Comparative analysis of augmented datasets performances of age invariant face recognition models

Bulletin of Electrical Engineering and Informatics
Vol. 10, No. 3, June 2021, pp. 1356~1367
ISSN: 2302-9285, DOI: 10.11591/eei.v10i3.3020  1356
Journal homepage: http://beei.org
Comparative analysis of augmented datasets performances of
age invariant face recognition models
Kennedy Okokpujie1
, Etinosa Noma-Osaghae2
, Samuel Ndueso John3
, Charles Ndujiuba4
,
Imhade Princess Okokpujie5
1,2
Department of Electrical and Information Engineering, Covenant University, Ota, Nigeria
3
Department of Electrical and Electronic Engineering, Nigerian Defence Academy, Kaduna, Nigeria
4
Department of Electrical and Electronic Engineering, Air Force Institute of Technology, Kaduna, Nigeria
5
Department of Mechanical Engineering, Covenant University, Ota, Nigeria
Article Info ABSTRACT
Article history:
Received Aug 23, 2020
Revised Oct 12, 2020
Accepted Apr 14, 2021
The popularity of face recognition systems has increased due to their non-
invasive method of image acquisition, thus boasting the widespread
applications. Face ageing is one major factor that influences the performance
of face recognition algorithms. In this study, the authors present a
comparative study of the two most accepted and experimented face ageing
datasets (FG-Net and morph II). These datasets were used to simulate age
invariant face recognition (AIFR) models. Four types of noises were added to
the two face ageing datasets at the preprocessing stage. The addition of noise
at the preprocessing stage served as a data augmentation technique that
increased the number of sample images available for deep convolutional
neural network (DCNN) experimentation, improved the proposed AIFR
model and the trait aging features extraction process. The proposed AIFR
models are developed with the pre-trained Inception-ResNet-v2 deep
convolutional neural network architecture. On testing and comparing the
models, the results revealed that FG-Net is more efficient over Morph with
an accuracy of 0.15%, loss function of 71%, mean square error (MSE) of
39% and mean absolute error (MAE) of -0.63%.
Keywords:
Age invariant face recognition
Convolutional neural network
Data augmentation
Fg-Net dataset
Morph dataset
This is an open access article under the CC BY-SA license.
Corresponding Author:
Kennedy Okokpujie
Department of Electrical and Information Engineering
Covenant University
Km 10 Idiroko Road, Ota, Ogun State, Nigeria
Email: kennedy.okokpujie@covenantuniversity.edu.ng
1. INTRODUCTION
The paper aim at carrying out a comparative analysis of augmented datasets (FG-Net dataset and
morph datasets). Both performances (accuracy, loss function, mean square error (MSE), and mean absolute
error (MAE)) for trait-ageing invariant face recognition (AIFR) systems are compared. The significate of the
study is that both datasets are used for AIFR. Data augmentation via the addition of noises to both datasets at
the preprocessing phase greatly increases the accuracy and other parameters of AIFR.
- Literature review
Many comparisons exist in the literature between the performances of augmented datasets on age
invariant recognition systems. The augmented dataset is usually used independently of each other to verify
the invariability of designed face recognition systems. Two of the most common face image datasets used in
age-invariant face recognition, FG-NET and MORPH [1] are usually at the centre of comparisons made to
check the performance of age-invariant face recognition system. The goal is to have a good performance for

