This research proposes a Federated Multi-Modal Deep Learning Framework (FedMM-AD) with explainability features and privacy-preserving techniques for early-stage AD prediction. Our federated model was constructed using MRI, PET, and CT scan data, protecting patient privacy and security while enabling cross-institution training of the model without sharing any patient information. The FedMM-AD model’s accuracy, AUC, sensitivity, and specificity were 98.59%, 97.3%, and 98.2%, respectively, according to the experiment conducted on the ADNI and OASIS datasets. In order to increase the model’s transparency and help doctors identify key brain regions for AD, we integrated explainable AI techniques employing cross-attention. The findings show that when additional stakeholders, like clinicians, are properly involved, the approach can be beneficial and have an impact.

Blockchain-based federated learning (BCFL) is emerging as a promising solution to these challenges. Unlike traditional machine learning frameworks, including standard federated learning, BCFL offers a secure framework for distributed analysis of agricultural big data while integrating a transparent reward system via smart contracts. The development of smart agriculture requires data analysis methods that preserve privacy while encouraging collaboration among stakeholders. This review analyses studies (2018-2023) to assess how the integration of blockchain in federated learning (BCFL) addresses this dual challenge. Three key findings emerge: (1) 78% of BCFL implementations improve model auditability while maintaining accuracy > 85% (average difference of -9.2% vs. conventional FL); (2) Reward smart contracts reduce participant dropout rates by 40%; and (3) Only 12% of existing frameworks integrate differential privacy (DP) mechanisms. Our analysis reveals key barriers such as scalability (average latency of 2.7 s/transaction) and the heterogeneity of agricultural devices. Identified future directions include the adoption of lightweight blockchains (IOTA) and hybrid FL-DP protocols. This comprehensive review examines the integration of federated learning and blockchain in the agricultural domain. By analysing current trends, technical challenges, and future perspectives, this work aims to establish a framework for decentralized, secure, and privacy-preserving solutions. Unlike traditional approaches, our analysis highlights the potential of this synergy to optimize distributed agricultural data management, strengthen supply chain traceability, and improve decision-making through collaborative AI models. By identifying existing limitations (latency, scalability, interoperability), we pave the way for future research aimed at making these technologies accessible to farms of various sizes, while ensuring transparency and energy efficiency.

Brain tumors, whether benign or malignant, present significant challenges in medical diagnosis and treatment. Timely and precise detection is critical for effective intervention and patient outcomes. This study introduces a pioneering method for brain tumor detection, employing a fusion of Convolutional Neural Networks (CNN) and Vision Transformer (ViT) architectures. By integrating these models, we exploit their complementary features in image analysis, particularly in medical imaging contexts. Our research assesses the performance of this integrated CNN-ViT framework across various brain tumor imaging modalities and clinical scenarios using extensive experimentation on benchmark datasets. Results validate the robustness and accuracy of our approach, achieving a remarkable precision, recall rates, and overall accuracy of 98%.

The processing of hyperspectral remote sensing images is a crucial field of study. As one of the most important phases in image processing right now, the classification task has attracted our attention. The Convolutional Neuronal Network (CNN) is one of the most demonstrative algorithms for deep learning since it is a sort of feed-forward neural network that uses convolutional computation. In this paper we proposed a customized model based on CNN applied on SAT 4 and SAT6 datasets. The experimental results outperformances methodology with accuracy value of 99.4%, loss of 2% and 99.8%, loss of 3.2% for SAT4 and SAT6, respectively.

Aerial imagery has emerged in numerous fields such as sustainable development, forestry, urban planning, agriculture, earth science and climate research. Extracting relevant information from satellite images, such as building detection, road extraction, and land cover classification, is crucial for decision-making. Semantic segmentation (SS),generates a dense pixel-wise segmentation map of a given satellite image where each pixel is related necessarily to a specific class. (SS) has become essential to reach the aforementioned goals. In this context, we presented an improved approach for semantic segmentation of aerial images using the pre-trained CNN VGG16 model and the modified U-Net architecture. Our results show that our proposed approach exceeds state of the art methods in accurately classifying land cover types in terms of dice coefficient and an average accuracy of 82.30% of 87.81% respectively.

