See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/370795714

Rock-Paper-Scissors Image Classification Using Transfer Learning

Conference Paper · March 2023

DOI: 10.1109/ICBATS57792.2023.10111433

CITATIONS READS

2 205

6 authors, including:

Fahad Ahmed Hamza Alrababah

University of South Asia Skyline University College
9 PUBLICATIONS 19 CITATIONS 10 PUBLICATIONS 39 CITATIONS

SEE PROFILE SEE PROFILE

Muhammad Farhan Khan

International Islamic University, Islamabad
22 PUBLICATIONS 441 CITATIONS

SEE PROFILE

All content following this page was uploaded by Fahad Ahmed on 26 May 2023.

The user has requested enhancement of the downloaded file.

 Rock-Paper-Scissors Image Classification Using
Transfer Learning
2023 International Conference on Business Analytics for Technology and Security (ICBATS) | 979-8-3503-3564-4/23/$31.00 ©2023 IEEE | DOI: 10.1109/ICBATS57792.2023.10111433

Fahad Ahmed
Department of Computer Science Wasim Ahmad Khan Muhammad Iqbal
National College of Business School of Computer Science, Department of Computer Science
Administration & Economics NCBA&E National University of Technology
Lahore, Pakistan Lahore, Pakistan (NUTECH)
fahadahmed4617@gmail.com Applied Science Research Center, Islamabad, Pakistan
Applied Science Private University, muhammad.iqbal@nutech.edu.pk
Ashraf riad Ahmad abazeed Amman 11937, Jordan, iqbalhaider.qau@gmail.com
Assistant professor, h_gaftim@asrc.asu.edu.jo
Abu Dhabi Vocational Education and wasimahmad.ucit@gmail.com Muhammad Farhan Khan
Training Institute, ADVETI, Department of Forensic Sciences,
Ashraf.abazeed@sts.adveti.ac.ae Hamza Alrababah University of Health Sciences
School of Information Technology, Lahore, Pakistan
Skyline University College, University mfarhankhan1987@uhs.edu.pk
City Sharjah, 1797, Sharjah, UAE
hamza.alrababah@skylineuniversity.ac.
ae

Abstract— Games are a genre of deliberate mental or Parents and governments must develop policies that enable
physical activities which can be done alone or in groups for the children to thrive during a pandemic. The only way to
purpose of enjoyment and recreation. Every game has its unique maintain social contact is via the internet. Children may form
collection of participants, movements, and tactics, in addition to networks with their peers over the internet, which is a novel
its own set of principles and guidelines. Rock-Paper-Scissors is
adaptation for everyone in this period. To avoid the harmful
probably the most well-known game for determining a winner
in the world. The rules are pretty straightforward. Rock, paper, effects of social distancing on children, social and networking
or scissors are displayed by each player simultaneously. The between children and their peers must remain active. They
rock defeats the scissors but is defeated by the paper. Scissors can play with each other through the internet. Rock-paper-
defeat paper but lose to rock and paper defeats rock but loses to scissor is a famous hand game for children. The hand game
scissors. Using hand images, this study proposed identifying rock-paper-scissors is played by two players [8]. The rules
images of hands playing rock-paper-scissors. Transfer learning are straightforward: "rock" trumps "scissors," "paper" trumps
is used in the proposed model to recognize and classify images "rock," and "scissors" trumps "paper." Although the game is
of rock-paper-scissors. The dataset for this study was acquired primarily popular among youngsters, it has also been utilized
from the Kaggle. There are 2188 images in the dataset, including
in art auctions [9], [10].
712 rock images, 726 paper images, and 750 scissor images. The
dataset is divided into two parts: 80% is utilized for training the The field of image processing has seen a significant
model, and 20% is utilized for model validation. The RPSICTL revolution as a result of artificial intelligence. Machine
model evaluation results good validation accuracy of 99.54%. learning approaches were utilized to evaluate and detect
images in image processing. Nevertheless, as technology
Keywords— Transfer learning (TL), AlexNet, Rock-Paper- advances, deep learning has demonstrated considerable gains
Scissors in image processing and image classification and
identification [11]. The significant benefit of convolutional
I. INTRODUCTION
neural network (CNN) is that the qualities of the objects do
Every facet of life in the planet as we know it has not have to be defined in advance; rather, the network learns
been irreversibly affected by the proliferation of this COVID- them on its own, without requiring a description of which trait
19 virus [1]. During this time, all governments' primary goal to search for precisely [12]. This article uses a unique
is to slow the massive expansion of COVID-19 throughout technology called transfer learning (TL) with a pre-trained
civilization [2]. Schools, colleges, universities, and other Alexnet to recognize and classify rock-paper-scissors images.
higher education institutions around the world have been AlexNet is a deep CNN structure suggested by Krizhevsky
closed as a result of the deadly COVID-19's spread [3]. and Sutskever [13], which debuted for the first time in the
Numerous students are required to attend school ImageNet LSVRC-2010 challenge and obtained astounding
from home in order to avoid a coronavirus epidemic. Physical results. In this article, rock-paper-scissors image dataset is
separation has varying effects on children depending on their used for the classification and identification, by utilizing deep
age [4]. Isolation impairs children's comprehension, learning. Samples of rock-paper-scissors images before pre-
confusion, and fear [5]. Covid-19 and social isolation are the processing are presented in fig. 1.
primary factors that contribute to the development of mental
health disorders in a significant number of persons [6], [7].

