Conference Articles

Permanent URI for this collection

https://dspaceusad7.4science.cloud/handle/20.500.12458/19

Browse

Now showing 1 - 20 of 26

A Lower Complexity Deep Learning Method for Drones Detection
(2023)
Mohamad Kassab
;
Amal El Fallah Seghrouchni
;
Frederic Barbaresco
;
Abu Zitar, Raed
Detecting objects such as drones is a challenging task as their relative size and maneuvering capabilities can deceive machine learning models and cause them to misclassify drones as birds or other objects. In this work, we investigate applying several deep-learning techniques to benchmark real data sets of flying drones. A Deep learning paradigm is proposed for the purpose of mitigating the complexity of those systems. The proposed paradigm consists of a hybrid between the AdderNet deep learning paradigm and the SSD paradigm. The goal was to minimize multiplication operations numbers in the filtering layers within the proposed system and, hence, reduce complexity. Some standard machine learning techniques such as SVM is also tested and compared to other deep learning systems. The data sets used for training and testing were either complete or filtered in order to remove the images with mall objects. The types of data were either RGB or IR data. Comparisons were made between all these types and conclusions are presented.
40Scopus© Citations 1
A Review for the Genetic Algorithm and the Red Deer Algorithm Applications
(2021)
Abu Zitar, Raed
The Red Deer algorithm (RD), a contemporary population-based meta heuristic algorithm, applications are thoroughly examined in this paper. The RD algorithm blends evolutionary algorithms' survival of the fittest premise with the productivity and richness of heuristic search approaches. On the other a well-known and relatively older evolutionary based algorithm called the Genetic Algorithm applications are also shown. The contemporary algorithm; the RDA, and the older algorithm; the GA have wide applications in computer science and engineering. This paper sheds the light on all those applications and enable researchers to exploit the possibilities of adapting them in any applications they may have either in engineering, computer science, or business.
Scopus© Citations 9 43
A Review of the Genetic Algorithm and JAYA Algorithm Applications
(2022)
Abu Zitar, Raed
This study throws the light on two metaheuristic algorithms and enable researchers to leverage the potential of adapting them in whatever applications they may have either in engineering, computer science, or business. The two algorithms are the GA and the JAYA. The JAYA algorithm is a modern population-based meta heuristic algorithm, its applications are presented in this work. The JA Y A algorithm integrates evolutionary algorithms' survival of the fittest concept with the productivity and richness of heuristic search methodologies. On the other a well-known and somewhat older evolutionary based method called the Genetic Algorithm with applications is also presented here. The recent two algorithms; the JA Y A and the GA have broad comparable applications in computer science and engineering applications.
19Scopus© Citations 4
Analysis of the Performance of Four Filter Types for Drone Tracking
(2023)
Mohsen, Amani
;
Segrouchni, Amal El Fallah
;
Barbaresco, Frederic
;
Abu Zitar, Raed
In this work, extensive simulations are done to compare the performance of the 4 filter types; Linear Kalman filter (LKF), Extended Kalman Filter (EKF), Unscented Kalman Filter (UKF), and Particle Filter (PF). A simple nearly constant velocity (NCV) motion model is used with a Gaussian noise measurement model. Simulations were done with different ground truths, different measurements covariance matrices, and different speeds of the drone. Stone soup software was used in the simulations. The analyses revealed informative results that gave us more understanding of the behavior of the four filters when a common type of motion model such as the NCV model is used.
17
Application of Red Deer Algorithm in Optimizing Complex functions
(2021)
Zitar, Raed
;
Abualigah, Laith
The Red Deer algorithm (RDA), a recently developed population-based meta-heuristic algorithm, is examined in this paper with the optimization task of complex functions. The RD algorithm blends evolutionary algorithms' survival of the fittest concept with heuristic search techniques' productivity and richness. It is critical to assess this algorithm's performance in comparison with other well-known heuristic methods. The findings are presented along with additional recommendations for increasing RDA performance based on the analysis. The readers of this paper will gain a grasp of the RD algorithm and its optimization ability to determine whether this algorithm is appropriate for their particular business, research, or industrial needs.