In this paper, we propose a novel approach for semantic segmentation based on Deep Learning methods for Autonomous Driving. Our main idea is to develop a fully convolutional neural network (CNN) architecture for semantic segmentation based on the U-Net model. For that, we propose a model with different layers such as a convolutional layer, pooling layer, dropout layer, and fully connected layer. The experiment results obtained, using the CamVid dataset, show that the choice of the number of periods and the depth of the network have a great influence on the best results.

Machine learning and artificial intelligence have led to notable progress in the field of deep learning and the identification of text within natural scene images in the past few years. Despite notable advancements, the effectiveness of deep learning and text detection in natural scene images, particularly for Arabic language, is frequently constrained by the scarcity of comprehensive datasets containing various multilingual scripts. YOLO (You Only Look Once) is a widely used deep learning neural network that has gained immense popularity for its versatility in handling diverse machine learning tasks, primarily in the field of computer vision. The YOLO algorithm has progressively garnered recognition for its exceptional performance to solving a complex problems, noisy data, as well as overcoming various challenges encountered in real-world scenarios. Our experiments provide a concise examination of text detection algorithms based on convolutional neural networks (CNNs), especially different versions of the YOLO models using a data augmentation technique applied to "SYPHAX" dataset, our new dataset of multilingual scripts in the wild. The objective of this paper is to offer insights into the future of YOLO in the field, highlighting potential research directions that can enhance text detection systems.

Remarkable performance on computer vision, and especially on pattern recognition field has been known for a long time to be produced by Deep learning algorithms. It is clear that amongst the successful applications in the pattern recognition domain, Arabic handwriting recognition (AHR) is a must. In this survey, we use two deep networks: Deep Belief Network (DBN) and Convolutional Neural Networks (CNN), for Arabic handwritten script (AHS) recognition. Despite the triumph of DBN and CNN methods, over-fitting is able to take place on these networks thanks to the massive number of parameters. In order to fight over-fitting, we have deeply inquired two regularization techniques called Dropout and DropConnect. While training with the two regularization methods, a randomly chosen subsets of activations/weights are dropped. Consequently, the assessment on the HACDB database to treat character level proves shows an improvement of classification error rate once adding Dropout and DropConnect techniques.

In this paper, we investigate an efficient palmprint texture modeling method that incorporates a robust analysis based on fusing multiple information. In fact, a single descriptor alone may not achieve a high accuracy in palmprint biometric system. Hence, we propose the fusion of various information features extracted by the different descriptors, such as the fractal and the Multi-fractal techniques which produce a robustness to face the numerous challenging and variation of palmprint in unconstrained environments. To increase the performance of palmprint biometric systems, information fusion is proposed as a key phase in multi-characteristic systems. The obtained information can be combined at different levels, i.e., at the feature level, the score level or the decision level. Nevertheless, the feature level fusion is considered more effective than both the matching score and the classifier decision levels, thanks to a feature vector set which contains more and richer information about the input palmprint image. In order to improve the discriminating texture information, our proposed method extracts the fractal dimension features from the preprocessed palmprint images and fuses them with the Multi-fractal dimension features using the Canonical Correlation Analysis (CCA) incorporating the Linear Discriminant Analysis (LDA) in order to reduce the feature dimensionality for each feature set. To demonstrate the feasibility and effectiveness of our proposed method, we performed the experimental results on two benchmark datasets. These results outperform other well-known state of the art methods and produce promising recognition rates by achieving 96.02% for PolyU-Palmprint database and 97.00% for CASIA-Palmprint database.

This paper focuses on hand pose estimation by proposing a system that solves real-time static hand gesture detection and recognition issues for interacting with smartphones. The first step of our work consists in detecting and tracking the hand in a complex background. The second step consists in recognizing hand gestures using SVM “Support Vector Machine”. Finally, we developed a grammar to generate gesture commands for mobile applications control. This work presents a system based on a real-time hand gesture recognition algorithm for Android devices. The idea here is to allow the user interacting with the mobile device without the need to touch the screen. In this system, the Smartphone is able to perform all necessary steps to recognize gestures without the need to connect to a distant device.