979-8-3503-3564-4/23/$31.00 ©2023 IEEE

Authorized licensed use limited to: KU Leuven Libraries. Downloaded on May 23,2023 at 19:16:15 UTC from IEEE Xplore. Restrictions apply.
 technique. The MLP classification's accuracy was also
evaluated using the cross validation approach.
Amnia Salma [22] investigated how to classify
images of hands playing rock-paper-scissors using a MLP.
Multilayer perceptron is the most often used model in back-
propagation neural network applications. MLP's architecture
Fig. 1. Image samples of rock-paper-scissors before pre-processing (a) is made up of three layers: an input layer, a hidden layer, and
Rock; (b) Paper; (c) Scissors
an output layer. A perceptron is fashioned after the
The remainder of the article is broken into the
fundamental unit of the human brain, the neuron, and
following sections: The section II describes the literature
possesses an incredible capacity for learning and problem
review, the material and methods are presented in section III,
solving. This article discusses how to use back propagation
the simulation and results are shown in section IV and the
in the MLP to identify hand playing pictures using hand
conclusion is presented in section V.
images. Ahmadi et al. [23] created RASA, an intelligent robot
that competed in a game of rock-paper-scissors against
humans. Using an MLP network and a Leap Motion sensor,
the created robot first recognizes the player's hand movement.
II. LITERATURE REVIEW He uses the Markov chain model in the second step to
Japan invented the game Jon Ken Pon in the 18th forecast the participant's forthcoming performance. Another
century, which is now known as rock-paper-scissors [14]. ingenious robot was created by Pozzato et al. [24]. In the
Another method for developing a smart rock-paper-scissors game of rock-paper-scissors, the robot uses the Gaussian
player was suggested by De Souza et al. [15]. In the rock- mixture model method to forecast the player's future
paper-scissors game, Cenggoro et al. [16] proposed using a performance.
feed-forward neural network to forecast a participant's future
performance. For this article, major contributions are as follow:
Salvetti et al. [17] introduced a new technique 1. By using this proposed RPSICTL model, rock-paper-
depending on entropy and LLE1 indications for predicting scissors classes have been predicted.
when to employ rock, paper, or scissors during the game. The 2. Accuracy, misclassification rate, precision, sensitivity,
opportune instant has arrived when the opponent's future specificity, false positive rate (FPR), false negative rate
conduct is further foreseeable. After that, the wise participant (FNR), and F1 score are among the evaluation criteria that
decides on how to beat the rival participant. Ghasemi et al. have shown good results.
[18] demonstrated an effective technique for identifying 3. Proposed RPSICTL model is quick does not need any
human behavioral patterns in the rock-paper-scissors game handmade features to handle small datasets.
by utilizing a multilayer perceptron neural network (MLP). 4. Finally, proposed RPSICTL model is compared with
The study's findings on 40 participants demonstrated the literature methodology Implementation of Multilayer
MLP network's superior effectiveness in recognizing patterns Perceptron for Image Classification [22] . RPSICTL model
of humanoid conduct. For the game of rock-paper-scissors, have better accuracy and less misclassification rate as
Matsumoto et al. [19] developed an automated judgment compared to this methodology.
system. Their study estimates the numbers and placements The Rock-Paper-Scissors image classification method using
of the hands in rock-paper-scissors images using the C-Means DNA level scrambling [29], a hyperchaotic system, DNA
classification approach and then makes winning and losing encoding [30], and blockchain technology [31] involves
judgments based on the difference in the compression of the converting the images to DNA sequences, scrambling them
categories. Additionally, studies have been conducted to using a hyperchaotic system, and encoding them before
identify aberrant hand positions. storing them in a secure blockchain database. Machine
Chen et al. [20] investigated this using image learning algorithms [32-35] can be trained on these encoded
processing. The hand image is first taken from sequential digital signatures to recognize the patterns associated with the
images and then utilized to generate a binary image with images, leading to a more accurate and reliable image
intervention, total cutoff, and skin tone, removed. Their classification process. Ensemble learning [36] and machine
primary objective in this piece was to recognize the scissors learning techniques [37-40] can be used to classify images of
due to successive errors involved, but recognizing rock and the Rock-Paper-Scissors game, achieving high accuracy and
paper is relatively straightforward. Rock is represented by the improving the robustness of the model.
number zero fingers, while paper is represented by the Secure communication [49] protocols and mobile agent
number five fingers. There are eight alternative arrangements protocols [50] can be used in conjunction with machine
for two fingers, but only one is accurate. They used the learning [51] and artificial neural networks [52,53] to
distance between two fingers as a condition for recognizing enhance the privacy and efficiency of Rock-Paper-Scissors
the placement of the scissors in specific circumstances, and Image Classification.
they solved the issue using division, histograms, and the
angle criterion to avoid inaccuracy. Gang et al. [21] proposed
a technique for classifying electromyographic (EMG) data
using MLP in terms of hand positions. Electromyographics III. MATERIAL AND METHODS
are put to the He-zajac-levine bilinear activation model, and In recent years, several deep learning methods for
the result is then used as multilayer perceptron inputs in this image recognition and evaluation have been widely used in

