With the interconnection of roads spanning not only cities but also multiple countries, transportation capabilities have significantly improved. This has resulted in vehicles being able to reach far-off destinations within shorter timeframes. However, this progress presents challenges in identifying and tracking vehicles, particularly high-security or pursuit vehicles involved in crimes like theft, as they move across various security operational centers. Traditionally, centralized data access through traffic zones has been used for monitoring, surveillance, and tracking purposes. Nevertheless, this approach is susceptible to manipulation and compromises privacy due to unauthorized access by limited stakeholders. To address these concerns, a decentralized alternative is proposed, utilizing a blockchain network that ensures integrity, confidentiality, and access control for vehicle surveillance. Furthermore, computer vision algorithms can be employed for automated surveillance, enabling discreet re-identification or tracking of vehicles. This paper proposes an alternate framework that diverges from the centralized solution, instead encompassing multiple surveillance zones connected through a blockchain-based design with controlled access. To achieve this, we consider traditional approaches to vehicle re-identification using image similarity matching as opposed to resource-intensive unsupervised models performed on a federated consortium blockchain, the Hyperledger fabric. The framework is evaluated based on latency, resource consumption, and the effectiveness of image processing algorithms for feature sharing through blockchain. Compared to existing frameworks with similar applications, our proposed framework demonstrates higher accuracy and shorter inference times while adhering to the specified constraints. It leverages the concept of Blockchain in multi-surveillance zones to enhance privacy and security in automated vehicle surveillance. • A framework for multi-zonal surveillance using a federated consortium blockchain using hyperledger fabric. • A comparison of vehicle make and model classification techniques enabled on the blockchain. • A comparison of multi-zonal vs single zone setup.

Abstract It is an immensely challenging task to segment the tissue regions and cells in histological images of breast cancer with precision, but the results of this task are extremely significant for the field of computational pathology as a whole. To address this challenge, the integration of the Denoising Diffusion Probabilistic Model (DDPM) and the Generative Adversarial Network (GAN) has been explored. Specifically, we employ a conditional DDPM as the generator within the GAN framework, alongside a conditional adversarial network serving as the discriminator, to achieve segmentation of breast cancer histology images both at the regional and cellular levels. The forward process of the DDPM is first applied to the image mask. As the noise is added step by step, it is conditioned with the pathological image and estimated by a denoising network. To improve the estimated noise, the estimated noise is again conditioned with the pathological image and fed into the discriminator as part of the training process. As part of the test phase, a noise image conditioned with a pathological image is fed into a denoising model trained taking into consideration each time step, which segments the images into regions and cells in the reverse process. Three datasets were used for the experiments, one at a regional level and two at a cellular level. This method outperforms both GAN and diffusion models, as well as current state-of-the-art methods. Specifically, our method shows notable improvements in terms of Dice and IoU metrics over existing state-of-the-art methods.

This paper presents a comparative study of two widely used convolutional neural network (CNN) fam-ilies-Residual and Efficient architectures-for disaster type classification using social media images. The goal is to evaluate the robustness of these architectures to strong data augmentations and to examine how augmentation strength influences key performance metrics. ResNet50 and EfficientNetB1 were chosen for this study based on their strong benchmark performance on the MEDIC dataset for disaster type classification. We manually designed a custom augmentation pipeline inspired by RandAugment, selecting transformations that maintain critical visual features in disaster imagery. Results show that EfficientNetB1 slightly outperformed ResNet50 in accuracy (74.22% vs. 73.88 %), while ResNet50 shows better performance on underrepresented disaster classes, as reflected by minority class F1 score. We also observed that while stronger augmentation improved accuracy (+0.45 %), weaker augmentation yielded a slightly higher F1-score (+0.17 %) on EfficientNetB1.

Breast cancer detection in mammographic images is critical for early diagnosis and treatment. In modern medical AI systems, performance, computational efficiency, and explainability are key determinants of clinical adoption. This study evaluates three convolutional neural network (CNN) models, ResNet-50 ($\mathbf{4 9. 7 M}$ parameters), EfficientNet-B0 ($\mathbf{2 0. 4 M}$), and EfficientNetB7 (96.7 M), each integrated with Grad-CAM for visual explanation. The models are trained and tested on the MIAS dataset following structured preprocessing and data augmentation. GradCAM is employed to generate heatmaps that highlight the image regions most influential to each prediction, enhancing interpretability and clinician trust. Among the models, EfficientNetB0 achieves the best trade-off between accuracy and efficiency, reaching $98.71 \%$ accuracy with minimal computational overhead, making it highly suitable for resource-constrained deployments. EfficientNet-B7 achieves a slightly lower accuracy of $97.00 \%$ but requires substantially more computational resources. ResNet-50 attains 94.85% accuracy and delivers the fastest inference time, offering a viable baseline for time-sensitive applications. These results provide practical guidance for selecting CNN architectures that balance diagnostic performance, computational cost, and explainability in real-world clinical workflows.

