A learning framework for age rank estimation based on face images with scattering transform

OUR OFFICES @CHENNAI/ TRICHY / KARUR / ERODE / MADURAI / SALEM / COIMBATORE / BANGALORE / HYDRABAD
CELL: +91 9894917187 | 875487 1111 / 2111 / 3111 / 4111 / 5111 / 6111
ECWAY TECHNOLOGIES
IEEE SOFTWARE | EMBEDDED | MECHANICAL | ROBOTICS PROJECTS DEVELOPMENT
Visit: www.ecwaytechnologies.com | www.ecwayprojects.com Mail to: ecwaytechnologies@gmail.com
A LEARNING FRAMEWORK FOR AGE RANK ESTIMATION BASED ON FACE
IMAGES WITH SCATTERING TRANSFORM
By
A
PROJECT REPORT
Submitted to the Department of electronics &communication Engineering in the
FACULTY OF ENGINEERING & TECHNOLOGY
In partial fulfillment of the requirements for the award of the degree
Of
MASTER OF TECHNOLOGY
IN
ELECTRONICS &COMMUNICATION ENGINEERING
APRIL 2016

CELL: +91 9894917187 | 875487 1111 / 2111 / 3111 / 4111 / 5111 / 6111
ECWAY TECHNOLOGIES
CERTIFICATE
Certified that this project report titled “A Learning Framework for Age Rank Estimation
Based on Face Images With Scattering Transform” is the bonafide work of Mr.
_____________Who carried out the research under my supervision Certified further, that to the
best of my knowledge the work reported herein does not form part of any other project report or
dissertation on the basis of which a degree or award was conferred on an earlier occasion on this
or any other candidate.
Signature of the Guide Signature of the H.O.D
Name Name

CELL: +91 9894917187 | 875487 1111 / 2111 / 3111 / 4111 / 5111 / 6111
ECWAY TECHNOLOGIES
DECLARATION
I hereby declare that the project work entitled “A Learning Framework for Age Rank
Estimation Based on Face Images With Scattering Transform” Submitted to
BHARATHIDASAN UNIVERSITY in partial fulfillment of the requirement for the award of the
Degree of MASTER OF APPLIED ELECTRONICS is a record of original work done by me the
guidance of Prof.A.Vinayagam M.Sc., M.Phil., M.E., to the best of my knowledge, the work
reported here is not a part of any other thesis or work on the basis of which a degree or award
was conferred on an earlier occasion to me or any other candidate.
(Student Name)
(Reg.No)
Place:
Date:

CELL: +91 9894917187 | 875487 1111 / 2111 / 3111 / 4111 / 5111 / 6111
ECWAY TECHNOLOGIES
ACKNOWLEDGEMENT
I am extremely glad to present my project “A Learning Framework for Age Rank Estimation
Based on Face Images With Scattering Transform” which is a part of my curriculum of third
semester Master of Science in Computer science. I take this opportunity to express my sincere
gratitude to those who helped me in bringing out this project work.
I would like to express my Director,Dr. K. ANANDAN, M.A.(Eco.), M.Ed., M.Phil.,(Edn.),
PGDCA., CGT., M.A.(Psy.)of who had given me an opportunity to undertake this project.
I am highly indebted to Co-OrdinatorProf. Muniappan Department of Physics and thank from
my deep heart for her valuable comments I received through my project.
I wish to express my deep sense of gratitude to my guide
Prof. A.Vinayagam M.Sc., M.Phil., M.E., for her immense help and encouragement for
successful completion of this project.
I also express my sincere thanks to the all the staff members of Computer science for their kind
advice.
And last, but not the least, I express my deep gratitude to my parents and friends for their
encouragement and support throughout the project.

