there
has been a significant increase in interest in facial recognition
technologies. Technical progress in this field can be separated into
four major areas of research interest, as shown in Figure 1. Holistic
approaches employ distributional concepts such as manifold
(Xiaofei et al., 2005; Yan et al., 2007; Deng et al., 2008) and sparse
representation (Wright et al., 2009; Zhang et al., 2001) and linear
subspace (Moghaddam et al., 1998;Belhumeur et al., 1997) to
create low-dimensional representations. These approaches were
however limited by the uncontrollable variations in facial
appearance, which naturally deviate from distribution assumptions.
 This limitation gave rise to the use of local feature-based
techniques such as Gabor wavelets (Chengjun et al., 2002;
Shen et al., 2007; Struc et al., 2009) and LBP (Ahonen et
al., 2006) which provided greater robustness due greater
invariance to transformations and environmental effects.
These methods were limited by lack of distinctiveness, a
problem which was partially rectified by the use of
shallow learning-based local descriptors (Can et al.,
2010;Chan et al., 2015;Lei et al., 2014) which improved
compactness and introduced automated encoding.
However, these shallow methods failed to achieve optimal
performance due to their inability to accurately represent
the complex non- linear variations in facial appearance.
 A major milestone in the development of facial recognition
techniques was achieved by the introduction of highly accurate
deep learning methods such as DeepFace (Taigman et al., 2014)
and DeepID (Sun et al., 2015). For the first time, face verification
in unconstrained settings was achieved with accuracy surpassing
human ability. This development was only allowed for by the
advent of significant improvements in hardware, such as high
capacity GPUs. Since then, the majority of research has focused
on the development of deep learning-based methods which
attempt to model the human brain, via high-level abstraction
achieved using a concurrence of non-linear filters resulting in
feature invariance. The majority of these methods rely on
increasingly deep CNNs, with an emphasis on promoting sparsity
and selectivity. Other deep learning methods.
 With the collection of various types of face data, research
concentrations have also focused on some specific tasks, such
as robustness to changes of pose, illumination, expression,
age, or improving performance of video, 3D, and
heterogeneous (e.g., NIR-VIS, Sketch-Photo, Still-to-Video)
face recognition. Although some related surveys overviewed
methods on handling pose (Ding and Tao, 2016), illumination
(Chan et al., 2014), expression (Murtaza et al., 2013),
occlusion (Lahasan et al., 2017), infrared (Ghiass et al.,
2014), single-modal and multimodal (Zhou et al., 2014),
video (Barr et al., 2012), 3D (Patilet al., 2015), heterogeneous
face matching (Guo, 2014; Ouyang et al.,2016a), etc., most of
them focused on the traditional methods, and few of them has
been related to deep learning methods.
 We present a complete, comprehensive overview of
face recognition works using deep learning, considering
both the deep architectures and specific recognition
problems. We also give a review of related
face databases. Fig. 3(a) shows some statistics of the
related papers. Applying deep learning to face image
analysis started several years ago, while the number of
papers have been growing rapidly, as shown in Fig.
3(b). This survey includes about 330 papers, and most
of them are within the recent five years. It is expected
to cover most, if not all, of the works incorporating
deep learning methods for face recognition.
Fig. 3. Statistical figures (best view in color) of the related papers: (a) distribution by category,
(b) distribution by year, (c) distribution of databases by face type, e.g., still, video,
heterogeneous faces.
 By this survey, we show that:
• deep learning methods have been fully applied to
face recognition and played important roles;
• many specific issues or challenges to address in
FR by DL, such as pose, illumination,
expression, 3D, heterogeneous matching, etc.;
• various face datasets collected in recent years,
including still
images, videos, and heterogeneous data that raises
the issue of
cross-modal face matching.
 Deep learning methods
Artificial Neural Network (ANN) (Haykin, 2009) is a
computational nonlinear model inspired by the biological systems
in information processing. It consists of artificial neurons or
processing elements and is typically organized in three types of
interconnected layers. Data are presented to the network via the
input layer, which communicates to one or more hidden layers
where the actual processing is done via weighted connections. The
hidden layers then link to an output layer to give the output. It is
possible to make the neural network more flexible and more
powerful by using additional hidden layers. Artificial neural
networks with many hidden layers between the input and output
layers are called Deep Neural Networks (DNNs), and they can
model complex relationships between the input and output.
 There are various deep neural networks used in face recognition.
Convolutional Neural Network (CNN) is the most popular. It
shows outstanding results in image and speech applications.
Autoencoder (AE) and its variants also gained much attention.
They process data without using class labels and the purpose is to
find patterns, such as latent subspaces. Generative Adversarial
Network (GAN) has increased rapidly recently. It usually
contains two nets, putting one against the other, thus called
adversarial. It can learn to mimic any distribution of data. Deep
Belief Network (DBN), Deep Boltzmann Machine (DBM) are
also used in FR. However, the Recurrent Neural Network (RNN),
Self-Organizing Map (SOM), Radial Basis Function Network
(RBFN), are not used very often in FR. Fig. 3 gives the
distribution of research works for different neural network
architectures.
 Convolutional neural networks
(CNNs)
In the last decade, Convolutional Neural Network (CNN) (LeCunet al.,
1998) has become one of the most popular techniques in computer
vision, such as image classification (He et al., 2016a), object detection
(Redmon et al., 2016), face recognition (Yi et al., 2014). Many vision
tasks have benefited from the robust and discriminative representation
learned via CNN and the performance has been improved significantly.
Probably the first successful real world application of CNN is LeNet
(LeCun et al., 1998) for hand-written digit recognition. AlexNet
(Krizhevsky et al., 2012) is considered one of the most influential
research work for DL in Computer Vision, having spurred many works
employing CNNs and GPUs to accelerate deep learning research and
development (Simonyan and Zisserman, 2014; Szegedy et al., 2015). In
face recognition, deep CNN has now become the technique of choice.
Single CNN
Typical deep face recognition methods adopt one single CNN,
which is usually trained in a supervised fashion. Table 1 gives
an overview some of FR methods based on a single deep
CNN. The earlier DeepFace (Taigman et al., 2014) is a 9-
layer CNN in which the input is RGB face images
preprocessed with 3D-alignment. Several locally connected
convolutional layers are adopted without weight sharing, and
every location in feature maps of these layers learns a
different set of filters. As an extension of DeepFace, Web-
Scale (Taigman et al., 2015) used a bootstrapping process to
select more efficient training set by replacing the naive
random subsampling.
Multi-CNN