Bulletin of Electr Eng & Inf ISSN: 2302-9285 
Comparative analysis of augmented datasets performances of age invariant face… (Kennedy Okokpujie)
1357
all datasets used for face recognition. The results got from augmented standard datasets [2] for face images
are usually based on the robustness of designed face recognition system models to variations in pose,
illumination, shape, and texture. It is customary to have several ways of augmenting the datasets based on the
goal of the researcher and the challenge that needs to be solved.
Comparisons between the performance of augmented datasets on age-invariant face recognition
systems extend to niche applications like finding missing children who are discovered at a much later time
(longer than ten years) [3]. The importance of comparisons, especially for niche applications, is emphasized
in [4]. The factors that degrade the performance of face recognition systems are so numerous that it is
sensible to have as many augmented datasets from as many providers as possible. The abundant evidence of
the robustness of any age-invariant face recognition system is usually presented after it has passed the
rigorous condition of being subject to varieties of augmented datasets [5]. The evidence is generally in the
form of performance metrics like accuracy [6]. These performance metrics are used to gauge how well face
recognition systems can accurately recognize face images of various subjects regardless of the source of the
image, the noise added to the image and other forms of augmentation.
The region of the face used [7] to develop the age-invariant face recognition model plays a
significant role in the design of age-invariant face recognition systems that are robust. The region of the face,
when extracted from various datasets, could give non-identical performances on the designed face
recognition model. This submission extends to other face recognition models designed to checkmate the
negative effect of trait ageing. At the centre of comparisons of augmented datasets is accuracy [8], [9]. The
precision with which the designed face recognition model can identify subjects' facial image after they have
been designed to discriminate between real and generated images, estimate age and identify subjects. New
applications of age-invariant face recognition systems like soft biometrics [10] take the comparison between
augmented datasets seriously. The verification/identification process is thoroughly confirmed for as many
augmented datasets as possible to verify the accuracy of the face recognition system. The algorithms used to
develop age-invariant face recognition systems such as support vector machine (SVM) [11], principal
component analysis (PCA) and the like, perform differently for various forms of augmented datasets. The
authors in [12] tested the recognition system performance of a modelled age-invariant face recognition
system after passing face images through the designed and optimized adaptive neuro-fuzzy inference system
(ANFIS) classifier. The reviews made in [13] and [14] give in-depth studies of the performances of various
augmented datasets on designed age-invariant face recognition system. The studies focused on the challenges
of face recognition as it relates to the verification of designed face recognition systems using different
augmented datasets. The studies were able to identify the challenges faced by adaptive and age-invariant face
recognition systems through extensive and thorough comparisons using different augmented datasets.
- Fg-Net dataset specifications and complexities
There are 1002 images of 82 various persons with ages spanning from birth to 69 years in the FG-
NET database. The most common age group in the dataset is within the (<41 years) age group. Some of the
pictures of subjects in the FG-NET database were digitally taken recently while others were scanned copies
of the original photographs taken from personal collections of the subjects. The quality of the images in the
FG-NET database depends significantly on the skill of the photographer, the condition of the photograph, the
sophistication of the imaging tools used and the durability of the photographic paper found in personal
collections. Thus there are variations in sharpness, illumination, resolution, background, facial expression,
camera angles, and facial hair. These variations make the FG-NET database a good one for AIFR research
and samples of same subject (person) ranging from ages 2 to 43 is as shown in Figure 1.
Figure 1. Image age progression of Fg-Net dataset subject 1 at ages 1,5,8,10,14,16,……28,29,33,40,43
- Morph dataset II specifications and complexities
A longitudinal face database, MORPH Album 2 is a well-known publicly available dataset for face
recognition research. The face images in the MORPH database vary in age, sex, background. The MORPH

 ISSN: 2302-9285
Bulletin of Electr Eng & Inf, Vol. 10, No. 3, June 2021 : 1356 – 1367
1358
dataset was collected in uncontrolled environments (the pictures were taken in real-world conditions) and
thus has a very unique range of facial expressions. The photographs in the MORPH database were taken over
a period of four years and the database is regularly updated. MORPH Album 2 contains 55,134 face images
of 13,000 subjects along with metadata that shows that majority of the images were acquired in a period of
four years. Example images, age progression, and statistics of MORPH Album 2, as shown in Figure 2.
Figure 2(a) for white male and Figure 2(b) for african-american female.
(a) (b)
Figure 2. These figures are, (a) Image progression for white male, (b) Image progression for african-american
female
- Comparative analysis of Fg-Net ageing dataset and morph dataset II
Some of the remarkable dissimilarity between the Fg-Net and the Morph datasets is that children
dominate photos from Fg-Net. In contrast, most pictures from Morph are mainly from adult persons [15].
Also, the age gap between the images of the same subjects in the Fg-Net dataset is significantly wide-ranging
as compare to the once in Morph dataset, which is relatively small [16], as shown in Table 1. Besides, Fg-Net
contains subjects from one caucasian race, whereas Morph dataset contains the caucasoid, negroid, and
mongoloid races [17]. Furthermore, the total images (samples) in Fg-Net are 1002 with 82 subjects, while
that of Morph is 55,134 with 13,658 subjects [18]-[22], while details of both datasets are as shown in Table 2
and Table 3 and Table 4 depicts the Morph numbers of facial image and decade-of-life. However, the
Similarity is that both datasets contain face images of the same subjects at various age gaps. The sole reason
makes both ageing datasets and can be compared experimentally base on this fact [5]-[8].
Table 1. Comparison of FGNET and MORPH ageing datasets [23]
Database Images Subjects Age range Resolution
FG-NET ageing
database
1,002 high-resolution colour
or grey-scale
82 multiple race
subjects
0 to 69 years High 294 images of
females 430 images of males
High
MORPH database Album 1 1,724 face images 515 46 days to 29 years 240×200 pixels
Album 2 more than 20,000 4,000 16 to 77 years -
Table 2. Number of sample images in each age group of FG-NET dataset [24]
Age group (in years) Number of samples
0-9 371
10-19 339
20-29 144
30-39 79
40-49 46
50-59 15
60-69 8
Total 1,002
Table 3. Number of facial images based on gender and ethnicity from MORPH dataset [25]
African European Asian Hispanic Other Total
Male 36,832 7,961 141 1,667 44 46,645
Female 5,757 2,598 13 102 19 8,489
Total 42,589 10,559 154 1,769 63 55,134