In disaster response and search and rescue operations, the immediate detection of survivors remains a critical challenge. This research pioneers a novel approach to rapidly detect survivors in disaster areas using UAVs and advanced computer vision techniques. Leveraging the YOLOv8 object detection model, the study explores how synthetic and real disaster-specific datasets, alongside transfer learning, enhance survivor detection capabilities. Results show promising adaptability, with the YOLOv8 model achieving an average precision (AP) of 0.864, marking a significant 32% improvement over the previous state-of-the-art (SOTA) performance of 0.654 achieved with a slower model. Furthermore, the combination of fine-tuning a pre-trained model on the newly built dataset surpassed, by a small margin, even the standard training method despite utilizing only half the number of epochs. Additionally, this paper proposes a UAV-based system model that integrates computer vision for rapid onsite detection, potentially revolutionizing disaster response frameworks. This paper highlights the potential of UAV technology and transfer learning in improving disaster management and guides future investigations in this critical field.

This study introduces an innovative methodology for accurately classifying physiotherapy exercises, integrating Pose Estimation and diverse Machine Learning (ML) techniques within the alwaysAI framework. The workflow includes data preprocessing, feature normalization, feature extraction, exploration of ML techniques, model training, and evaluation using the accuracy metric, applied to eight diverse exercise datasets. Unlike traditional approaches relying solely on Support Vector Machines (SVM), this study explores a range of ML techniques adaptable to high-dimensional data, showcasing the effectiveness of the proposed methodology. The results demonstrate precise exercise classification based on pose information, affirming the robustness of the approach and highlighting its potential integration into physiotherapy practices. This research contributes to advancing technology-driven solutions in healthcare by emphasizing the versatility of combining pose estimation with ML techniques for precise physiotherapy exercise classification.

Over the last decade, videos uploaded and shared through web-based multimedia platforms and mobile applications have proliferated worldwide. This is because cloud-based applications such as iCloud, YouTube, Facebook, Twitter, and WhatsApp offer affordable and secure environments for video storage and sharing. However, new challenges have emerged alarming forensic analysts and investigators since videos can be used to commit heinous crimes such as blackmail, fraud, and forgery. Source Camera Identification (SCI) has become of paramount importance in the field of image and video forensics. Camera model identification can also help identify the perpetrators or narrow down the search and can be used to enhance SCI systems. In this context, existing approaches such as the Photo Response Non-Uniformity (PRNU) based methods and machine learning techniques such as the support vector machine (SVM) and deep learning models are commonly used solutions. This work exploits these two categories of methods by exploring a hierarchical deep learning model for camera model identification based on smartphone videos. The PRNU features are extracted by CNN-based structures during the training process. Proposed six-stream networks are leveraged to extract both low-level and high-level features through the network. A fusion layer is created based on joint sparse representation using forward and backward functions defined for fusing the proposed six streams. The proposed approach has been implemented and evaluated through intensive experiments, and results showed successful camera model identification with a performance at the frame level reaching an average accuracy of 69.9% for the Daxing dataset and 81.6% for the QUFVD dataset.

Abstract Learning disabilities, which primarily interfere with basic learning skills such as reading, writing, and math, are known to affect around 10% of children in the world. The poor motor skills and motor coordination as part of the neurodevelopmental disorder can become a causative factor for the difficulty in learning to write (dysgraphia), hindering the academic track of an individual. The signs and symptoms of dysgraphia include but are not limited to irregular handwriting, improper handling of writing medium, slow or labored writing, unusual hand position, etc. The widely accepted assessment criterion for all types of learning disabilities including dysgraphia has traditionally relied on examinations conducted by medical expert. However, in recent years, artificial intelligence has been employed to develop diagnostic systems for learning disabilities, utilizing diverse modalities of data, including handwriting analysis. This work presents a review of the existing automated dysgraphia diagnosis systems for children in the literature. The main focus of the work is to review artificial intelligence-based systems for dysgraphia diagnosis in children. This work discusses the data collection method, important handwriting features, and machine learning algorithms employed in the literature for the diagnosis of dysgraphia. Apart from that, this article discusses some of the non-artificial intelligence-based automated systems. Furthermore, this article discusses the drawbacks of existing systems and proposes a novel framework for dysgraphia diagnosis and assistance evaluation.