Lately, Online Arabic Handwriting Recognition has been gaining more interest because of the advances in technology such as the handwriting capturing devices and impressive mobile computers. And since we always try to improve recognition rates, we propose in this work a new system based on a deep recurrent neural networks on which the dropout technique was applied. Our approach is very practical in sequence modelling due to their recurrent connections, also it can learn intricate relationship between input and output layers because of many non-linear hidden layers. In addition to these contributions, our system is protected against overfitting due to powerful performance of dropout. This proposed system was tested with a large dataset ADAB to show its performance against difficult conditions as the variety of writers, the large vocabulary and diversity of style.

Recognizing the online Arabic handwritten script has been gaining more interest because of the impressive advances in mobile device requiring more and more intelligent handwritten recognizers. Since it was demonstrated within many previous research that Deep Neural Networks (DNN) exhibit a great performance, we propose in this work a new system based on a DNN in which we try to optimize the training process by a smooth construct of the deep architecture. The Output’s error of each unit in the previous layer will be computed and only the smallest error will be maintained in the next iteration. This paper uses LMCA database for training and testing data. The experimental study reveals that our proposed DBNN using generated Bottleneck features can outperform state of the art online recognizers.

As a machine learning algorithms, deep learning algorithms developed in recent years, have been successfully practiced in many fields of computer vision, like face recognition, object detection and image classification. These Deep algorithms look for drawing out a very performing representation of the data, among which image and speech, through multi-layers in a deep hierarchical structure. In this study, a deep learning model based on Support Vector Machine (SVM) named Deep SVM (DSVM) is represented. We applied the dropout technique on the Deep SVM (DSVM). It is worth noting that this model has an inherent capacity to choose data points crucial to classify good generalization capacities. The deep SVM is built by a stack of SVMs permitting to extracting/learning automatically features from the raw images and to realize classification, too. We chose and tested the Multi-class Support Vector Machine with an RBF kernel, as non-linear discriminative features for classification, on Handwritten Arabic Characters Database (HACDB). Further to these advantages, our model is safeguarded against over-fitting because of strong performance of dropout. Simulation outcomes prove the efficiency of the suggested model.

Recently, personal identification, which is based on the palmprint texture features analysis, has widely attracted the attention of several researchers and has gained a great popularity in the pattern recognition field. In this paper, we present a novel methodology based on texture information extracted from palmprint. Firstly, we propose an algorithm to robustly locate the Region Of Interest (ROI) of the hand. Secondly, we combine multiple descriptors to extract the palmprint texture information, which are Gray-Level Co-occurrence Matrix (GLCM) and the Gabor filters using feature level fusion. These descriptors have been broadly applied in various tasks, specifically in the image processing domain to analyze the image texture. Then, we apply the generalized discriminant analysis (GDA) to reduce the length of the feature vectors and their redundancies. Finally, we classify these final resulting features by developing the SVM method which supports several kernel functions to reach a best recognition rate. We have conducted extensive experiments on the “CASIA-Palmprint” and “PolyU-palmprint” datasets. The obtained results of the proposed approach provide promising results compared to other well-known state-of-the-art approaches.

This paper presents a new method to recognize the person's identity through their palmprints. Palmprint recognition is among the most reliable physiological characteristics that can be used especially in forensic applications thanks to its simplicity and its ease of use, its user friendliness and high identification reliability. Accordingly, it has gained great popularity within the pattern recognition field over the past three decades. In this paper, we suggest a new approach for personal identification based on palmprint features extracted using the various methods of fractal theory. These methods have been broadly applied in image processing fields to estimate the fractal dimensions of an image as an important parameter for the analysis of objects of irregular shapes of the texture image. The novelty of this approach is two-fold. On the one hand, we apply the Box counting (BC), the Mass Radius (MS) and the Cumulative Intersection (CumInt) methods to extract the palmprint texture information. On the other hand, the combination of efficient information from the three descriptors has been presented in order to make identification system more efficient and achieve better performances. Then, we explore such texture information features by using classical machine learning techniques: the K-Nearest Neighbor (KNN), the Support Vector Machine (SVM) and the Multiclass Random Forest classification algorithms. The results of the experiments conducted on two large datasets show that our proposed method gives better recognition rates of about 96.35% for CASIA-Palmprint dataset and 95.98% for IITD-Touchless-Palmprint dataset. These results obtained are compared to other well-known state-of-the-art approaches.