Scopus© Citations 4 20
Deep Learning Techniques for Colorectal Cancer Detection: Convolutional Neural Networks vs Vision Transformers
(2024)
Sari, Meriem
;
Moussaoui, Abdelouahab
;
Hadid, Abdenour
Colorectal cancer (CRC) is one of the most common cancers among humans, its diagnosis is made through the visual analysis of tissue samples by pathologists; artificial intelligence (AI) can automate this analysis based on histological images generated from different tissue samples. In this paper we aim to enhance this digital pathology process by proposing two deep learning (DL) based methods that are extremely accurate and reliable despite several limitations. Our first method is based on Convolutional Neural Networks (CNN) in order to classify different classes of tissues into cancerous and non-cancerous cells based on histological images. Our second method is based on Vision Transformers and also classifies images into cancerous and non cancerous cells. Due to the sensitivity of the problem, the performance of our work will be estimated using accuracy, precision, recall and F -score metrics since they ensure more credibility to the classification results; our models have been tested and evaluated with a dataset collected from LC25000 database containing 10000 images of cancerous and non-cancerous tissues, our models achieved promising results with an overall accuracy of 99.84 % and 98.95 % respectively with precision= 100%, recall= 100% and Fl-score= 100%, we observed that both of our models overcame several state-of-the-art results.
9
Deepfakes Signatures Detection in the Handcrafted Features Space
(2023)
Hamadene, Assia
;
Ouahabi, Abdeldjalil
;
Hadid, Abdenour
In the Handwritten Signature Verification (HSV) literature, several synthetic databases have been developed for data-augmentation purposes, where new specimens and new identities were generated using bio-inspired algorithms, neuromotor synthesizers, Generative Adversarial Networks (GANs) as well as several deep learning methods. These synthetic databases contain synthetic genuine and forgeries specimens which are used to train and build signature verification systems. Researches on generative data assume that synthetic data are as close as possible to real data, this is why, they are either used for training systems when used for data augmentation tasks or are used to fake systems as synthetic attacks. It is worth, however, to point out the existence of a relationship between the handwritten signature authenticity and human behavior and brain. Indeed, a genuine signature is characterised by specific features that are related to the owner’s personality. The fact which makes signature verification and authentication achievable. Handcrafted features had demonstrated a high capacity to capture personal traits for authenticating real static signatures. We, therefore, Propose in this paper, a handcrafted feature based Writer-Independent (WI) signature verification system to detect synthetic writers and signatures through handcrafted features. We also aim to assess how realistic are synthetic signatures as well as their impact on HSV system’s performances. Obtained results using 4000 synthetic writers of GPDS synthetic database show that the proposed handcrafted features have considerable ability to detect synthetic signatures vs. two widely used real individuals signatures databases, namely CEDAR and GPDS-300, which reach 98.67% and 94.05% of successful synthetic detection rates respectively.
14Scopus© Citations 1
Deploying model obfuscation: towards the privacy of decision-making models on shared platforms
(2024)
Sadhukhan, Payel
;
Chakraborty, Tanujit
;
Sengupta, Kausik
The automation of the industrial paradigms characterizes the era of Industry 4.0. The implementation nuances involve data and model sharing among allies and partners working on the same domain. Privacy and security of data and models are fundamental necessities that must be satisfied for this protocol's proper functioning. To this end, we propose a conceptual and algorithmic framework of a model obfuscation scheme. It is built upon the extant data obfuscation paradigm. The future work lies with the implementation and establishment of its viability. This research is expected to develop into deployable model obfuscation technique which practitioners from the industrial domain can adopt.
24 2