Abstract In multiple-instance learning (MIL), the existing bag encoding and attention-based pooling approaches assume that the instances in the bag have no relationship among them. This assumption is unsuited, as the instances in the bags are rarely independent in diverse MIL applications. In contrast, the instance relationship assumption-based techniques incorporate the instance relationship information in the classification process. However, in MIL, the bag composition process is complicated, and it may be possible that instances in one bag are related and instances in another bag are not. In present MIL algorithms, this relationship assumption is not explicitly modeled. The learning algorithm is trained based on one of two relationship assumptions (whether instances in all bags have a relationship or not). Hence, it is essential to model the assumption of instance relationships in the bag classification process. This paper proposes a robust approach that generates vector representation for the bag for both assumptions and the representation selection process to determine whether to consider the instances related or unrelated in the bag classification process. This process helps to determine the essential bag representation vector for every individual bag. The proposed method utilizes attention pooling and vision transformer approaches to generate bag representation vectors. Later, the representation selection subnetwork determines the vector representation essential for bag classification in an end-to-end trainable manner. The generalization abilities of the proposed framework are demonstrated through extensive experiments on several benchmark datasets. The experiments demonstrate that the proposed approach outperforms other state-of-the-art MIL approaches in bag classification.

The domain of image classification has been seen to be dominated by high-performing deep-learning (DL) architectures. However, the success of this field, as seen over the past decade, has resulted in the complexity of modern methodologies scaling exponentially, commonly requiring millions of parameters. Quantum computing (QC) is an active area of research aimed toward greatly reducing problems of complexity faced in classical computing. With growing interest toward quantum machine learning (QML) for applications of image classification, many proposed algorithms require usage of numerous qubits. In the noisy intermediate-scale quantum (NISQ) era, these circuits may not always be feasible to execute effectively; therefore, we should aim to use each qubit as effectively and efficiently as possible, before adding additional qubits. This article proposes a new single-qubit-based deep quantum neural network for image classification that mimics traditional convolutional neural network (CNN) techniques, resulting in a reduced number of parameters compared with previous works. Our aim is to prove the concept of the initial proposal by demonstrating classification performance of the single-qubit-based architecture, as well as to provide a tested foundation for further development. To demonstrate this, our experiments were conducted using various datasets including MNIST, Fashion-MNIST, and ORL face datasets. To further our proposal in the context of the NISQ era, our experiments were intentionally conducted in noisy simulation environments. Initial test results appear promising, with classification accuracies of 94.6%, 89.5%, and 82.5% achieved on the subsets of MNIST, FMNIST, and ORL face datasets, respectively. In addition, proposals for further investigation and development were considered, where it is hoped that these initial results can be improved.

A major challenge in brain tumor treatment planning is determination of the tumor extent. Brain tumor disease can be identified with imaging techniques such as MRI. The images produced by an MRI scan can provide a clear view of the brain's internal structures, including the presence of any abnormal growths or tumors. Tumors can be seen on the images as areas of abnormal tissue that have a different signal intensity from normal brain tissue. In addition, the MRI images can also provide information about the size, shape, location, and characteristics of the tumor, such as its blood flow and whether it is solid or cystic. This information can be very helpful in determining the best course of treatment for the patient. The main objective of this paper is to recognize the existence of tumors in the brain from MRI images using machine learning techniques. Our results show that the k-nearest approach is capable of precisely identifying brain cancers with more than 97%.

This paper introduces a database of multi-script (Arabic and English) for indoor scene text detection, taken from different angle-of-view. This database can be used in a variety of real-world applications, such as image search, robot navigation, and assisting the visually impaired. The database contains 944 images taken with smartphones in an indoor environment at Qatar University. These images were taken from at least three angles, making the database even more challenging. To evaluate the database, an OCR method based on multiple language detection is considered. The results show that multi-language detection should be given more attention in practice. The database is publicly available11https:/www.dropbox.com/s/7s7f936y4etzsu7/QU_door_dataset%20%282%29.zip?dl=0.