1359
Table 4. Morph numbers of facial image and decade-of-life [26]
Age group (in years) Number of samples –Male Number of samples –Female Total Number of Samples
< 20 6,638 831 7,469
20-29 14,016 2,309 16,325
30-39 12,447 2,910 15,357
40-49 10,062 1,988 12,050
50+ 3,482 451 3,933
Total 46,645 8,489 55,134
2. RESEARCH METHOD
2.1. Pre-processing the FG-NET database for deep learning
A mammoth amount of data is needed to train a deep neural network. The FG-NET dataset has only
10-15 face images of each subject at different ages amounting to 1002 images. The size of the FG-NET
dataset is too small for deep neural network application. We preprocessed the images in the database by
adding noise to it. The addition of noise to the FG-NET dataset helped increase the total amount of pictures
available for deep learning application. The augmentation of the dataset was done at the preprocessing stage
to allow for improved feature extraction. The following steps were followed to augment the FG-NET dataset
with noise.
a. Convert all images to three channels with matrix entries for red, green an blue (RGB) for uniformity.
b. Viola-Jones face detector crops all face images and removes all background details from the face images
for richer feature extraction by the proposed deep learning model.
c. Five different versions of each image is created by the addition of four types of noise namely:
− No noise (original cropped image preserved)
− Poisson noise
− Salt and pepper noise
− Speckle noise
− Gaussian noise
The number of images available for deep learning experimentation was increased from 1002 to 5010
with up to about 45-90 images per subject. The addition of noise also helped with getting the deep neural
network to extract richer features from the face image for AFIR. Algorithm 1 shows the noise injection image
(data augmentation) procedures.
2.2. Pre-processing the morph database for deep learning
A mammoth amount of data is needed to train a deep neural network. The MORPH Album 2 dataset
has only 1-5 face images of each subject at different ages amounting to 13,000 images. The size of the is too
small for deep neural network application. We preprocessed the images in the database by adding noise to it.
The addition of noise to the MORPH Album 2 dataset helped increase the total amount of pictures available
for deep learning application. The augmentation of the dataset was done at the preprocessing stage to allow
for improved feature extraction. The following steps were followed to augment the MORPH Album 2 dataset
with noise:
a. Convert all images to three channels with matrix entries for red, green an blue (RGB) for uniformity.
b. Viola Jones face detector crops all face images and removes all background details from the face images
for richer feature extraction by the proposed deep learning model.
c. Five different versions of each image is created by the addition of four types of noise namely:
− No noise (original cropped image preserved)
− Poisson noise
− Salt and pepper noise
− Speckle noise
− Gaussian noise
The number of images available for deep learning experimentation was increased from 13,000 to
27,5000 with up to about 5-25 images per subject. The addition of noise also helped with getting the deep
neural network to extract richer features from the face image for AFIR.

 ISSN: 2302-9285
1360
2.3. Feature extraction and classification using convolutional neural network
Over a million images from the ImageNet database was used to train the Inception-ResNet-v2
convolutional neural network (CNN). The images that was used to train the Inception-ResNet-v2 CNN forms
part of the databased for the imagenet large-scale visual recognition challenge. Inception-ResNet-v2 has 164
layers and can classify images into 1000 object classes. The CNN accept images of size 299x299 for
classification. The Inception-ResNet-v2 was used in this study to learn features for age invariant face
recognition using a process called transfer learning. Transfer learning is the process of adapting a pre-trained
neural network for another task for which it was not originally trained. Transfer learning was used to learn
age invariant features from the FG-NET and MORPH datasets for AIFR. Figure 3 and Table 5 shows a
summary of the network architecture of Inception-ResNet-v2. In order to use the Inception-ResNet-v2
network, MATLAB R2018b was installed and downloaded the installer of the deep learning toolbox model
for Inception-ResNet-v2 network from [27]. Run the installer to install the Inception-ResNet-v2 network in
MATLAB R2018b.