CELL: +91 9894917187 | 875487 1111 / 2111 / 3111 / 4111 / 5111 / 6111
ECWAY TECHNOLOGIES
ABSTRACT:
This paper presents a cost-sensitive ordinal hyperplanes ranking algorithm for human age
estimation based on face images. The proposed approach exploits relative-order information
among the age labels for rank prediction. In our approach, the age rank is obtained by
aggregating a series of binary classification results, where cost sensitivities among the labels are
introduced to improve the aggregating performance. In addition, we give a theoretical analysis
on designing the cost of individual binary classifier so that the misranking cost can be bounded
by the total misclassification costs. An efficient descriptor, scattering transform, which scatters
the Gabor coefficients and pooled with Gaussian smoothing in multiple layers, is evaluated for
facial feature extraction. We show that this descriptor is a generalization of conventional
bioinspired features and is more effective for face-based age inference. Experimental results
demonstrate that our method outperforms the state-ofthe-art age estimation approaches.

CELL: +91 9894917187 | 875487 1111 / 2111 / 3111 / 4111 / 5111 / 6111
ECWAY TECHNOLOGIES
INTRODUCTION:
Automatic age estimation, which involves evaluating a person’s exact age or age-group, is a
crucial topic in human face image understanding. The task of estimating exact human age adopts
a dense representation of the age labels (e.g., from 0 to 80), and the task of age-group estimation
divides the labels only into rough groups (e.g., elder, adult, and teenage/children). In this paper,
we focus on the setting of the former task that can be applicable to more general situations.
Nevertheless, the proposed method can be used for age-group estimation as well. Some video-
based age estimation approaches utilize temporal dynamic features, In this paper, we focus on
the study of image-based approaches. Two main components for building an effective age
estimator are facial feature extraction and estimator learning.
We discuss them Manuscript received June 8, 2014; revised October 20, 2014; accepted
December 11, 2014. Date of publication January 5, 2015; date of current version January 20,
2015. This work was supported by the National Science Council of Taiwan under Grant
MOST103-2221-E-001-010. The associate editor coordinating the review of this manuscript and
approving it for publication was Prof. Stefano Tubaro. K.-Y. Chang is with the Institute of
Information Science, Academia Sinica, Taipei 11529, Taiwan C.-S. Chen is with the Institute of
Information Science, Academia Sinica, Taipei 11529, Taiwan, and also with the Research Center
for Information Technology Innovation, Academia Sinica, Taipei 11529, Taiwan Color versions
of one or more of the figures in this paper are available online
Digital Object Identifier 10.1109/TIP.2014.2387379 briefly in the following to motivate our
approach. To learn an age estimator, most approaches formulate it as either a multi-class
classification problem, or a regression problem, Multi-class approaches simply treat the age
values as independent labels and learn a classifier to infer the person age. Many standard
approaches such as k Nearest Neighbors, Multilayer Perceptrons, Adaboost, Support Vector
Machine (SVM), can be employed to predict the specific age or age group. Regression
approaches basically learn a function that best fits the mapping from the feature space to the age-
value space with appropriate regularization. Typical nonlinear regression approaches such as
quadratic regression, Gaussian Process,and Support Vector Regression (SVR),have been used to
solve the age estimation problem as well. However, the classification approaches merely regard

CELL: +91 9894917187 | 875487 1111 / 2111 / 3111 / 4111 / 5111 / 6111
ECWAY TECHNOLOGIES
the age labels independent to each other but overlook the inter-relationship among the age
values.
Although the regression approaches take the inter-relationship into account, the labels are treated
as linearly increased values that cannot reflect the non-stationarity of human aging process
Instead of classification or regression, recent studies have proposed learning-to-rank approaches
to solve the age estimation problem recently. Unlike the classification approaches that treat the
labels naively as independent tags or the regression approaches that treat the labels as simply
proportional quantities, ranking approaches can be trained by adopting the ordering property of
the labels, thereby achieving superior performance.
The earliest ranking-based age estimation approach has been given in This study employed the
ranking method in for age rank estimation. Ranking (or ordinal regression) models are suitable
for age estimation because the relative ordering information among the age labels is employed
appropriately. However, the performance improvement is still limited because multiple
hyperplanes parallel to each other are used in a single kernel space for dense-labeled age
estimation. In a feature-selection approach has been proposed for the age-ranking method
introduced in These approaches use multiple hyperplanes in the feature or kernel spaces and
aggregate the hyperplane classification results to infer the age rank, which have been shown
effective for improving the age inference performance. In this paper, we introduce a learning-to-
rank approach for age estimation.
Our approach utilizes the relative order of age labels to conduct an effective age estimator.
Beside, we propose a cost-sensitive ordinal ranking framework and provide 1 a theoretical bound
guarantee that can be applied to common performance indices (such as mean absolute error
(MAE) and cumulative score (CS) for age estimation. In addition to learning, feature
representation is a critical problem when estimating age based on facial images. To derivefacial
aging features, early study,adopts skin wrinkles and distances between facial components (such
as eyes and noses), but such simplified features are inadequate for accurate age estimation.
To extract more details, active appearance model (AAM), learns both shape and appearance
models via principal component analysis (PCA), which becomes one of the most popular