Authorized licensed use limited to: KU Leuven Libraries. Downloaded on May 23,2023 at 19:16:15 UTC from IEEE Xplore. Restrictions apply.
image processing. Fig. 2 shows the proposed RPSICTL It has one input (data), five convolutional, seven
model at the application level. ReLU, two normalization, three pooling, three fully
Fig. 2. Proposed RPSICTL model for application-level connected, two dropout, one softmax (prob), and one output
Fig. 3. Detailed proposed RPSICTL model
layers. AlexNet is an eight-layer network with learnable
The proposed RPSICTL model accepts rock-paper- parameters that contains three fully connected layers and five
scissors images and uses deep learning techniques to detect convolutional layers with a combination of max pooling. The
and classify them. It has two layers: pre-processing and abbreviation ReLU stands for non-linear activation function
application. The training data, which comprised of images of in each layer. Images from the Pre-processed layer are read
rock-paper-scissors, was acquired using the Kaggle source, by the network's input layer. Pre-processing of images is a
and the data was collected in raw state. The raw data was crucial stage in obtaining appropriate datasets, and it can be
handled by the Pre-processing layer, which converted the accomplished in a variety of ways, such as by enhancing
image to RGB dimensions of 227x227x3. The application some image attributes or resizing images. Since images come
in a variety of sizes, resizing them is a necessary step. As a
layer imports and customizes the pre-trained model AlexNet
for transfer learning. If the learning conditions aren't met, the
model will need to be retrained; otherwise, the trained model
will be saved on the cloud. Fig. 3 shows a thorough
explanation of the proposed RPSICTL model
Images from people playing rock-paper-scissors are
captured during the validation phase and sent to the pre-
processing layer for image dimension alteration. Then result, images were downsized to 227x 227x3, where
proposed RPSICTL model imports cloud data to intelligently 227x227 indicates the input images' height and width, and 3
classify rock-paper-scissors images. denotes the number of channels. Fig. 5 shows rock-paper-
scissors image samples representing different classes after
pre-processing.
Fig. 5. Image samples of rock-paper-scissors after pre-processing (a) Rock;
(b) Paper; (c) Scissors
The network's fully connected layers understand the
A. Transfer Learning traits of rock-paper-scissors to classify images into specific
For various applications, various pre-trained and classes; whereas the initial Convolutional layers retrieve
deep learning algorithms are utilized. In this article, a deep generic features of images by applying filters like edge
learning-based network called AlexNet is employed for detection and maintaining the spatial correlation among
transfer learning with the goal of classifying rock-paper- pixels. This proposed model incorporates the AlexNet
scissors images. AlexNet is a pre-trained CNN model. network, which is a pre-trained CNN network that has had a
significant influence on current deep learning
applications.[26]. In this proposed methodology, a
customized version of the AlexNet model is utilized. AlexNet
layers (except the last three) are retrieved for the specific
motive of attaining TL. The fully connected layer, softmax
layer, and output classification layer are the three last layers
Transfer learning refers to the usage of a pre-trained model, of the architecture that have been updated in response to the
problem statement. Table I shows the updated network that
was utilized for transfer learning.