Driver's Facial Expression Recognition Using Global Context Vision Transformer
(2023)
Saadi, Ibtissam
;
Cunningham, Douglas W
;
Abdelmalik, Taleb-Ahmed
;
Hadid, Abdenour
;
El Hillali, Yassin
Driver's facial expression recognition plays a critical role in enhancing driver safety, comfort, and overall driving experience by proactively mitigating potential road risks. While most existing works in this domain rely on CNN - based approaches, this paper proposes a novel method for driver facial expression recognition using Global Context Vision Transformer (DFER-GCViT). With its inherent capabilities of transformer-based architectures and global context modeling, the proposed method handles challenges commonly encountered in real-world driving scenarios, including occlusions, head pose variations, and illumination conditions. Our method consists of three modules: preprocessing for face detection and data augmentation, facial feature extraction of local and global features, and expression classification using a modified GC-ViT classifier. To evaluate the performance of DFER-GCViT, extensive experiments are conducted on two benchmarking datasets namely the KMU-FED driver facial expression dataset and FER2013 general facial expression dataset. The experimental results demonstrate the superiority of DFER-GCViT in accurately recognizing driver's facial expressions, achieving an average accuracy of 98.27 % on the KMU-FED dataset and 73.78% on the FER2013 dataset, outperforming several state-of-the-art methods on these two benchmarking datasets.
Scopus© Citations 1 4
Drone Tracking Based on the Fusion of Staring Radar and Camera Data: An Experimental Study
(2023)
De Rouchechouart, Maxence
;
Ahmad, Bashar I.
;
Seghrouchni, Amal El Fallah
;
Barbaresco, Frederic
;
Harman, Stephen
;
Abu Zitar, Raed
This paper presents an experimental study on tracking a small drone target with a high resolution camera and a staring radar. The objective is to assess the benefits of fusing the outputs of both sensors using real data collected during live drone trials. We examine the impact of losing the signal from one sensor, which often occurs in practice for various reasons such as occlusions, high background noise-clutter, target sharp maneuvers, etc. We demonstrate that fusion with filtering, namely employing interacting multiple models with unscented Kalman filter in modified spherical coordinates or a simple extended Kalman filter, can deliver improved overall target tracking performance under such degraded sensing conditions.
21Scopus© Citations 1
Drone/Bird Classification Based on Features of Tracks Trajectories
(2023)
Kengeskanov, Maksat
;
Seghrouchni, Amal El Fallah
;
Abu Zitar, Raed
;
Barbaresco, Frederic
This paper presents the outcome of several machine learning techniques used for the task of bird/drone classification based on their tracks. Instead of using static images, the dynamics and features extracted from the trajectories captured in videos are used to provide a more accurate and reliable recognition task. Standard Machine Learning methods such as SVM and Random Forest are used for learning this classification. Features based on the kinematics, Gabor filter, and Gray Level Co-occurrence Matrix are utilized. Several comparisons and experiments based on benchmark data sets are show
39
End to End Camera only Drone Detection and Tracking Demo within a Multi-agent Framework with a CNN-LSTM Model for Range Estimation
(2024)
De Rochechouart, Maxence
;
Abu Zitar, Raed
;
Seghrouchni, Amal El Fallah
;
Barbaresco, Frederic
We present an end-to-end camera-only drone tracking approach in a multi-agent framework. We show implementation and simulation of such a system and test the tracking components utilizing a CNN-LSTM model for range estimation tested on real data. A video of the demo is available at this link (https://drive.google.com/file/d/1AlV89lgfi5nqwZCHTXC0pLm6BsZh_cBk/view?usp=drive_link).
28 3
Evaluation of Pre-Trained CNN Models for Geographic Fake Image Detection
(2022)
Hadid, Abdenour
;
Fezza, Sid
;
Ouis, Mohammed
;
Kaddar, Bachir
;
Hamidouche, Wassim
Thanks to the remarkable advances in generative adversarial networks (GANs), it is becoming increasingly easy to generate/manipulate images. The existing works have mainly focused on deepfake in face images and videos. However, we are currently witnessing the emergence of fake satellite images, which can be misleading or even threatening to national security. Consequently, there is an urgent need to develop detection methods capable of distinguishing between real and fake satellite images. To advance the field, in this paper, we explore the suitability of several convolutional neural network (CNN) architectures for fake satellite image detection. Specifically, we benchmark four CNN models by conducting extensive experiments to evaluate their performance and robustness against various image distortions. This work allows the establishment of new baselines and may be useful for the development of CNN-based methods for fake satellite image detection.