Governments around the world are increasingly using digital transformation to improve their services and operations. Digital maturity assessment models however are mainly designed for business organizations, and there are significant differences between private and government sectors, which means that such models may not be effective in the government context. This paper reevaluates the factors for assessing digital transformation maturity from the perspective of government institutes, integrating factors selected from maturity assessment models of recognized consultancy firms and leveraging insights from Subject Matter Experts (SMEs) in government organizations to prove their viability for government sector. A comprehensive set of assessment factors relevant to digital transformation maturity are systematically identified and subjected to evaluation by the SMEs, to reconsider their weights accordingly taking into account experiences and practical considerations unique to government institutes to enhance its relevance and applicability in real-context settings. The resulting Digital Transformation Maturity Assessment factors serve as a valuable tool for government organizations to guide strategic initiatives towards the successful implementation of digital transformation initiatives.

Image and video dehazing is a difficult subject that has received a lot of attention in the field of computer vision. The presence of air haze in photos and movies can reduce visual quality dramatically, resulting in a loss of contrast, color accuracy, and sharpness. To address this issue, in this paper, we propose a deep-learning-based method for image dehazing. The proposed network consists of using multi-scale representation at every VGG-16 block to conserve the high quality of the image during the learning process. The collaboration of convolutional layers and the multi-scale block make the network learn from different scales combined with the outputs of the previous layers of the networks. This can conserve the high quality as well as remove the haze. The proposed method is trained and tested on four datasets including BESIDE, DENSE, O-HAze, and I-HAZE, and hives promising results compared to some of the state-of-the-art methods.

A key area in computer vision and biometric authentication systems is detecting and classifying face anti-spoofing. The novel method for face anti-spoofing presented in this paper focuses on color invariant methods. The proposed approach consists to compare two different feature extraction methods: LBP and HOG, used with two different machine-learning models such as KNN and SVM. The suggested method improves the system's ability to discriminate by utilizing color properties and overcoming obstacles like varying lighting and image quality. The structure-texture decomposition is used as a feature that is used to improve the performance of the face anti-spoofing methods. After the experimental results using different techniques, we noticed that structure-texture decomposition can be a good feature for face anti-spoofing detection features.

Video Source device identification plays a crucial role in video forensics as the proliferation of video capturing devices has given rise to crimes with videos that are challenging to trace. Reliance on metadata extraction is insufficient as it can be corrupted or manipulated to conceal the source of the crime. Another technique employed for source identification is noise pattern extraction, which generates a unique identification for the video camera. However, this method is susceptible to capture faults and can produce diverse noise patterns for each video. In addressing these challenges, there is a need to identify distinctive features that are consistent across all videos captured by the same camera. This has led to the adoption of computer vision techniques utilizing machine learning and deep learning. Classifiers play a crucial role in machine learning and data analysis, as they are responsible for categorizing or predicting results based on input data. Our experiments show that the subject is sensitive to classifiers and developing a good classifier or classifier-level fusions can improve results in practice for all datasets.

Traffic problems constitute one of the major issues addressed worldwide. Some universities with an increasing number of students moving at a fast pace face transportation problems almost every day. Hence, this paper aims to solve this problem by developing an alternative transportation service, “Smart Scooter,” which is a scooter that can be driven autonomously. The Smart Scooter is used to serve the purpose of reducing the number of problems caused by traffic within campuses. Sensors associated with other hardware components are used to accomplish the goal of this project. An application is also created to assist the users in accessing the scooter's functions. Additionally, a survey is also conducted to gather the feedback of students and people in general to better understand their needs and demands related to these types of services. Testing the scooter for different routes shows an average error of 4.8% in reaching the final destination and 100% accuracy in obstacle detection at the front.

Recently, developing automated video surveillance systems (VSSs) has become crucial to ensure the security and safety of the population, especially during events involving large crowds, such as sporting events. While artificial intelligence (AI) smooths the path of computers to think like humans, machine learning (ML) and deep learning (DL) pave the way more, even by adding training and learning components. DL algorithms require data labeling and high-performance computers to effectively analyze and understand surveillance data recorded from fixed or mobile cameras installed in indoor or outdoor environments. However, they might not perform as expected, take much time in training, or not have enough input data to generalize well. To that end, deep transfer learning (DTL) and deep domain adaptation (DDA) have recently been proposed as promising solutions to alleviate these issues. Typically, they can (i) ease the training process, (ii) improve the generalizability of ML and DL models, and (iii) overcome data scarcity problems by transferring knowledge from one domain to another or from one task to another. Although the increasing number of articles proposed to develop DTL- and DDA-based VSSs, a thorough review that summarizes and criticizes the state-of-the-art is still missing. To that end, this paper introduces, to the best of the authors’ knowledge, the first overview of existing DTL- and DDA-based video surveillance to (i) shed light on their benefits, (ii) discuss their challenges, and (iii) highlight their future perspectives.