Choosing more than one CNNs to extract different deep

features and concatenate them in some way as the final
face representation is commonly investigated too. It
mainly contains two types: (1) extracting deep features of
different regions of a face, and (2) extracting features of
different aspects of faces. These deep models (as shown
in Table 2) usually require additional training data to train
each CNN. It is necessary to explore particular modalities
that can contribute to enhance their performance, which
requires significant efforts in terms of data preparation or
selection and computing resources.
Variants of CNN

Except the traditional CNN based architectures,

some research works (in Table 3) have been done
to investigate unconventional CNN based
framework by (1) designing totally different
layout of CNNs, (2) modifying kernel learning
components or the way of kernel activation, (3)
fusing CNNs with other types of modules, (4)
adopting weakly-supervised or unsupervised
learning, and (5) others.
 Face recognition models based
on CNN

 CNN based models are different

– Architecture of CNN
– Depth of neural network
– Number of parameters
– Scale of training dataset
– Similarity metric
– Alignment vs. non-alignment pre-processing
 Deep Face Model

 First CNN-based face recognition method (2014)

– By Face book research group
 Includes 4 main steps
– Detection
– 3D Alignment
– Feature representation
– Classification
 Similarity metric learning
– Siamese energy based neural network
Deep Face Model
Web-scaled Deep Face Model

 Based on Deep Face (2015)