Handwriting recognition ranks among the highest and the most triumphant applications in the pattern recognition domain. Despite being a developed field, many enquiries are still needed and still represent a defiance mainly for the Arabic Handwritten Script (AHS). Recently, more regard has been given to Support Vector Machines (SVM) classifier for script recognition. Nevertheless, it has not been put in application yet to the handwritten Arabic field if compared with the other methods like ANN, CNN, RNN and HMM. SVMs for AHS recognition is examined in this paper. Handcrafted feature is handled as input by the suggested method and gets going with a supervised learning algorithm. We chose the Multi-class Support Vector Machine with an RBF kernel and we tested it on Handwritten Arabic Characters Database (HACDB) as well. It was proven that the proposed method was effective thanks to the simulation results. We compared the well-functioning of this method with character recognition reliabilities coming from state-of-the-art Arabic OCR which resulted in commendatory outcomes.

Hand gestures are natural and intuitive communication way for the human being to interact with his environment. They serve to designate or manipulate objects, to enhance speech, or communicate in a noisy place. They can also be a separate language. Gestures can have different meanings according to the language or culture. They can also be a way to interact with machines. The subject of our research concerns the design and development of computer vision methods for recognizing hand gestures by a mobile device. We have proposed a system based on SVM for recognizing various hand gestures. The system consists of four steps: hand segmentation, smoothing, feature extraction and classification. The idea here is to allow the smartphone to perform all necessary steps to recognize gestures without the need to connect to a computer in which a database is located to perform training process. With this system, all steps can be done by the smartphone. In this paper, for image acquisition, frontal camera of the smartphone is used. After that frames are gotten from the video, the color sampling is done which is followed by making binary representation of the hand, and then contours representing the hand were described with convex polygons to get information about fingertips and finally the input gesture was recognized using proper classifier.

The Palmprint recognition is a reliable method to identify the person. Actually, several researchers are interested in this method in the last few years thanks to its intense contributions to public security. This paper aims to present a biometric system based on a new approach that focuses on the dominant features of palmprint for recognition. It also introduces new techniques to locate the Region Of Interest (ROI) of the hand by using the Chinese Database, which is called CASIA-Palmprint. We used various methods which employ the image processing techniques including the use of the Otsu's method to binarize the original image, the boundary extraction, the smoothing determination to remove the noise and to eliminate the holes curve and the detection of the centroid to extract the key points applying the Euclidean distance. These methods are used to finally extract the ROIs from the left and right hand. We resized the ROIs to 150 × 150 and we rotated it according to the detected angle. Experimental results were obtained on the “CASIA-Palmprint” database show promising results and demonstrate the effectiveness of the proposed system. These results obtained are compared with some of the state of the arts techniques.

In this paper, we propose a new system for identification and verification of offline handwriting signature. The identification allows knowing the owner of such a signature, whereas the verification allows verifying the authenticity of a given signature. In recognition systems, the phase of feature extraction represents the most pertinent step. Therefore, we opted to use the fractal approach which is very sophisticated for objects with irregular forms. In this work, we have implemented three methods for calculating fractal dimensions. Experimental results have been assumed on a large database, namely, “FUM-PHSDB” showing promising result and demonstrating the effectiveness of the proposed system. These experiments show an identification rate of 95% and a verification rate of 83%.

This paper deals with the improvement of an on-line Arabic handwriting modeling system based on graphemes segmentation. The presented strategy consists in the integration of off-line features to assimilate and take up the handwriting style variation in a multi-writer context. The main contribution of the presented work consists in making off-line fuzzy template for each on-line segmented graphemes trajectory and the extraction of geometric moments invariants by using a method adapted to the irregular spatial sampling of their on-line trajectory. The experimental results prove the added value of the introduced features on the discriminative power of the developed handwriting modeling system.

The aim of this paper is to address the task of writer Identification of on-line handwriting. A new method for analytical on-line writer identification is proposed. However, although it is possible to measure the degree of handwriting irregularity thanks to the fractal dimension, the fractal analysis with a single exponent is not enough sufficient to characterize handwriting styles variation, instead, a continuous spectrum of exponents is necessary. In this purpose Multi-Fractal analysis was used to characterize styles of writing of writers. The main objective of this study is to explore the utility of this novel statistical tool for the purpose of distinguishing styles of on-line writings. Furthermore, a new method to estimate Multi-Fractal dimensions for on-line handwriting is presented and a procedure to find the most distinctive graphemes is elaborated. To evaluate our method, we have used the writings of 100 writers from the ADAB database. Our experimental results demonstrate the effectiveness of our proposed method and show a large capability of Multi-fractal features to characterize on-line handwriting styles.