Dysgraphia is a neurological disorder that hinders the acquisition process of normal writing skills in children, resulting in poor writing abilities. Poor or underdeveloped writing skills in children can negatively impact their self-confidence and academic growth. This work proposes various machine learning methods, including transfer learning via fine-tuning, transfer learning via feature extraction, ensembles of deep convolutional neural network (CNN) models, and fusion of CNN features, to develop a preliminary dysgraphia diagnosis system based on handwritten images. In this work, an existing online dysgraphia dataset is converted into images, encompassing various writing tasks. Transfer learning is applied using a pre-trained DenseNet201 network to develop four distinct CNN models separately trained on word, pseudoword, difficult word, and sentence images. Soft voting and hard voting strategies are employed to ensemble these CNN models. The pre-trained DenseNet201 network is used for CNN feature extraction from each task-specific handwritten image data. The extracted CNN features are then fused in different combinations. Three machine learning algorithms support vector machine (SVM), AdaBoost, and Random forest are employed to assess the performance of the CNN features and fused CNN features. Among the task-specific models, the SVM trained on word data achieved the highest accuracy of 91.7%. In the case of ensemble learning, soft voting ensembles of task-specific CNNs achieved an accuracy of 90.4%. The feature fusion approach substantially improved the classification accuracy, with the SVM trained on fused features from the task specific-data achieving an accuracy of 97.3%. This accuracy surpasses the performance of state-of-the-art methods by 16%.

Abstract Video steganography approach enables hiding chunks of secret information inside video sequences. The features of video sequences including high capacity as well as complex structure make them more preferable for choosing as cover media over other media such as image, text, or audio. Video steganography is a prominent as well as the evolving field in the information security domain and significant number of video steganography methods are proposed in recent years. This article provides a comprehensive review of video steganography methods proposed in the literature. This article initially reviews various raw domain-based video steganography methods. In particular, the raw domain-based methods include spatial domain approaches such as least significant bits (LSB), transform domain-based methods such as discrete wavelet transform, discrete cosine transform, etc. Furthermore, the article looks into various compressed domain steganography methods. A critical comparative analysis is included in the article to analyze and contrast the steganography methods proposed in the literature. A brief description of various evaluation matrices for video steganography methods is provided in this article. Moreover, a brief introduction to steganalysis and video steganalysis is provided. The article concludes with a discussion focused on the limitations and challenges of the video steganography methods. Further, a brief insight into future directions in video steganography systems is provided.

Abstract Feature-level-based fusion has attracted much interest. Generally, a dataset can be created in different views, features, or modalities. To improve the classification rate, local information is shared among different views by various fusion methods. However, almost all the methods use the views without considering their common aspects. In this paper, wavelet transform is considered to extract high and low frequencies of the views as common aspects to improve the classification rate. The fusion method for the decomposed parts is based on joint sparse representation in which a number of scenarios can be considered. The presented approach is tested on three datasets. The results obtained by this method prove competitive performance in terms of the datasets compared to the state-of-the-art results.

People counting is one of the computer vision tasks that can be useful for crowd management. In addition, estimating the crowdedness of a surveilled scene for crowd behavior analysis is one of the prominent challenges in video surveillance systems. With the introduction of deep learning, this operation has become doable with a convincing performance. However, this task still represents a challenge for these methods. In this regard, we propose a combination of deep reinforcement learning (DRL) networks and deep learning architecture for crowd counting. DRL network used the Context-Aware Attention (CAA) module for segmenting the crowd region, Then, on the segmented results, the crowd density estimation is performed using an encoder-decoder. The proposed method is evaluated and compared with and without the segmentation parts on the existing datasets including UCF_QNRF, UCF_CC_50, ShangaiTech_(A, B), while the obtained results in terms of MAE metric achieved 84,8, 179.2, 44.6, and 8.2 respectively.