– Higher identification acc.
– Lower verification acc.
– Lower feature vector dimension
 Applies bootstrapping on large training dataset

– Select harder recognition cases

– Ignore easy recognition cases
 Claim

– High dimensional feature vectors do not necessarily result in

better accuracy!
 Deep ID model series

 Inspired by Deep Face

 Model Series

– DeepID (2014)
– DeepID2 (2014)
– DeepID2+ (2015)
– DeepID3 (2015)
 Deep ID model

 CNN structure
– 4 convolutional layers
– 3 max-pooling layers
– 1 fully connected layer
 Alignment
– Center of two eyes, two corners of mouth, nose tip.
 Multiple patches
– Extract different features for different part of face
 Similarity metric learning
– Joint-Bayesian
Deep ID2 Model
Deep ID2+ Model

 Based on DeepID2
– More deep network
– Uses supervisory signals
– Uses fully connected layers
 Fully connected layers
– Early feature extraction
 Claim
– Deep CNN based networks
are more robust to
corruption of image!
Deep ID3 Model

 DeepID3
– More Deep than DeepID2+
– Less deep than FaceNet&
VGG
 Includes
– Very deep neural networks
– 15 feature extraction layers
– Early fully connected layers
 Similarity metric learning
– Joint-Bayesian
Face Net Model

 FaceNetmodel: by Google
research group
– Same framework for
identification and verification
– Very deep network
– No alignment
– Efficient representation of
features
 Similarity metric learning
– Triplet loss
 VGG Model

 VGG model: by Visual Geometry Group

– Inspired by the very deep FaceNetnetwork
– Very deep CNN
– 36 level of feature extraction
 Similarity metric
– Triplet loss
 Contributions
– Automatic collection of large face dataset
– Publically available pre-trained CNN model
 Lightened CNN Model

 Shallow network
– 4 convolutional layers
– 4M parameters
 Less computational intensive