1361
Figure 3. Block diagram of the adapted Inception-Resnet-V2 architecture network [28]
Table 5. Architecture of adapted Inception-Resnet-V2 network [29]
Block Type Repeat Depth Filter/Stride Output size Branch 1 Branch 2 Branch 3
1 Convolution 3×3/2 149×149×32 (32)
1 Convolution 3×3/1 147×147×32 (32)
1 Convolution 3×3/1 147×147×64 (64)
1 Max Pooling 3×3/2 73×73×160
1 Convolution 3×3/2 73×73×160 (96)
1 Convolution 3x3/1 71×71×192 (64, 96) (64, 64,64,96)
1 Convolution 3×3/2 35×35×384 (192)
1 Max Pooling 3x3/2 35×35×384
2 Inception-A 5 3 35x35x256 (32)
(32, 32/2)
(32,48,64/2)
3 Reduction-A 1 3 17x17x256 (384) (256,256,384)
4 Inception-B 10 3 17X17X896 (192) (128,160,192)
5 Reduction-B 1 3 8x8x1792
(256,384/2)
(256,288/2)
(256,288,320/2)
6 Inception-C 5 3 8x8x1792 (192) (192,224,256)
7
Average
Pooling
8×8 1792
8 Dropout Keep 0.8 1792
9 Softmax Classifier 82/13000
2.4. Training the deep learning model
The preprocessed FG-NET dataset was used to retrain the Inception-ResNet-v2 neural network for
AIFR. The process was possible via Transfer learning. The transfer learning process is enumerated below:
a. The preprocessed FG-NET images are loaded into MATLAB using the image datastore object.
b. The images are then splited into a validation set (20% images) and a train set (80%).
c. The images in the train set are resized to 299x299 for compatipility with Inception-ResNet–v2.
d. Inception-ResNet_v2 is run by MATLAB.
e. Training preferences are specified.
f. Begin the transfer learning process using the augmented FG-NET dataset.
g. Check the validity of the transfer learning process using the validation set.
h. Estimate the network’s accuracy.
2.5. Using the trained deep learning model for face recognition in FG-NET and Morph datasets
The retrained neural network was used for testing images from the Morph Album 2 dataset using the
following process:
a. Image is read from the MORPH Album 2 dataset.
b. All images are converted into an RGB matrix
c. Viola-Jones algorithm is used to cop and detect faces.
d. All images are resized to 299x299, resize the image to 299x299.
e. The retrained Inception-ResNet-v2 neural network is loaded.
f. Load the image into the retrained neural network for classification.
g. Compare predicted class to the ground truth.

 ISSN: 2302-9285
1362
3. RESULTS AND DISCUSSION
3.1. Evaluation methodology
MAE, MSE, Accuracy, and Loss function were used to check the performance of the proposed
AIFR model [30]-[35].
3.1.1. Accuracy
Accuracy is derived from the true positive (TP), true negative (TN), false positive (FP), and false
negative (FN) values as shown in (1). True positives are correct positive classifications. True negatives are
correct negative classifications. False positives are wrong positive classifications and false negatives are
wrong negative classifications.
Accuracy =
TP+TN
TP+TN+FP+TN
∗ 100% (1)
3.1.2. Mean squared error
The mean squared error (MSE) is a predictor value that is always positive. A score closer to zero
better. Where, N, in this instance, is the sums of iteration,f_i is the training loss values and y_i is the testing
loss values. Consequently, MSE is calculated, as presented in (2) [36], [37].
𝑀𝑆𝐸 =
1
𝑁
∑ (𝑓𝑖 − 𝑦𝑖)2
𝑁
𝑖=1 (2)
The MSE is the average (
1
𝑁
∑ .
𝑁
𝑖=1 ) of the squares of the inaccuracies (𝑓𝑖 − 𝑦𝑖)2
.
3.1.3. Mean absolute error
The mean absolute error (MAE) is a measure of the disparity between two values. In this
circumstance between y_i which is the values of training loss and y ̂_i, which is the value of the testing loss,
n is the sums of iteration. Consequently, MAE is calculated, as presented in (3) [38].
MAE =
∑ |𝑦𝑖−𝑦
̂𝑖|
𝑛
𝑖=1
𝑛
(3)
The MAE is the mean of the total errors (|𝑦𝑖 − 𝑦
̂𝑖|).
3.1.4. Loss function
Categorical cross-entropy is a loss function used to calculate the variation concerning two
probability disseminations. This dissimilarity is computed for respectively iteration in the training and testing
dataset. The technique to calculate the likelihood variation is as shown in the (4) [39].
ℒ𝑐𝑟𝑜𝑠𝑠(𝑦,ŷ) = − ∑ .
𝑁
𝑡=1 ∑ 𝑦𝑖
𝑡
. log(ŷ𝑖
𝑡
)
𝐶
𝑖=1 (4)
Where 𝑥 is the input value, 𝑦 is the true value, ŷ is forecast value by the method, 𝑁 is the sum of iteration and
𝐶 is the sum of class labels. Wen et al. [40] recommended a loss function called centre loss in adding to using
the definite cross-entropy loss. The idea is to growth the discriminative power of the completely learned
features by declining the intra-class variations. The centre loss function is
as shown in (5).
ℒ𝑐𝑒𝑛𝑡𝑒𝑟(𝑦,ŷ) =
1
2
∑ .
𝑁
𝑡=1 ∑ (ŷ𝑖
𝑡
− 𝑐𝑦𝑖
𝑡
)2
𝐶
𝑖=1 (5)
While 𝑐𝑦𝑖
is the 𝑦𝑖𝑡ℎ class centre of the features and 𝑁 is the sum of iterations. Wen et al. [40] detected that
(5) seen not accomplish the expected result. Two modifications were done by Wen et al. [40] to decide this
problem. First, the modification is to bring up to data the centers founded on a mini-batch as a additional for
the entire dataset. For the second modification is the institution of two new variables 𝛼 and the 𝛿 −
𝑓𝑢𝑛𝑐𝑡𝑖𝑜𝑛. α is used to regulate the learning rates of the centre, and the δ-function is a Boolean that results in
1 if the situation is true and 0 if the situation is false. In (6) defines the updated function of the class centre.
Δ𝑐𝑗 (𝑦, ŷ) =
∑ 𝛿(𝑦𝑖
𝑁
𝑡=1 =𝑗).(𝑐𝑗−ŷ𝑖)
1+∑ 𝛿(𝑦𝑖=𝑗)
𝑁
𝑡=1
(6)