TABLE I. MODIFIED LAYERS OF ALEXNET

Layers Original Layer Modified Layer
23 1000 fully connected layers 3 fully connected layers
24 Prob (Softmax) Softmax
25 Classification Output Classification Output
(crossentropyex with (crossentropyex)
‘tench’ and 999 other
classes)

To implement softmax functions to the input,

softmax layers are utilized. During the training process, it was
discovered that a convolutional layer represents generic
and it's a concept that's becoming popular in deep learning image properties like feature extraction, whereas a fully
applications. The AlexNet algorithm was developed using connected layer learns class specific features to distinguish
over 15 million annotated images from the ImageNet datasets among classes. As a result, fully connected layers are updated
[25]. Fig. 4 depicts the AlexNet network's architecture. based on class-specific characteristics of rock-paper-scissors.
Learning rate, validation frequency, Number of
Fig. 4. AlexNet's architecture iterations per epoch, optimization algorithm, and Number of
epochs are some of the parameters that can be used as training

Authorized licensed use limited to: KU Leuven Libraries. Downloaded on May 23,2023 at 19:16:15 UTC from IEEE Xplore. Restrictions apply.
choices. 1e–4 was discovered to be the best learning rate to 2 13 25 227x22 1e − 4 25
train the network. Validation frequency and number of 7x3
iterations per epoch are 25 and 13 respectively. For training, 5 13 25 227x22 1e − 4 25
7x3
the stochastic gradient descent with momentum (SGDM) 10 13 25 227x22 1e − 4 25
optimization algorithm is utilized. The training was 7x3
conducted on a number of epochs, including 1, 2, 5, and 10, Fig. 6 shows the training progress plot on 10th epoch
and the optimum epoch was discovered to be 10. The with training parameters of the proposed RPSICTL model.it
algorithm was retrained multiple times to reach the accuracy also shows the training and validation behavior of the
threshold in order to get maximum accuracy. proposed RPSICTL model from 1 to 10th epoch.

B. Dataset
Rock-paper-scissors dataset was acquired from
kaggle source [27]. There are three classes of rock, paper, and
scissors in this image dataset. Table II shows the image
samples for each class.

TABLE II. DATASET OF ROCK-PAPER-SCISSORS

Layers Class Image Samples
23 Rock 726
24 Paper 712
25 Scissors 750

IV. SIMULATION AND RESULTS

Fig. 6. Training Progress Plot on 10th epoch
For simulation and results, MATLAB 2020a is As shown in table IV, multiple epochs like 1, 2, 5
being used. Using pre-trained AlexNet, the proposed and 10 were used to train the data. The best validation
RPSICTL model for image classification and detection was accuracy 99.54% for RPSICTL model was achieved on 10th
constructed. There are two phases to the proposed RPSICTL epoch. This proposed RPSICTL model has a higher accuracy
model image dataset: training and validation. For the training than the prior methodology [22], which had 95% accuracy.
and validation phases, the dataset was divided into 80 % and
20%, respectively. Accuracy, misclassification rate, TABLE IV. PROPOSED RPSICTL MODEL VALIDATION WITH DIFFERENT
precision, sensitivity, specificity, FPR, FNR, and F1 score are EPOCHS