30 40
Face Sketch Synthesis using Generative Adversarial Networks
(2022)
Hadid, Abdenour
;
Mahfoud, Sami
;
Daamouche, Abdelhamid
;
Bengherabi, Messaoud
;
Boutellaa, Elhocine
Face Sketch Synthesis is crucial for a wide range of practical applications, including digital entertainment and law enforcement. Recent approaches based on Generative Adversarial Networks (GANs) have shown compelling results in image-to-image translation as well as face photo-sketch synthesis. However, these methods still have considerable limitations as some noise appears in synthesized sketches which leads to poor perceptual quality and poor preserving fidelity. To tackle this issue, in this paper, we propose a Face Sketch Synthesis technique using conditional GAN to generate facial sketches from facial photographs named cGAN-FSS. Our cGAN-FSS framework generates high perceptual quality of face sketch synthesis while maintaining high identity recognition accuracy. Image Quality Assessment metrics and Face Recognition experiments confirm our proposed framework's performs better than the state-of-the-art methods.
38 38
Heart Rate Variability Analysis Reveals a Non-monotonic Relationship between Humanin Concentration and Cardiac Autonomic Regulation
(2022)
Feng, Samuel
;
Yousef, Hibba
;
Khandoker, Ahsan H.
;
Tarvainen, Mika P.
;
Jelinek, Herbert F.
Oxidative stress (OS) has been shown to have a negative effect on the autonomic nervous system (ANS) and on ANS modulation of heart rate. Mitochondrial ATP production is the main source of reactive oxygen species (ROS) and hence the regulation of ROS becomes an important issue in maintaining optimal ANS functionality. Humanin (HN), a mitochondrial-derived peptide, plays an important role in lowering OS. Sympathovagal balance was assessed in 124 healthy participants through heart rate variability (HRV) analysis and compared across changes in HN concentrations divided into quintiles, with values of HN ranging from 64.6 to 343.2 pg/mL. Significant differences included various frequency domain and nonlinear HRV parameters, particularly between first and fourth HN quintiles with p values < 0.001 for recurrence plot analysis (RPA), detrended fluctuation analysis (DFA) a1 and Poincaré plot ratio SD1/SD2. The results revealed non-monotonic relationships between measures of HRV and HN concentration. A mitohormetic type of relationship was observed with HRV features increasing and then decreasing with increasing HN concentration. These results are consistent with previous findings of the importance of HN levels in regulating OS and extend these by revealing a concomitant effect on the modulation of cardiac rhythm by the ANS.
50Scopus© Citations 1 1
Knowing the class distinguishing abilities of the features, to build better decision-making models
(2024)
Sadhukhan, Payel
;
Sengupta, Kausik
;
Palit, Sarbani
;
Chakraborty, Tanujit
Explainability allows end-users to have a transparent and humane reckoning of an ML scheme's capability and utility. ML model's modus opernadi can be explained via the features which trained it. To this end, we found no work explaining the features' importance based on their class-distinguishing abilities. In a given dataset, a feature is not equally good at distinguishing between the data points' possible categorizations (or classes). This work explains the features based on their class or category-distinguishing capabilities. We estimate the variables' class-distinguishing capabilities (scores) for pair-wise class combinations, utilize them in a missing feature context, and propose a novel decision-making protocol. A key novelty of this work lies in the refusal to render a decision option when the missing feature (of the test point) has a high class-distinguishing potential for the likely classes. Two real-world datasets are used empirically to validate the explainability of our scheme.
14 1
Machine Learning Approach for the Design of an Assessment Outcomes Recommendation System
(2021)
Abu Zitar, Raed
;
Fatime Al-Zahra
;
Shaimaa Mounir
;
Lamees Dalbah