1363
The novel centre of each class is as shown in (7):
𝑐𝑗
𝑡+1
= 𝑐𝑗
𝑡
− 𝛼. Δ𝑐𝑗
𝑡
(7)
While 𝛼 ∈ [0, 1]. Wen et al., [40] introduce λ to balance the two-loss functions of the total loss function.
The complete function is shown in (8).
ℒ = ℒ𝑐𝑟𝑜𝑠𝑠 + 𝜆ℒ𝑐𝑒𝑛𝑡𝑒𝑟 (8)
In the event 𝜆 is set to 0, the total loss function is equal to the categorical cross-entropy function is used.
3.2. Results and discussion
This section deal wth the results and comparative analysis of augmented datasets (FG-Net and
Morph II) performances for trait-ageing invariant face fecognition system. Figure 4 shows FG-Net and
Morph datasets training, and testing accuracies results in comparative analysis. With FG-Net dataset
outperforming the Morph dataset with a mean testing accuracy of 0.15%. While Figure 5 shows FG-Net and
Morph training and testing loss (error) results in comparative analysis. With mean FG-Net dataset output
performance, the Morph dataset testing loss of 71%. Table 6 shows a summary of the result performance of
augmented datasets of Fg-Net and Morph dataset. All this implied that FG-Net dataset have will perform
better than Morph dataset during deployment of these model in age invariant face recognition (AIFR) system
Table 6. The performance of augmented datasets (Fg-Net and Morph II)
Variable Fg-Net Dataset Morph Dataset Percentage Difference
Accuracy (Testing) 99.94% 99.79% 0.15%
Loss Function (Testing) 0.0039% 0.0067% 71%
Mean Square Error (MSE) 0.0155 0.0094 39%
Mean Absolute Error (MAE) 0.0634 0.0638 -0.63%
Furthermore, Figure 4 emphasize in graphical form the characteristics of FG-Net and morph training
and testing accuracies results in comparative analysis. While Figure 5 highlights in graphical form the
attributes of FG-Net and Morph training and testing loss (error function) results in comparative analysis.
Figure 6 in graphical form the characteristics of FG-Net and morph squared error results comparative
analysis. Finally, Figure 7 in graphical form the attributes of FG-Net and morph absolute error results
comparative analysis.
Figure 4. FG-Net and morph training and testing accuracies results comparative analysis (the result is best
viewed in colour)

 ISSN: 2302-9285
1364
Figure 5. FG-Net and morph training and testing loss (error function) results in comparative analysis
(the result is best viewed in colour)
Figure 6. FG-Net and morph squared error results comparative analysis (the result is best viewed in colour)
Figure 7. FG-Net and Morph absolute error results comparative analysis (the result is best viewed in colour)