CELL: +91 9894917187 | 875487 1111 / 2111 / 3111 / 4111 / 5111 / 6111
ECWAY TECHNOLOGIES
features for face-related tasks An evaluation of AAM can be found in.Besides the employment of
facial landmarks, some approaches build the feature vector based on appearances directly. For
example, Ahonen et al. and Zhou et al. use local binary patterns (LBP) and Haar-like wavelets as
feature representations to estimate the human age, respectively.
In addition, previous studies have developed dimension reduction models, such as manifold
learning ,to construct facial aging features. One of the most effective appearancebased features is
the Bio-inspired Features (BIF) of which is designed by simplifying the feed-forward model in
the ventral stream of primate visual cortex In this paper, we introduce the use of scattering
transform (ST) for human age estimation based on face images. In BIF, features are constructed
by convolving the input image with Gabor filters of different scales and orientations.
Subsequently, the pixels of each transformed image are pooled to form a feature vector. BIF has
shown their high efficacy for face-based age estimation.
However, pooling the convolved coefficients of Gabor-wavelets facilitates local translation
invariance and reduces texture detail. To address this problem, ST retains the highfrequency
information by a cascaded structure, which can recover the lost texture details so that the
discriminating capability is better preserved. In Section V, we will explain that BIF is analogous
to the first layer of ST though a slightly different pooling operator is adopted. We also compare
their performance to demonstrate the effectiveness of ST for face-based age estimation.
Characteristics of this work are summarized as follows.
1) We introduce an effective divide-and-conquer approach for ordinal regression, which divides
the age rank estimation problem into a set of cost-sensitive binary classification problems and
then the binary results areaggregated for rank inference. 2) We conduct a theoretical bound to
support our framework and explain why the ranking performance of the divide-and-conquer
approach can be improved whenmaking the binary classifiers better. 3) We give an insightful
interpretation of BIF by showing that it can be considered as the first layer of a more general
model, ST. 4) Our approach that employs ST in ranking inference can achieve state-of-the-art
performance on a large human-face age dataset.The rest of this paper is organized as follows.
Section II reviews previous studies on human age inference based on face images. Sections III

CELL: +91 9894917187 | 875487 1111 / 2111 / 3111 / 4111 / 5111 / 6111
ECWAY TECHNOLOGIES
outlines the proposed learning framework for age ranking. Section IV introduces our cost-
sensitive rank learning algorithm. Section V presents the feature representations employed in our
approach. Experimental results are shown in Section VI. Finally, conclusions are given in
Section VII.