– 9 times less than VGG model

 References:
 Ahonen, T., A. Hadid, and M. Pietikainen, Face Description with Local Binary Patterns: Application to
Face Recognition. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2006. 28(12): p.
2037-2041.
 Barr, J.R., Bowyer, K.W., Flynn, P.J., Biswas, S., 2012. Face recognition from video: A review. Intl. J.
Pattern Recognit. Artif. Intell. 26 (05), 1266002.
 Belhumeur, P.N., J.P. Hespanha, and D.J. Kriegman, Eigenfaces vs. Fisherfaces: recognition using class
specific linear projection. IEEE Transactions on Pattern Analysis and Machine Intelligence, 1997. 19(7):
p. 711-720.
 Bodla, N., Zheng, J., Xu, H., Chen, J.-C., Castillo, C., hellappa, R., 2017. Deep heterogeneous feature
fusion for template-based face recognition. In: Applications of Computer Vision, Winter Conf. on. IEEE,
pp. 586–595.
 Cao, Z., et al. Face recognition with learning-based descriptor. in 2010 IEEE Computer Society
Conference on Computer Vision and Pattern Recognition. 2010.
 Chan, C.H., Zou, X., Poh, N., Kittler, J., 2014. Illumination invariant face recognition:a survey. Face
Recognit. Adverse Cond. 147–166.
 Chan, T.H., et al., PCANet: A Simple Deep Learning Baseline for Image Classification? IEEE
Transactions on Image Processing, 2015. 24(12): p. 5017-5032.
 Chan, T.-H., Jia, K., Gao, S., Lu, J., Zeng, Z., Ma, Y., 2015. Pcanet: A simple deep learning baseline for
image classification? IEEE Trans. Image Process. 24 (12), 5017–5032.
 Chen, B., Deng, W., 2016. Weakly-supervised deep self- learning for face recognition.In:
Multimedia and Expo, Intl. Conf. on. IEEE, pp. 1–6.
 Chen, J.-C., Ranjan, R., Kumar, A., Chen, C.-H., Patel, V.M., Chellappa, R., 2015b. An end-
to-end system for unconstrained face verification with deep convolutional neural networks.
In: Proceedings of IEEE Intl. Conf. on Computer Vision Workshops.pp. 118–126.
 Chen, J.-C., Zheng, J., Patel, V.M., Chellappa, R., 2016b. Fisher vector encoded deep
convolutional features for unconstrained face verification. In: Image Processing, Intl. Conf.
on. IEEE, pp. 2981–2985.
 Chengjun, L. and H. Wechsler, Gabor feature based classification using the enhanced fisher
linear discriminant model for face recognition. IEEE Transactions on Image Processing,
2002. 11(4): p. 467-476.
 Chowdhury, A.R., Lin, T.-Y., Maji, S., Learned-Miller, E., 2015. Face identification with
bilinear cnns. arXiv preprint arXiv:1506.01342.
 Chowdhury, A.R., Lin, T.-Y., Maji, S., Learned-Miller, E., 2016. One-to-many face
recognition with bilinear cnns. In: Applications of Computer Vision, Winter Conf. on. IEEE,
pp. 1–9.
 Deng, W., et al., Comments on "Globally Maximizing, Locally Minimizing: Unsupervised
Discriminant Projection with Application to Face and Palm Biometrics". IEEE Transactions
on Pattern Analysis and Machine Intelligence, 2008. 30(8): p. 1503-1504.
 Ding, C., Tao, D., 2016. A comprehensive survey on pose-invariant face recognition. ACM
Trans. Intell. Syst. Technol. 7 (3), 37.
 Ding, Y., Cheng, Y., Cheng, X., Li, B., You, X., Yuan, X., 2017. Noise-resistant network: a
deep-learning method for face recognition under noise. EURASIP J. Image Video Process.
2017 (1), 43.
 Dong, B., An, Z., Lin, J., Deng, W., 2017a. Attention-based template adaptation for face
verification. In: Automatic Face & Gesture Recognition, Intl. Conf. on. IEEE, pp.941–946.
 Fan, H., Cao, Z., Jiang, Y., Yin, Q., Doudou, C., 2014. Learning deep face representation.
arXiv preprint arXiv:1403.2802.
 Fu, T.-C., Chiu, W.-C., Wang, Y.-C.F., 2017. Learning guided convolutional neural networks
for cross-resolution face recognition. In: Machine Learning for Signal Processing (MLSP),
2017 IEEE 27th International Workshop on. IEEE, pp. 1–5.