among the various performance indicators used to assess No. of Accuracy (%) Misclassification rate (%)
epochs
performance. Different evaluation metrics can be used to 1 35.01 64.99
evaluate the results achieved from the proposed RPSICTL 2 98.86 1.14
model [28]. 5 99.31 0.69
( ⁄ ) ( ⁄ ) 10 99.54 0.46
Accuracy = ∗ 100 (1)
( # ⁄ #) ( $ ⁄ $) The proposed RPSICTL model's confusion matrix
Misclassification Rate = ∗ 10 (2)
( ⁄ )
for validation on the 10th epoch is shown in table V. Total 437
Percision = ⁄
∗ 100 (3) images were used during validation. For the classification of
( ) ( # ⁄ #)
( ⁄ ) rock images, all 145 images were correctly identified as rock.
Sensitivity = ⁄
∗ 100 (4) For the classification of paper images, 140 images were
( ) ( $ ⁄ $)
( ⁄ ) correctly identified as paper while two images were
Specificity = ( ∗ 100 (5)
⁄ ) ( # ⁄ #) incorrectly identified as rock and scissors respectively. For
( # ⁄ #)
FPR = ∗ 100 (6) the classification of scissors images, all 150 images were
( # ⁄ #) ( ⁄ )
( $ ⁄ $) correctly identified as scissors.
FNR = ( ∗ 100 (7)
$⁄ $) ( + )
,∗(-./012134∗ .421516157) TABLE V. PROPOSED RPSICTL MODEL CONFUSION MATRIX WITH
F1 Score = (8) VALIDATION
-./012134 .421516157
Ntotal=437 Rock Paper Scissors

The proposed RPSICTL model classifies images Rock 145 1 0

into rock-paper-scissors. Different parameters like no. of Paper 0 140 0
epochs, no. of iterations per epoch, frequency, input image Scissors 0 1 150
size, no. of layers and learning rate are shown in table III for
the training of proposed RPSICTL model.
Table VI shows Statistical validation measures of
TABLE III. PARAMETERS OF SIMULATION proposed RPSICTL model including accuracy,
No. of No. of Frequency Input Learning No. of misclassification rate, precision, sensitivity, specificity, FPR,
epochs iterations Image rate layers FNR, and F1 score.
per epoch size
1 13 25 227x22 1e − 4 25 TABLE VI. STATISTICAL VALIDATION MEASURES OF PROPOSED RPSICTL
7x3 MODEL
Performance Parameters Validation