1365
4. CONCLUSION
This paper compared two of the most acceptable and experimented face ageing dataset (FG-NET
and Morph II). These datasets were used to simulate age invariant face recognition (AIFR) models. The
obtained results show that FG-Net and Morph datasets are similar, and the little difference may be due to
randomness for augmenting the dataset.
ACKNOWLEDGMENTS
This paper is sponsor by Covenant University Center of Research, Innovation, and Discovery
(CUCRID) Covenant University, Ota, Ogun State, Nigeria.
REFERENCES
[1] S. Riaz, Z. Ali, U. Park, J. Choi, I. Masi, and P. Natarajan, "Age-invariant face recognition using gender-specific
3D ageing modeling," Multimed. Tools Appl., vol. 78, no. 17, pp. 25163–25183, 2019, DOI:10.1007/s11042-019-
7694-1.
[2] A. Yousaf, M. J. Khan, M. J. Khan, A. M. Siddiqui, and K. Khurshid, "A robust and efficient convolutional deep
learning framework for age‐invariant face recognition," Expert Syst., vol. 37, no. 3, 2019, doi: 10.1111/exsy.12503.
[3] D. Deb, D. Aggarwal, and A. K. Jain, "Finding Missing Children: Ageing Deep Face Features," arXiv Prepr., vol.
arXiv:1911, 2019.
[4] S. Anwarul and S. Dahiya, "A Comprehensive Review on Face Recognition Methods and Factors Affecting Facial
Recognition Accuracy," in Proceedings of ICRIC 2019: Springer, pp. 495-514, 2020, DOI:10.1007/978-3-030-
29407-6_36.
[5] Y. Wu, H. Hu, and H. Li, "Age-Invariant Face Recognition Using Coupled Similarity Reference Coding," Neural
Process. Lett., vol. 50, no. 1, pp. 397-411, 2019, doi:10.1007/s11063-018-9930-5.
[6] A. Caroppo, A. Leone, and P. Siciliano, "Facial Expression Recognition in Ageing Adults: A Comparative Study,"
Ital. Forum Ambient Assist. Living, Springer, pp. 349–359, 2018, doi: 10.1007/978-3-030-05921-7_29.
[7] K. K. Kamarajugadda and T. R. Polipalli, "Age-invariant face recognition using multiple descriptors along with
modified dimensionality reduction approach," Multimed. Tools Appl., vol. 78, no. 19, pp. 27639–27661, 2019, doi:
10.1007/s11042-019-7741-y.
[8] S. Zhao, J. Li, and J. Wang, "Disentangled representation learning and residual GAN for age-invariant face
verification," Pattern Recognit., vol. 100, 2019, doi: 10.1016/j.patcog.2019.107097.
[9] C. Atuegwu, S. A. Daramola, K. O. Okokpujie, and E. Noma-Osaghae. "Development of an Improved Fingerprint
Feature Extraction Algorithm for Personal Verification," International Journal of Applied Engineering Research,
vol. 13, no. 9, pp. 6608-6612, 2018.
[10] E. S. Jaha, "Augmenting Gabor-based Face Recognition with Global Soft Biometrics," 2019 7th International
Symposium on Digital Forensics and Security (ISDFS), Barcelos, Portugal, 2019, pp. 1-5, doi:
10.1109/ISDFS.2019.8757553.
[11] P. Panchal and H. Mewada, "Robust Illumination and Pose Invariant Face Recognition System using Support
Vector Machines," Int. J. Appl. Eng. Res., vol. 13, no. 16, pp. 12689-12701, 2018.
[12] S. A. Padmanabhan and N. K. Kumar, "An effective pose invariant face recognition system with the aid of ABC
optimized ANFIS," Cluster Comput., vol. 22, no. 1, pp. 177–187, 2019, doi: 10.1007/s10586-018-2032-3.
[13] S. B. Ahmed, S. F. Ali, J. Ahmad, M. Adnan, and M. M. Fraz, "On the frontiers of pose invariant face recognition:
a review," Artif. Intell. Rev., pp. 2571-2634, 2020, doi: 10.1007/s10462-019-09742-3.
[14] S. Singh and S. Prasad, "Techniques and Challenges of Face Recognition: A Critical Review," Procedia Comput.
Sci., vol. 143, pp. 536–543, doi: 10.1016/j.procs.2018.10.427.
[15] J. Zeng, H. Ling, L. J. Latecki, S. Fitzhugh, and G. Guo, "Analysis of Facial Images across Age Progression by
Humans," ISRN Mach. Vis., vol. 2012, pp. 1–7, 2012, doi: 10.5402/2012/505974.
[16] G. Panis and A. Lanitis, "An Overview of Research Activities in Facial Age Estimation Using the FG-NET Aging
Database," IET Biometrics, vol. 5, no. 2, pp. 37–46, 2015, doi:10.1049/iet-bmt.2014.0053.
[17] P. Grd and M. Baca, "Creating a face database for age estimation and classification," 2016 39th Int. Conv. Inf.
Commun. Technol. Electron. Microelectron. MIPRO 2016-Proc., pp. 1371–1374, 2016, doi:
10.1109/MIPRO.2016.7522353.
[18] B. Yip et al., "Preliminary Studies on a Large Face Database," 2018 IEEE International Conference on Big Data
(Big Data), Seattle, WA, USA, 2018, pp. 2572-2579, doi: 10.1109/BigData.2018.8622432.
[19] M. S. Shakeel and K. M. Lam, "Deep-feature encoding-based discriminative model for age-invariant face
recognition," Pattern Recognit., vol. 93, pp. 442–457, 2019, doi: 10.1016/j.patcog.2019.04.028.
[20] S. Zaghbani, N. Boujneh, and M. S. Bouhlel, "Age estimation using deep learning," Comput. Electr. Eng., vol. 68,
pp. 337–347, 2018, doi: 10.1016/j.compeleceng.2018.04.012.
[21] H. El Khiyari and H. Wechsler, "Face Recognition across Time Lapse Using Convolutional Neural Networks," J.
Inf. Secur., vol. 7, no. 3, pp. 141–151, 2016, doi: 10.4236/jis.2016.73010.
[22] H. Zhou and K. M. Lam, "Age-invariant face recognition based on identity inference from appearance age," Pattern
Recognit., vol. 76, pp. 191–202, 2018, doi: 10.1016/j.patcog.2017.10.036.
[23] M. Nimbarte and K. K. Bhoyar, "Distinctive feature representations and classifiers for age invariant face
recognition-a survey," Int. J. Comput. Syst. Eng., vol. 2, no. 4, pp. 222, 2016, doi:10.1504/IJCSYSE.2016.081391.