 Ghiass, R.S., Arandjelović, O., Bendada, A., Maldague, X., 2014. Infrared face recognition: A
comprehensive review of methodologies and databases. Pattern Recognit. 47 (9), 2807–2824.
 Gruber, I., Hlav‫ل‬č, M., Železn`y, M., Karpov, A., 2017. Facing face recognition with resnet:
Round one. In: Intl. Conf. on Interactive Collaborative Robotics. Springer, pp. 67–74.
 Grundstrِm, J., 2015. Face verification and open-set dentification for real-time video
applications. In: Master’s Theses in Mathematical Sciences.
 Guo, G., Wen, L., Yan, S., 2014. Face authentication with makeup changes. IEEE Trans.
Circuits Syst. Video Technol. 24 (5), 814–825.
 Haykin, S., 2009. Neural Networks and Learning Machines, Vol. 3. Pearson education Upper
Saddle River.
 He, K., Zhang, X., Ren, S., Sun, J., 2016a. Deep residual learning for image recognition. In:
Proceedings of the IEEE Conf. on Computer Vision and Pattern Recognition. pp.770–778.
 Hsieh, H.-L., Hsu, W., Chen, Y.-Y., 2017. Multi-task learning for face identification and
attribute estimation. In: Acoustics, Speech and Signal Processing, Intl. Conf. on. IEEE, pp.
2981–2985.
 Hu, G., Hua, Y., Yuan, Y., Zhang, Z., Lu, Z., Mukherjee, S.S., Hospedales, T.M., Robertson,
N.M., Yang, Y., 2017a. Attribute-enhanced face recognition with neural tensor fusion
networks. In: Proceedings of IEEE Conf. on Computer Vision and Pattern Recognition. pp.
3744–3753.
 Huang, Z., Shan, S., Wang, R., Zhang, H., Lao, S., Kuerban, A., Chen, X., 2015. A
benchmark and comparative study of video-based face recognition on cox face database.
IEEE Trans. Image Process. 24 (12), 5967–5981.
 Kang, B.-N., Kim, Y., Kim, D., 2017. Deep convolutional neural network using triplets of
faces, deep ensemble, and score-level fusion for face recognition. In: Proceedings of IEEE
Conf. on Computer Vision and Pattern Recognition Workshops. pp. 109–116.
 Krizhevsky, A., Sutskever, I., Hinton, G.E., 2012. Imagenet classification with deep
convolutional neural networks. In: Advances in Neural Information Processing Systems. pp.
1097–1105.
 Lahasan, B., Lutfi, S.L., San-Segundo, R., 2017. A survey on techniques to handle face
recognition challenges: occlusion, single sample per subject and expression. Artif.
Intell. Rev. 1–31.
 LeCun, Y., Bottou, L., Bengio, Y., Haffner, P., 1998. radient-based learning applied to
document recognition. Proc. IEEE 86 (11), 2278–2324.
 Lei, Z., M. Pietikäinen, and S.Z. Li, Learning Discriminant Face Descriptor. IEEE
Transactions on Pattern Analysis and Machine Intelligence, 2014. 36(2): p. 289-302.
 Li, S., Xing, J., Niu, Z., Shan, S., Yan, S., 2015a. Shape driven kernel adaptation in
convolutional neural network for robust facial traits recognition. In: Proceedings of the
IEEE Conf. on Computer Vision and Pattern Recognition. pp. 222–230.
 Lin, T.-Y., RoyChowdhury, A., Maji, S., 2015. Bilinear cnn models for fine-grained
visual recognition. In: Proceedings of the IEEE Intl. Conf. on Computer Vision. pp.
1449–1457.
 Liu, J., Deng, Y., Bai, T., Wei, Z., Huang, C., 2015. Targeting ultimate accuracy: Face
recognition via deep embedding. arXiv preprint arXiv:1506.07310.
 Lu, J., Liong, V.E., Wang, G., Moulin, P., 2015a. Joint feature learning for face
recognition. IEEE Trans. Inf. Forensics Secur. 10 (7), 1371–1383.
 Lu, X., Yang, Y., Zhang, W., Wang, Q., Wang, Y., 2017c. Face verification with multi-
task and multi-scale feature fusion. Entropy 19 (5), 228.
 Lumini, A., Nanni, L., Ghidoni, S., 2016. Deep features combined with hand-crafted
features for face recognition. Intl. J. Comput. Res. 23 (2), 123.
 Moghaddam, B., W. Wahid, and A. Pentland. Beyond eigenfaces: probabilistic