Authorized licensed use limited to: KU Leuven Libraries. Downloaded on May 23,2023 at 19:16:15 UTC from IEEE Xplore. Restrictions apply.
 Accuracy 99.54% Available: http://arxiv.org/abs/2201.03791.
Misclassification Rate 0.46% [13] A. Krizhevsky, I. Sutskever, and G. E. Hinton, “ImageNet
Precision 99.53% Classification with Deep Convolutional Neural Networks,” Int. Conf.
Sensitivity 99.55% Neural Inf. Process. Syst., pp. 1097–1105, 2012.
Specificity 99.78% [14] Y. Hasuda, S. Ishibashi, H. Kozuka, H. Okano, and J. Ishikawa, “A
FPR 0.22% robot designed to play the game ‘Rock, Paper, Scissors,’” IEEE Int.
FNR 0.45% Symp. Ind. Electron., pp. 2065–2070, 2007.
F1 Score 0.995 [15] D. F. P. De Souza, H. C. C. Carneiro, F. M. G. Franca, and P. M. V.
Lima, “Rock-paper-scissors WiSARD,” Proc. - 1st BRICS Ctries.
The proposed RPSICTL model is compared to the Congr. Comput. Intell. BRICS-CCI 2013, no. September, pp. 178–182,
literature in Table VII and it shows the higher accuracy and 2013.
less misclassification rate. [16] T. W. Cenggoro, A. H. Kridalaksana, E. Arriyanti, and M. I. Ukkas,
“Recognition of a human behavior pattern in paper rock scissor game
using backpropagation artificial neural network method,” 2014 2nd Int.
TABLE VII. PROPOSED RPSICTL MODEL COMPARISON WITH LITERATURE
Conf. Inf. Commun. Technol. ICoICT 2014, pp. 238–243, 2014.
Author Technique Validation
[17] F. Salvetti, P. Patelli, and S. Nicolo, “Chaotic time series prediction for
Accuracy (%) Misclassification rate (%)
the game, Rock-Paper-Scissors,” Appl. Soft Comput. J., vol. 7, no. 4,
Amnia Multilayer 95 5
pp. 1188–1196, 2007.
Salma [22] Perceptron
[18] M. Ghasemi, G. H. Roshani, and A. Roshani, “Detecting Human
Proposed Alexnet 99.54 0.46 Behavioral Pattern in Rock, Paper, Scissors Game Using Artificial
RPSICTL Intelligence,” Comput. Eng. Phys. Model., vol. 3, no. 1, pp. 25–35,
2020, [Online]. Available:
http://www.jcepm.com/article_104442.html.
V. CONCLUSION [19] Y. Matsumoto, T. Yamamoto, K. Honda, A. Notsu, and H. Ichihashi,
“Application of cluster validity criteria to Rock-Paper-Scissors game
Since the dawn of time, games have played a judgment,” IEEE Int. Conf. Fuzzy Syst., 2012.
significant role in human life. Rock-paper-scissors, among all [20] W. Y. Chen, C. H. Chung, S. Y. Heish, and C. C. Ku, “Rock, paper and
games, has a pretty straightforward ambiance and is regarded scissors image identification scheme,” 4th Int. Conf. New Trends Inf.
a popular and traditional game that almost everyone of any Sci. Serv. Sci. IEEE, United States, pp. 748–753, 2010.
[21] T. Gang, Y. Cho, and Y. Choi, “Classification of rock-paper-scissors
age can enjoy. Transfer learning is used to detect and classify using electromyography and multi-layer perceptron,” 2017 14th Int.
rock-paper-scissors images in this article. This proposed Conf. Ubiquitous Robot. Ambient Intell. URAI 2017, pp. 406–407,
model has a validation accuracy of 99.54% which is better 2017.
than the previous applied methodology. This proposed [22] A. Salma, “Implementation of Multilayer Perceptron for Image
Classification,” vol. 4, no. February, pp. 212–215, 2021.
RPSICTL model can be effectively used for the image [23] E. Ahmadi, A. G. Pour, A. Siamy, A. Taheri, and A. . Meghdari,
classification of hand playing rock-paper-scissors. “Playing rock-paper-scissors with RASA: A case study on intention
prediction in humanrobot interactive games,” Int. Conf. Soc. Robot.
REFERENCES Springer, Cham, pp. 347–357, 2019.
[24] G. Pozzato, S. Michieletto, and E. Menegatti, “Towards smart robots:
[1] X. Y. Wang, G. Li, S. Malik, and A. Anwar, “Impact of COVID-19 on
Rock-Paper-Scissors gaming versus human players,” PAI@ AI* IA, pp.
achieving the goal of sustainable development: E-learning and
89–95, 2013.
educational productivity,” Econ. Res. Istraz. , vol. 0, no. 0, pp. 1–18,
[25] M. Z. Alom, “The History Began from AlexNet: A Comprehensive
2021.
Survey on Deep Learning Approaches,” ArXiv, vol. 18. pp. 1–12, 2018.