 ISSN: 2302-9285
1366
[24] “The FG-NET Aging Database.” [Online]. Available: http://www.fgnet.rsunit.com. [Accessed: 21-Dec-2019].
[25] “MORPH Non-Commercial Release Whitepaper.” [Online]. Available: http://www.faceaginggroup.com.
[Accessed: 21-Dec-2019].
[26] “Morph Dataset II.” [Online]. Available: http://people.uncw.edu/vetterr/MORPH-NonCommercial-Stats.pdf.
[Accessed: 24-Dec-2019].
[27] MathWorks, “Inception-ResNetv2.” [Online]. Available:
https://www.mathworks.com/matlabcentral/fileexchange/66607-deep-learning-toolbox-model-for-inception-resnet-
v2-network. [Accessed: 02-Oct-2019].
[28] K. Okokpujie, S. John, C. Ndujiuba, and E. Noma-Osaghae, "Development of an Adaptive Trait-Aging Invariant
Face Recognition System Using Convolutional Neural Networks," in Information Science and Applications:
Springer, pp. 411-420, 2020, doi:10.1007/978-981-15-1465-4_41.
[29] K. Okokpujie and S. Apeh, "Predictive Modeling of Trait-Aging Invariant Face Recognition System Using
Machine Learning," in Information Science and Applications: Springer, pp. 431–440, 2020, DOI:10.1007/978-981-
15-1465-4_43.
[30] D. Wang, C. Otto, S. Member, and A. K. Jain, "Face Search at Scale : 80 Million Gallery," MSU Tech. Rep. MSU-
CSE-, pp. 1–14, 2015.
[31] C. Chen and A. Ross, "Matching thermal to visible face images using hidden factor analysis in a cascaded subspace
learning framework," Pattern Recognit. Lett., vol. 72, pp. 25-32, 2016, doi: 10.1016/j.patrec.2015.06.021.
[32] S. Banerjee, W. J. Scheirer, K. W. Bowyer and P. J. Flynn, "Fast Face Image Synthesis With Minimal
Training," 2019 IEEE Winter Conference on Applications of Computer Vision (WACV), Waikoloa, HI, USA, 2019,
pp. 2126-2136, doi: 10.1109/WACV.2019.00230.
[33] B. F. Klare and A. K. Jain, "Heterogeneous Face Recognition Using Kernel Prototype Similarities," in IEEE
Transactions on Pattern Analysis and Machine Intelligence, vol. 35, no. 6, pp. 1410-1422, June 2013, doi:
10.1109/TPAMI.2012.229.
[34] Y. Li, G. Wang, L. Nie, Q. Wang, and W. Tan, "Distance metric optimization driven convolutional neural network
for age invariant face recognition," Pattern Recognit., vol. 75, pp. 51–62, 2018, doi: 10.1016/j.patcog.2017.10.015.
[35] F. B. Alvi and R. Pears, "A composite spatio-temporal modeling approach for age invariant face recognition,"
Expert Syst. Appl., vol. 72, pp. 383–394, 2017, doi: 10.1016/j.eswa.2016.10.042.
[36] Q. Zhu, P. Zhang, Z. Wang and X. Ye, "A New Loss Function for CNN Classifier Based on Predefined Evenly-
Distributed Class Centroids," in IEEE Access, vol. 8, pp. 10888-10895, 2020, doi:
10.1109/ACCESS.2019.2960065.
[37] K. O. Okokpujie, E. Noma-Osaghae, O. J. Okesola, S. N. John and O. Robert, "Design and Implementation of a
Student Attendance System Using Iris Biometric Recognition," 2017 International Conference on Computational
Science and Computational Intelligence (CSCI), Las Vegas, NV, USA, pp. 563-567, 2017, doi:
10.1109/CSCI.2017.96.
[38] R. Angulu, J. R. Tapamo, and A. O. Adewumi, "Age estimation via face images : a survey," EURASIP Journal on
Image and Video Processing, vol. 1, 2018, DOI:10.1186/s13640-018-0278-6.
[39] K. Okokpujie, E. Noma-osaghae, O. Okesola, O. Omoruyi, Osemwegie Chinonso, S. John, and I. P. Okokpujie,
"Fingerprint Biometric Authentication Based Point of Sale Terminal," in Kim K., Baek N. (eds) Information
Science and Applications 2018. ICISA 2018. Lecture Notes in Electrical Engineering, Springer, Singapore,
vol 514, 2018, https://doi.org/10.1007/978-981-13-1056-0_24.
[40] Y. Wen, K. Zhang, Z. L. B, and Y. Qiao, "A Discriminative Feature Learning Approach for Deep Face
Recognition," in European conference on computer vision. Springer, Cham, 2016, vol. 1, pp. 499–515, 2016, doi:
10.1007/978-3-319-46478-7_31.
BIOGRAPHIES OF AUTHORS
Kennedy Okokpujie holds a Bachelor of Engineering (B.Eng.) in Electrical and Electronics
Engineering, Master of Science (M.Sc.) in Electrical and Electronics Engineering, Master of
Business Administration (MBA), Master of Engineering (M.Eng.) in Electronics and
Telecommunication Engineering and Ph.D in Information and Communication Engineering. Dr.
Kennedy Okokpujie is currently a Researcher/Lecturer with the Department of Electrical and
Information Engineering at Covenant University, Ota, Ogun State, Nigeria. He is a member
numerous organizations including; The Nigeria Society of Engineers and the Institute of
Electrical and Electronic Engineers (IEEE) and reviews for numerous high impact Journals. His
research areas of interest include Biometrics, Artificial Intelligence, Communication
Engineering, Digital signal Processing, Cryptography and its applications, Network security and
Pravicy.