It is believed that the evaluation of the outcomes of the course, based on grades, is necessary to improve the teaching and learning process. Our research processes and workflows supported by AI utilize machine learning technology in order to interpret big data, analyze broad data sets and recognize associations with more reliably. The course learning outcomes will be assessed on the basis of QF-Emirates guidelines and use it to suggest teaching and learning measures. It will be used to determine courses learning results based on the empirical knowledge presented. We research and test the design of the right neural networks that achieves our goal. A modern algorithm was improvised for this reason. For our proposed recommendation system, a database program was created to store data and include details in the analysis of course learning outcomes. As a machine-learning system, the proposed approach is tested and results are competitive.
21
Multi-Target Tracking Resources Allocation Using Multi-Agent Modeling and Auction Algorithm
(2023)
De Rochechouart, Maxence
;
Seghrouchni, Amal El Fallah
;
Barbaresco, Frederic
;
Abu Zitar, Raed
This paper proposes a multi-agent technique for modeling and simulating multiple objects and multiple sensor tracking for drones. The goal is to achieve efficient resource allocation based on different preset cost functions that take into account several local and global attributes. The target dynamics, the sensor selection, and the measurements are simulated by way of an agent-based modeling framework called MESA. The results show an optimum resource allocation with an information-based cost function that is optimized by an Auction Algorithm.
22
Optimum Track to Track Fusion Using CMA-ES and LSTM Techniques
(2024)
Fares, Samar
;
Seghrouchni, Amal El Fallah
;
Barbaresco, Frederic
;
Abu Zitar, Raed
This paper presents two different methods for track-to-track fusion of drone tracks. The sensors are unbiased radars with fixed locations. The first method uses an offline technique based on a global optimizer called the CMA-ES algorithm and the second one uses LSTM in its different forms to learn the online adjustment of the fusion weights between the two tracks. An objective function utilizing the covariance of the fused tracks is used by the first algorithm while a cost function based on the Kullback-Leibler (KL) divergence measure is used in the second case for training the LSTM. The two methods are compared with other baseline methods using performance metrics such as SIAP and OSPA. Simulations are done for a single object (drone) and repeated for multiple objects in the presence of two radars to demonstrate the validity of the two proposed techniques. The JPDA (Joint Probability Data Association) with fixed gating and moderate clutter is used in the case of multiple objects. Stone Soup was chosen as the radar simulation environment.
17
Predictive Model of Psychoactive Drugs Consumption using Classification Machine Learning Algorithms
(2023)
Almahmood, Mothanna
;
Najadat, Hassan
;
Alzubi, Dalia
;
Abualigah, Laith
;
Abu Zitar, Raed
;
Abualigah, Sayel
;
AL-Saqqar, Faisal
It is difficult to predict the effect of drugs on the individuals, as its results are unpredictable and most often dangerous. For a police purpose that concerned with the protection of individuals, the problem of predicting drug abusing is highly important. A dataset was used from open-source website UCI, that includes specific attributes about using up of eighteen different psychoactive drugs. Our study aimed to use data mining classification techniques, in order to classify the individual into two categories: user or non-user. Eighteen classification models were built using different classification algorithms such as Gaussian Naive Bais, Logistic Regression, k-nearest neighbors, Random Forest, and Decision Tree. The accurate classifier was chosen by studying the accuracy, recall, precision, and f1-score measures for each one, and it was evaluated by the Holdout method. The results were obtained optimally, and we got 18 models, where each one had different high accurate outputs, that classify an individual to user and non-user. The final model is a combination of 18 models for 18 critical psychoactive drugs: Alcohol, Amphet, Amyl, Benzos, Caff, Cannabis, Choc, Coke, Crack, Ecstasy, Heroin, Ketamine, Legalh, LSD, Meth, Mushrooms, Nicotine and VSA. This study in turn may give a chance for the decision makers to reduce the risk of these drugs consumption, in order to avoid healthcare issues and keep the community in safe.
27 1

Browse

Browsing Conference Articles by Discipline "Physics, Mathematics, Computer science"

Results Per Page

Sort Options