[2] M. E. Osman, “Global impact of COVID-19 on education systems: The
[26] J. Schmidhuber, “Deep Learning in neural networks: An overview,”
emergency remote teaching at Sultan Qaboos University,” J. Educ.
Neural Networks, vol. 61, pp. 85–117, 2015.
Teach., vol. 46, no. 4, pp. 463–471, 2020.
[27] “Rock-Paper-Scissors Images | Kaggle.”
[3] M. P. Murphy, “COVID-19 and emergency eLearning: Consequences
https://www.kaggle.com/drgfreeman/rockpaperscissors (accessed Jan.
of the securitization of higher education for post-pandemic pedagogy,”
20, 2022).
Contemp. Secur. Policy, vol. 41, no. 3, pp. 492–505, 2020.
[28] N. Seliya, T. M. Khoshgoftaar, and J. Van Hulse, “A study on the
[4] A. Kyriazis, G. Mews, E. Belpaire, J. Aerts, and S. A. Malik, “Physical
relationships of classifier performance metrics,” Proc. - Int. Conf.
distancing, children and urban health,” Cities Heal., vol. 00, no. 00, pp.
Tools with Artif. Intell. ICTAI, pp. 59–66, 2009.
1–6, 2020.
[29] [3] Iqbal, N., Hanif, M., Abbas, S., Khan, M.A., Almotiri, S.H. and
[5] A. A. Pradana, Casman, and Nur’aini, “Pengaruh Kebijakan Social
Al Ghamdi, M.A., 2020. DNA strands level scrambling based color
Distancing pada Wabah COVID-19 terhadap Kelompok Rentan di
image encryption scheme. IEEE Access, 8, pp.178167-178182
Indonesia,” J. Kebijak. Kesehat. Indones. JKKI, vol. 9, no. 2, pp. 61–
[30] [1] Iqbal, N., Hanif, M., Abbas, S., Khan, M.A. and Rehman, Z.U.,
67, 2020, [Online]. Available:
2021. Dynamic 3D scrambled image based RGB image encryption
https://jurnal.ugm.ac.id/jkki/article/view/55575.
scheme using hyperchaotic system and DNA encoding. Journal of
[6] H. Brock, S. Šabanović, and R. Gomez, Remote you, haru and me:
Information Security and Applications, 58, p.102809.
Exploring social interaction in telepresence gaming with a robotic
[31] Malik, J.A. and Saleem, M., 2023. Blockchain and Cyber-Physical
agent, vol. 1, no. 1. Association for Computing Machinery, 2021.
System for Security Engineering in the Smart Industry. In Security
[7] K. Usher, N. Bhullar, and D. Jackson., “Life in the pandemic: Social
Engineering for Embedded and Cyber-Physical Systems (pp. 79-98).
isolation and mental health,” J. Clin. Nurs., 2020.
CRC Press.
[8] F. F. Ali, Z. Nakao, and Y. W. Chen, “Playing the Rock-Paper-Scissors
[32] [16] Batool, T., Abbas, S., Alhwaiti, Y., Saleem, M., Ahmad, M., Asif,
game with a genetic algorithm,” Proc. 2000 Congr. Evol. Comput.
M. and Elmitwal, N.S., 2021. Intelligent Model Of Ecosystem For
CEC 2000, vol. 1, pp. 741–745, 2000.
Smart Cities Using Artificial Neural Networks. Intelligent Automation
[9] E. Brockbank and E. Vul, “Formalizing opponent modeling with the
& Soft Computing, 30(2).
rock, paper, scissors game,” Games, vol. 12, no. 3, 2021.
[33] Saleem, M., Khadim, A., Fatima, M., Khan, M.A., Nair, H.K. and Asif,
[10] C. Vogel, “Rock, paper, payoff: Child’s play wins auction house an art
M., 2022, October. ASSMA-SLM: Autonomous System for Smart
sale,” New York Times, 2005, [Online]. Available:
Motor-Vehicles integrating Artificial and Soft Learning Mechanisms.
http://www.nytimes.com/2005/04/29/arts/design/rock-paper-payoff-
In 2022 International Conference on Cyber Resilience (ICCR) (pp. 1-
childs-play-wins-auction-house-an-art-sale.html.
6). IEEE.
[11] M. Zahid, M. A. Khan, F. Azam, M. Sharif, S. Kadry, and J. R.
[34] Saleem, M., Abbas, S., Ghazal, T.M., Khan, M.A., Sahawneh, N. and
Mohanty, “Pedestrian identification using motion-controlled deep
Ahmad, M., 2022. Smart cities: Fusion-based intelligent traffic
neural network in real-time visual surveillance,” Soft Comput., vol. 9,
congestion control system for vehicular networks using machine
2021.
learning techniques. Egyptian Informatics Journal, 23(3), pp.417-426.
[12] P. Hohlfeld, T. Ostermeier, and D. Brandl, “Classification of Beer
[35] Al-Dmour, N.A., Salahat, M., Nair, H.K., Kanwal, N., Saleem, M. and
Bottles using Object Detection and Transfer Learning,” 2022, [Online].