1367
Etinosa Noma-Osaghae is with the Department of Electrical and Information Engineering. He
has a master’s degree in Electronic and Telecommunication Engineering. Information Theory,
Signal Processing, Biometrics, E-learning, E-health and Internet of Things are his research areas
of interest.
Samuel Ndueso John is a Professor of Computer Systems and Network Engineering. He has his
higher education at Donetsk National Technical University, Ukraine where he successfully
defended and obtained B.Sc, M.SC, MPhil and Ph.D. degrees in Computer Systems and Network
Engineering, specializing in Computer Science, Computing Machines, Complex Systems,
Security and Networks in 1993, 1994, 2000 and 2005, respectively. Presently, John is a
Professor of Computer Systems and Network Engineering in the Department of
Electrical/Electronic Engineering, Faculty of Engineering and Technology, Nigerian Defence
Academy, Kaduna, Nigeria. John has acquired valuable knowledge and practical experience in
the use of information technology as an enabler of industrial and national development goals. He
has a vast knowledge of computing and has applied it in the pursuance of a wide range of
indigenous ICT Convergence, Data Efficiency Management, Cyber Security, Cybercrime and
Telemedicine solutions.
Charles U. Ndujiuba is a professor of Communication Engineering with Air Force Institute of
Technology Nigerian Air Force, Kaduna state, Nigeria. Prof. C. U. Ndujiuba holds a Ph.D. in
Electrical & Electronics Engineering from the University College London (University of
London); Master Specialize (Masters with Specialization) in Radio Communications from Ecole
Superieure d’Electricite (SUPELEC) France; MSc in Electrical Engineering from the University
of Lagos; BSc in Electronics & Communications Engineering from the London Metropolitan
University. Prof. Ndujiuba is a Chartered Electronics Engineer (CEng) and a highly skilled
wireless professional. He has more than 25 years of RF, Microwave, Fixed-line (SDH and PDH),
and PMR experience, with considerable international exposure. Prof. Ndujiuba has attended
several conferences and published many technical papers in major professional journals. Prior to
joining Covenant University Ota Nigeria in 2011, Ndujiuba was the Technical Director of Globe
Trunk Ltd UK. His research interests include Monolithic Microwave Integrated Circuits
(MMIC), Active Filters, Ultra-Low Noise Amplifiers, Active Devices and Circuits, UWB
Transmitter, Modelling & Simulation, Dielectric Resonator Antennas, and Detection and
Collision Avoidance of Unmanned Aerial Vehicles.
Engr. Dr. (Mrs) Okokpujie Imhade Princess is a researcher/lecturer in the Department of
Mechanical Engineering Covenant University, Ota, Ogun State Nigeria. She is currently the
Chief Editor to Covenant Journal of Engineering Technology (CJET). She is also a reviewer and
editor of many international/local journals and conferences. Her areas of research interest are
Design and Production, Advanced Manufacturing such as Machining, Tool Wear, Vibration,
Nano-lubricant, Energy Systems, Mathematical Modeling, Optimization, Mechatronics and also
a Multi-disciplinary Researcher. Dr. I. P. Okokpujie is an active researcher who has authored/co-
authored over 106 peer-reviewed publications in reputable journals and international
conferences. In 2017 to 2019, she was the technical secretary to the International Conference on
Engineering for a Sustainable World (ICESW) index in Scopus and ISI data based through the
IOPs publisher. She is a Registered Engineer of the Council for the Regulation of Engineering in
Nigeria (COREN), a member of the Nigerian Society of Engineers (NSE), member Nigerian
Institution of Mechanical Engineers and Association of Professional Women Engineers of
Nigeria (APWEN). She is currently the National Technical Secretary and Vice-Chairman of
APWEN Ota Chapter in Ogun State, also, the Technical Editor to Journal of Nigeria Institution
of Mechanical Engineers. She is one of the top-rated researchers in her institution, and she is
happily married and blessed with children. Dr. I. P. Okokpujie is very passionate about the
Education of the Girl Child and also offers quality mentorship to the Young Engineers.

Comparative analysis of augmented datasets performances of age invariant face recognition models

More Related Content

What's hot (20)

Similar to Comparative analysis of augmented datasets performances of age invariant face recognition models (20)

More from journalBEEI (20)

Recently uploaded (20)

Comparative analysis of augmented datasets performances of age invariant face recognition models