Writer identification still remains as a challenge area in the field of off-line handwriting recognition because only an image of the handwriting is available. Consequently, some information on the dynamic of writing, which is valuable for identification of writer, is unavailable in the off-line approaches, contrary to the on-line approaches where temporal and spatial information for the handwriting is available. In this paper we present a new method for writer identification based on Multi-Fractal features for both types of presented approaches. This method consists to extract the multi-fractal dimensions from the images of Arabic words and the on-line signals for the same words. In order to enhance the performance of our writer identification system, we have combined both on-line and off-line approaches, taking the advantage it provides ADAB database, which allows to recover the on-line signal and image for the same handwriting. In this way, our work consists to take advantage of static and dynamic representations of handwriting, in order to identify the writer in realistic conditions. The tests are performed on the writing of 100 writers from the ADAB database. The obtained results show the effectiveness of the proposed writer identification system.

In this paper we designed an adaptation module (AM) with the objective to increase the performance of a recognition system for a new user or new writing style. The developed adaptation module is added after the recognition system, and its role is to examine the output of the independent system and produce a more correct output vector close to the desired response of the user. To achieve this end, we conceive an adaptation module based on Radial Basis Function Neural Network (RBF-NN) which is built using an incremental training algorithm. Two adaptation strategies are applied for adaptation module training: increase the number of new hidden units and adjust the parameters of the nearest unit (weights and location of center) using the standard descent gradient. This new architecture is evaluated by the adaptation of two recognition systems, one for digit recognition and one for alphanumeric character recognition. The results, reported according to the cumulative error, show that the adaptation module (AM) leads to decreasing the classification error and is capable of fast adaptation to the users handwriting. Moreover, results are compared with those carried out using the weights updating strategy of the nearest center apart from the addition of new units. In fact, the adaptation module decreases an average of 50% the error rate with standard recognition systems.

Arabic script presents a challenge complexity and variability for handwriting recognition. The first on line Arabic Database called ADAB is known as a standard benchmark in the ICDAR competition of 2009. This paper describes the Online Arabic handwriting recognition competition held at ICDAR 2011. 3 groups with 5 systems are participating in the competition. The systems were tested on known data (sets 1 to 4) and on two test datasets which are unknown to all participants (set 5 and set 6). The systems are compared on the most important characteristic of classification systems, the recognition rate. Additionally, the relative speed of every system was compared. A short description of the participating groups, their systems, the experimental setup, and the performed results are presented.

The aim of this paper is to understand the importance of modeling the dynamic of industrial systems using Recurrent Neural Network (RNN) and report the results obtained by training the RNN on a textile process to identify the relationship between yarn color and fabric color. The importance of RNN can be highlighted by the fact that the information about the underlying dynamics of such systems is not available. The dynamics can only be observed with the help of certain measurable variables. In this context, Recurrent Neural Network based approach is a powerful tool with promising results. In conventional Neural Network based approaches, periodic training is required. The time between retraining is still an open issue.

In this paper we present a new modeling approach for online Arabic handwriting which is based on fuzzy graphemes segmentation. In the literature, the result of the graphemes segmentation of a cursive writing not often reaches its optimum. This fact is due to the crisp aspect of the segmentation decision. In order to overcome this problem, we propose to introduce a fuzzy effect in this segmentation decision by overlapping the segmented graphemes in proportion to the confidence degrees associated with the detection of the particular points that separate them. The fuzzified boundary shapes of the extracted fuzzy graphemes are then modeled taking into account the coefficient of fuzzy membership of their points. The obtained results by using the ADAB database show an improvement of the recognition rate given by the fuzzy segmentation approach compared to the crisp one.

We present in this paper a new approach of online Arabic handwriting modeling based on the graphemes segmentation. This segmentation rests on the previous detection of baseline. It involves the detection of two types of topologically meaningful points: the backs of the valleys adjoining the baseline and the angular points. The stage of features extraction allows to model the shapes of segmented graphemes by relevant geometric parameters and to estimate their diacritics fuzzy affectation rates. The test results show a significant improvement in recognition rate with the introduction of new pertinent parameters.