matching for face recognition. in Proceedings Third IEEE International Conference on
Automatic Face and Gesture Recognition. 1998.
 Murtaza, M., Sharif, M., Raza, M., Shah, J.H., 2013. Analysis of face recognition under
varying facial expression: a survey.. Int. Arab J. Inf. Technol. 10 (4), 378–388.
 Ouyang, S., Hospedales, T., Song, Y.-Z., Li, X., Loy, C.C., Wang, X., 2016a. A survey on
heterogeneous face recognition: Sketch, infra-red, 3d and low-resolution. Image Vis.
Comput. 56, 28–48.
 Park, S., Yu, J., Jeon, M., 2017. Learning feature representation for face verification. In:
Advanced Video and Signal Based Surveillance, Intl. Conf. on. IEEE, pp. 1–6.
 Patil, H., Kothari, A., Bhurchandi, K., 2015. 3-d face recognition: features, databases,
algorithms and challenges. Artif. Intell. Rev. 44 (3), 393–441.
 Ranjan, R., Sankaranarayanan, S., Castillo, C.D., Chellappa, R., 2016. An all-in-one
convolutional neural network for face nalysis. arXiv preprint arXiv:1611.00851.
 Redmon, J., Divvala, S., Girshick, R., Farhadi, A., 2016. You only look once: Unified,
real-time object detection. In:Proceedings of the IEEE Conf. on Computer Vision and
Pattern Recognition. pp. 779–788.
 Shen, L., L. Bai, and M. Fairhurst, Gabor wavelets and General Discriminant Analysis for
face identification and verification. Vol. 25. 2007. 553-563.
 Simَn, M.O., Corneanu, C., Nasrollahi, K., Nikisins, O., Escalera, S., Sun, Y., Li, H., Sun,
Z., Moeslund, T.B., Greitans, M., 2016. Improved rgb-dt based face recognition. IET
iometrics 5 (4), 297–303.
 Simonyan, K., Zisserman, A., 2014. Very deep convolutional networks for large-scale
image recognition. arXiv preprint arXiv:1409.1556.
 Smirnov, E., Melnikov, A., Novoselov, S., Luckyanets, E., Lavrentyeva, G., 2017.
Doppelganger mining for face representation learning. In: Proceedings of IEEE Conf. on
Computer Vision and Pattern Recognition. pp. 1916–1923.
 Štruc, V., R. Gajšek, and N. Pavešić. Principal Gabor filters for face recognition. in 2009
IEEE 3rd International Conference on Biometrics: Theory, Applications, and Systems.
2009.
 Sun, Y., Chen, Y., Wang, X., Tang, X., 2014a. Deep learning face representation by joint
identification-verification. In: Advances in Neural Information Processing Systems.pp.
1988–1996.
 Sun, Y., Deep Learning Face Representation by Joint Identification-Verification, J. Huang,
Editor. 2015, ProQuest Dissertations Publishing.
 Sun, Y., Liang, D., Wang, X., Tang, X., 2015a. Deepid3: Face recognition with very deep
neural networks. arXiv preprint arXiv:1502.00873.
 Sun, Y., Wang, X., Tang, X., 2014b. Deep learning face representation from predicting
10,000 classes. In: Proceedings of the IEEE Conf. on Computer Vision and Pattern
Recognition. pp. 1891–1898.
 Sun, Y., Wang, X., Tang, X., 2015b. Deeply learned face representations are sparse,
selective, and robust. In: Proceedings of the IEEE Conf. on Computer Vision and Pattern
Recognition. pp. 2892–2900.
 Szegedy, C., Liu, W., Jia, Y., Sermanet, P., Reed, S., Anguelov, D., Erhan, D.,
Vanhoucke,Taigman, Y., et al., DeepFace: Closing the gap to human-level performance in
face verification. 2014. p. 1701-1708.
 Taigman, Y., Yang, M., Ranzato, M., Wolf, L., 2014. Closing the gap to human-level
performance in face verification. deepface. In: IEEE Computer Vision and Pattern
Recognition (CVPR).
 Taigman, Y., Yang, M., Ranzato, M., Wolf, L., 2015. Web-scale training for face
identification. In: Proceedings of the IEEE Conf. on Computer Vision and Pattern
Recognition. pp. 2746–2754.
 Tian, L., Fan, C., Ming, Y., 2016. Multiple scales combined principle component analysis
deep learning network for face recognition. J. Electron. Imaging 25 (2), 023025.