Authorized licensed use limited to: KU Leuven Libraries. Downloaded on May 23,2023 at 19:16:15 UTC from IEEE Xplore. Restrictions apply.
 Aziz, N., 2022, October. Intelligence Skin Cancer Detection using IoT [46] Issa, G., 2021. A new two-step ensemble learning model for improving
with a Fuzzy Expert System. In 2022 International Conference on stress prediction of automobile drivers. The International Arab Journal
Cyber Resilience (ICCR) (pp. 1-6). IEEE. of Information Technology, 18(16).
[36] Ghazal, Taher M., et al. "An IoMT based Ensemble Classification [47] Abu-Arqoub, M., Issa, G., Banna, A.E. and Saadeh, H., 2020.
Framework to Predict Treatment Response in Hepatitis C Patients." Interactive Multimedia-Based Educational System for Children Using
2022 International Conference on Business Analytics for Technology Interactive Book with Augmented Reality. Journal of Computer
and Security (ICBATS). IEEE, 2022. Science, 15(11), pp.1648-1658.
[37] Ghazal, Taher M., et al. "Private blockchain-based encryption [48] Hasan, M.K., Khan, M.A., Issa, G.F., Atta, A., Akram, A.S. and
framework using computational intelligence approach." Egyptian Hassan, M., 2022, February. Smart waste management and
Informatics Journal 23.4 (2022): 69-75. classification system for smart cities using deep learning. In 2022
[38] Ghazal, Taher M., and G. Issa. "Alzheimer disease detection International Conference on Business Analytics for Technology and
empowered with transfer learning." Computers, Materials & Continua Security (ICBATS) (pp. 1-7). IEEE.
70.3 (2022): 5005-5019 [49] Han, K., Yeun, C.Y., Shon, T., Park, J. and Kim, K., 2011. A scalable
[39] Hasan, Mohammad Kamrul, et al. "Fischer linear discrimination and and efficient key escrow model for lawful interception of IDBC‐
quadratic discrimination analysis–based data mining technique for based secure communication. International Journal of Communication
internet of things framework for Healthcare." Frontiers in Public Systems, 24(4), pp.461-472.
Health (2021): 1354 [50] Shehada, D., Yeun, C.Y., Zemerly, M.J., Al Qutayri, M., Al Hammadi,
[40] Ahmed, Usama, et al. "Prediction of diabetes empowered with fused Y., Damiani, E. and Hu, J., 2017. BROSMAP: A novel broadcast based
machine learning." IEEE Access 10 (2022): 8529-8538 secure mobile agent protocol for distributed service applications.
[41] Saleh, Y. and Issa, G., 2020. Arabic sign language recognition through Security and Communication Networks, 2017.
deep neural networks fine-tuning. [51] Al Alkeem, E., Kim, S.K., Yeun, C.Y., Zemerly, M.J., Poon, K.F.,
[42] Issa, G.F., El-Ghalayini, H.A., Shubita, A.F. and Abu-Arqoub, M.H., Gianini, G. and Yoo, P.D., 2019. An enhanced electrocardiogram
2014. A Framework for Collaborative Networked Learning in Higher biometric authentication system using machine learning. IEEE Access,
Education: Design & Analysis. International Journal of Emerging 7, pp.123069-123075.
Technologies in Learning, 9(4). [52] Kim, S.K., Yeun, C.Y. and Yoo, P.D., 2019. An enhanced machine
[43] Issa, G.F., Al-Bahadili, H. and Abuhamdeh, M., 2011. A scalable learning-based biometric authentication system using RR-interval
framework to quantitatively evaluate success factors of mobile learning framed electrocardiograms. IEEE Access, 7, pp.168669-168674.
systems. International Journal of Mobile Learning and Organisation, [53] Byon, Y.J., Ha, J.S., Cho, C.S., Kim, T.Y. and Yeun, C.Y., 2017. Real-
5(3-4), pp.299-316. time transportation mode identification using artificial neural networks
[44] Issa, G., Shen, S. and Chew, M.S., 1994. Using analogical reasoning enhanced with mode availability layers: A case study in Dubai. Applied
for mechanism design. Ieee Expert, 9(3), pp.60-69. Sciences, 7(9), p.923.
[45] Issa, G.F., Hussain, S.M. and Al-Bahadili, H., 2010. A framework for
building an interactive satellite TV based m-learning environment.
International Journal of Interactive Mobile Technologies, 4(3).

Authorized licensed use limited to: KU Leuven Libraries. Downloaded on May 23,2023 at 19:16:15 UTC from IEEE Xplore. Restrictions apply.

View publication stats