CELL: +91 9894917187 | 875487 1111 / 2111 / 3111 / 4111 / 5111 / 6111
ECWAY TECHNOLOGIES
CONCLUSION:
We proposed an age ranking approach, CSOHR, for human face age estimation. The proposed
approach employs relative order information among age ranks, and utilizes cost sensitivity to
regulate a series of K binary classiﬁcations that is aggregated for rank inference. We also showed
a theoretical bound analysis of CSOHR. Additionally, a translationinvariant and deformation-
stable descriptor, ST, is used and evaluated for feature extraction of facial components.
Experimental results demonstrated that our learning framework, CSOHR, outperforms
conventional classiﬁcation, regression and ranking approaches. Besides, by combining CSOHR
and ST, the MAE of the large dataset, MORPH Album 2, can be reduced to 3.74 years for the
selected Caucasian subset and 3.82 years for the whole dataset, which are the best among the
associated results of current studies. Currently, we focus on age estimation from faces of nearly
neutral facial expressions, and evaluate our methods on the datasets (FG-NET and MORPH
Album 2) without serious facial expression variations. However, expression changes could affect
the age estimation results. Estimating both the age rank and facial expression intensity rank is a
possible way to solve this problem. Weplan to extend the proposed approach by transfer learning
like to tackle this problem in the future.

CELL: +91 9894917187 | 875487 1111 / 2111 / 3111 / 4111 / 5111 / 6111
ECWAY TECHNOLOGIES
REFERENCES:
[1] A. Hadid, “Analyzing facial behavioral features from videos,” in Human Behavior
Understanding. Berlin, Germany: Springer-Verlag,2011.
[2] H. Dibeklio˘glu, T. Gevers, A. A. Salah, and R. Valenti, “A smile can reveal your age:
Enabling facial dynamics in age estimation,” in Proc. 20th ACM Int. Conf. Multimedia, 2012,
pp. 209–218.
[3] A. Lanitis, C. Draganova, and C. Christodoulou, “Comparing different classiﬁers for
automatic age estimation,” IEEE Trans. Syst., Man, Cybern. B, Cybern., vol. 34, no. 1, pp. 621–
628, Feb. 2004.
[4] C. Li, Q. Liu, J. Liu, and H. Lu, “Learning ordinal discriminative features for age
estimation,” in Proc. IEEE Int. Conf. Comput. Vis.Pattern Recognit., Jun. 2012, pp. 2570–2577.
[5] C.-C. Wang, Y.-C. Su, C.-T. Hsu, C.-W. Lin, and H. Y. M. Liao,“Bayesian age estimation
on face images,” in Proc. IEEE Int. Conf.Multimedia Expo, Jun./Jul. 2009, pp. 282–285.
[6] X. Geng, Z.-H. Zhou, and K. Smith-Miles, “Automatic age estimation based on facial aging
patterns,” IEEE Trans. Pattern Anal. Mach. Intell.,vol. 29, no. 12, pp. 2234–2240, Dec. 2007.
[7] Y. Fu and T. S. Huang, “Human age estimation with regression on discriminative aging
manifold,” IEEE Trans. Multimedia, vol. 10, no. 4,pp. 578–584, Jun. 2008.
[8] B. Ni, Z. Song, and S. Yan, “Web image mining towards universal age estimator,” in Proc.
17th ACM Int. Conf. Multimedia, 2009, pp. 85–94.
[9] G. Guo, G. Mu, Y. Fu, and T. S. Huang, “Human age estimation using bio-inspired features,”
in Proc. IEEE Int. Conf. Comput. Vis. Pattern Recognit., Jun. 2009, pp. 112–119.
[10] G. Guo, Y. Fu, C. R. Dyer, and T. S. Huang, “Image-based human age estimation by
manifold learning and locally adjusted robust regression,” IEEE Trans. Image Process., vol. 17,
no. 7, pp. 1178–1188, Jul. 2008.
[11] Y. Zhang and D.-Y. Yeung, “Multi-task warped Gaussian process for personalized age
estimation,” in Proc. IEEE Int. Conf. Comput. Vis.Pattern Recognit., Jun. 2010, pp. 2622–2629.

A learning framework for age rank estimation based on face images with scattering transform

Empfohlen

Empfohlen

Weitere ähnliche Inhalte

Kürzlich hochgeladen

Kürzlich hochgeladen (20)

Empfohlen

Empfohlen (20)

A learning framework for age rank estimation based on face images with scattering transform