In recent years, fractal and multi-fractal analysis have been widely applied in many domains, especially in the field of image processing. In this direction we present in this paper a novel method for Arabic text-dependent writer identification based on fractal and multi-fractal features; thus, from the images of Arabic words, we calculate their fractal dimensions by using the “Box-counting” method, then we calculate their multi-fractal dimensions by using the method of DLA (Diffusion Limited Aggregates). To evaluate our method, we used 50 writers of the ADAB database, each writer wrote 288 words (24 Tunisian cities repeated 12 times) with 2/3 of words are used for the learning phase and the rest is used for the identification. The results obtained by using knearest neighbor classifier, demonstrate the effectiveness of our proposed method.

This paper describes the Online Arabic handwriting recognition competition held at ICDAR 2009. This first competition uses the ADAB-database with Arabic online handwritten words. This year, 3 groups with 7 systems are participating in the competition. The systems were tested on known data (sets 1 to 3) and on one test dataset which is unknown to all participants (set 4). The systems are compared on the most important characteristic of classification systems, the recognition rate. Additionally, the relative speed of the different systems were compared. A short description of the participating groups, their systems, the experimental setup, and the performed results are presented.

In this paper we present a new method of baseline detection of online or offline short handwriting. This work is part of a large project for the edification of a dual online / offline Arabic handwriting recognition system. Compared to the existing approaches in the literature, this new method brings three specific novelties: First, the consideration of the agreement between the alignment of the points and their trajectory tangent directions for the detection of aligned points regroupings. Then, the consideration of a topologic characteristics specific to the used writing language, to value the pertinence of the pretender points regroupings to be recognized as baseline. Finally, we showed the aptitude of the algorithm to detect curved baseline.

In this paper we present a system of the off-line handwriting recognition. Our recognition system is based on temporal order restoration of the off-line trajectory. For this task we use a genetic algorithm (GA) to optimize the sequences of handwritten strokes. To benefit from dynamic informations we make a sampling operation by the consideration of trajectory curvatures. We proceed to calculate the curvilinear velocity signal and use the beta-elliptical modelling which is developed in on-line systems to calculate other characteristics. Our approach is validated by Hmm Tool Kit (HTK) recognition system using IFN/ENIT database.

In this paper, we present a new approach to the temporal order restoration of the off-line handwriting. After the pre-processing steps of the word image, a suitable algorithm makes it possible to segment its skeleton in three types of strokes. After that, we developed a genetic algorithm GA in order to optimize the best trajectory of these segments. The repetition of a segment will be studied in a secondary algorithm so that we do not disturb the GA operations. The techniques used in GA are the selection, crossover and the mutation. The fitness function value depends on right-left direction (direction of the Arab writing), the segments repetition and angular deviation on the crossing of the occlusion stroke. To validate our approach, we tested it on the on/off LMCA dual Arabic handwriting, the Latin IRONOFF and the off-line IFN/ENIT datasets.

This paper presents an off-line Arabic handwriting recognition system based on the selection of different state of the art features and the combination of multiple hidden Markov models classifiers. Beside the classical use of the off-line features, we add the use of on-line features and the combination of the developed systems. The designed recognizer is implemented using the HMM-Toolkit. In a first step, we use different features to make the classification and we compare the performance of single classifiers. In a second step, we proceed to the combination of the on-line and the off-line based systems using different combination methods. The system is evaluated using the IFN/ENIT database. The recognition rate is in maximum 63.90% for the individual systems. The combination of the on-line and the off-line systems allows to improve the system accuracy to 81.93% which exceeds the best result of the ICDAR 2005 competition.

In this paper, we present a new design for an interactive information service based on on-line recognition of the handwriting and quick news stories browsing. A person communicates with server PC using PDA and Bluetooth headset technology in order to consult same key frame that represent a summaries of video news. The result of the server research will by returned to the PDA.

In this article, we present initially arguments supporting the idea of the approximation of a cursive handwriting trajectory by arcs of ellipses. Then, we introduce a new strategy which improves the dynamic and geometrical features of the online handwritten trajectory modeling. We show that the curvilinear velocity can be rebuilt with the superposition of two components successively named the "Beta" model and the "Carrying" dragged component. After that, we integrated the geometrical characteristics as the arcs of ellipses for the layout modeling.