 Tian, L., Fan, C., Ming, Y., Jin, Y., 2015a. Stacked pca network (spcanet): an effective deep
learning for face recognition. In: Digital Signal Processing, Intl. Conf. on. IEEE, pp. 1039–
1043.
 Tian, L., Fan, C., Ming, Y., Shi, J., 2015b. Srdanet: an efficient deep learning algorithm for
face analysis. In: Intl. Conf. on Intelligent Robotics and Applications. Springer, pp. 499–510.
 Turk, M. and A. Pentland, Eigenfaces for recognition. J. Cognitive Neuroscience, 1991. 3(1):
p. 71-86.
 V., Rabinovich, A., 2015. Going deeper with convolutions. In: Proceedings of the IEEE
Conf. on Computer Vision and Pattern Recognition. pp. 1–9.
 Viola, P. and M. Jones, Rapid object detection using a boosted cascade of simple features.
2001. p. I511-I518.
 Wang, W., Wang, R., Shan, S., Chen, X., 2017c. Discriminative covariance oriented
representation learning for face recognition with image sets. In: Proceedings of the IEEE
Conf. on Computer Vision and Pattern Recognition. pp. 5599–5608.
 Wang, W., Yang, J., Xiao, J., Li, S., Zhou, D., 2014. Face recognition based on deep learning.
In: Intl. Conf. on Human Centered Computing. Springer, pp. 812–820.
 Wright, J., et al., Robust Face Recognition via Sparse Representation. IEEE Transactions on
Pattern Analysis and Machine Intelligence, 2009. 31(2): p. 210-227.
 Wu, W., Kan, M., Liu, X., Yang, Y., Shan, S., Chen, X., 2017a. Recursive spatial transformer
(rest) for alignment-free face recognition. In: Proceedings of IEEE Conf.on Computer Vision
and Pattern Recognition. pp. 3772–3780.
 Xi, M., Chen, L., Polajnar, D., Tong, W., 2016. Local binary pattern network: a deep learning
approach for face recognition. In: Image Processing, Intl. Conf. on. IEEE, pp. 3224–3228.
 Xiaofei, H., et al., Face recognition using Laplacianfaces. IEEE Transactions on Pattern
Analysis and Machine Intelligence, 2005. 27(3): p. 328-340.
 Xiong, C., Zhao, X., Tang, D., Jayashree, K., Yan, S., Kim, T.-K., 2015. Conditional
convolutional neural network for modality-aware face recognition. In: Proceedings of the IEEE
Intl. Conf. on Computer Vision. pp. 3667–3675.
 Xiong, L., Karlekar, J., Zhao, J., Feng, J., Pranata, S., Shen, S., 2017. A good practice towards
top performance of face ecognition: Transferred deep feature fusion. arXiv preprint
arXiv:1704.00438.
 Yan, S., et al., Graph Embedding and Extensions: A General Framework for Dimensionality
Reduction. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2007. 29(1): p. 40-
51.
 Yang, J., Ren, P., Zhang, D., Chen, D., Wen, F., Li, H., Hua, G., 2017a. Neural aggregation
network for video face recognition. In: Proceedings of the IEEE Conference on Computer
Vision and Pattern Recognition. pp. 4362–4371.
 Yi, D., Lei, Z., Liao, S., Li, S.Z., 2014. Learning face epresentation from scratch. arXiv preprint
arXiv:1411.7923.
 Zhang, L., M. Yang, and F. Xiangchu. Sparse representation or collaborative representation:
Which helps face recognition? in 2011 International Conference on Computer Vision. 2011.
 Zheng, J., Chen, J.-C., Bodla, N., Patel, V.M., Chellappa, R., 2016. Vlad encoded deep
convolutional features for unconstrained face verification. In: Pattern Recognition, Intl. Conf.
on. IEEE, pp. 4101–4106.
 Zhou, E., Cao, Z., Yin, Q., 2015. Naive-deep face recognition: Touching the limit
of lfw benchmark or not? arXiv preprint arXiv:1501.04690.
 Zhou, H., Mian, A., Wei, L., Creighton, D., Hossny, M., Nahavandi, S., 2014.
Recent advances on singlemodal and multimodal face recognition: a survey. IEEE
Trans. Hum.-Mach. Syst. 44 (6), 701–716.
 Zhu, Z., Luo, P., Wang, X., Tang, X., 2014b. Recover anonical-view faces in the
wild with deep neural networks. arXiv preprint arXiv:1404.3543.
Teşekkürler