Photo response non-uniformity (PRNU) noise is a sensor pattern noise characterizing imperfections in the imaging device. The PRNU is a unique noise for each sensor device, and it has been generally utilized in the literature for source camera identification and image authentication. In video forensics, the traditional approach estimates the PRNU by averaging a set of residual signals obtained from multiple video frames. However, due to lossy compression and other non-unique content-dependent noise components that interfere with the video data, constant averaging does not take into account the intensity of these undesirable noise components which are content-dependent. Different from the traditional approach, we propose a video PRNU estimation method based on weighted averaging. The noise residual is first extracted for each single video. Then, the estimated noise residuals are fed into a weighted averaging method to optimize PRNU estimation. Experimental results on two video datasets captured by various smartphone devices have shown a significant gain obtained with the proposed approach over the conventional state-of-the-art one.

Recent advances in digital imaging have meant that every smartphone has a video camera that can record high-quality video for free and without restrictions. In addition, rapidly developing Internet technology has contributed significantly to the widespread distribution of digital video via web-based multimedia systems and mobile applications such as YouTube, Facebook, Twitter, WhatsApp, etc. However, as the recording and distribution of digital video has become affordable nowadays, security issues have become threatening and have spread worldwide. One of the security issues is the identification of source cameras on videos. Generally, two common categories of methods are used in this area, namely Photo Response Non-Uniformity (PRNU) and Machine Learning approaches. To exploit the power of both approaches, this work adds a new PRNU-based layer to a convolutional neural network (CNN) called PRNU-Net. To explore the new layer, the main structure of the CNN is based on the MISLnet, which has been used in several studies to identify the source camera. The experimental results show that the PRNU-Net is more successful than the MISLnet and that the PRNU extracted by the layer from low features, namely edges or textures, is more useful than high and mid-level features, namely parts and objects, in classifying source camera models. On average, the network improves the results in a new database by about 4%.

The education of children with learning difficulties is a challenging task especially during COVID-19 pandemic. In fact, these children need to go regularly to specialized school, receive focused education, and interact with teachers to learn. Instructors allocate important time to teach them and use different approaches including, attractive stories, tangible photos, physical plays, sites visit, awards and gifts. However, these modes of education become nowadays very hard to achieve as instructors are teaching from homes or offices through the Internet. They cannot have face-to-face meetings with children in classrooms. The major issue is how to find the necessary materials to teach them the new Arabic vocabularies and explain their meanings in an effective manner. Instructors can use textbooks, online libraries and search engines looking for Arabic educational resources while most of them are in English or in other western languages. The process is very long, time consuming and does not fill-in the gap. We propose to build a new educational large-scale multimedia Arabic corpus that provides thousands of vocabularies and chunks associated with best representative images. The instructors can use the bimodal corpus directly during the learning sessions to explain new Arabic words through images. It currently covers the animals' domain and contains thousands of well-structured and interconnected entities. Instructors can collaborate in enhancing the corpus by adding new materials through a web-based platform and build then rich source of learning materials.

COVID-19 imposed a new paradigm in education especially in elementary schools. Children are no longer able to go regularly to schools as normal. They should then rely mainly on themselves to learn and acquire knowledge. The Understand My World app is a new technological solution that allows children to learn new Arabic vocabularies interactively and independently using their smart phones or tablets (i.e., iPad, iPhone). They can use the devices’ camera and microphone to explore the world, understand spoken words, and read written language properly. Images captured by the camera are recognized, labeled, and defined in both speech and writing. Children can also speak into their device to record their speech and listen to it. They can address questions and receive answers. The app then presents the dictated words in both spoken and written form. Using the required Internet connection, the interface automatically provides exceedingly accurate image and speech recognition, all while preserving the participants’ anonymity. We use two AI platforms namely Clarifai and Houndify. The application is designed to be intuitive and seamless to use, which makes it very attractive to children to learn through multimedia.

Abstract Human action recognition and video summarization represent challenging tasks for several computer vision applications including video surveillance, criminal investigations, and sports applications. For long videos, it is difficult to search within a video for a specific action and/or person. Usually, human action recognition approaches presented in the literature deal with videos that contain only a single person, and they are able to recognize his action. This paper proposes an effective approach to multiple human action detection, recognition, and summarization. The multiple action detection extracts human bodies’ silhouette, then generates a specific sequence for each one of them using motion detection and tracking method. Each of the extracted sequences is then divided into shots that represent homogeneous actions in the sequence using the similarity between each pair frames. Using the histogram of the oriented gradient (HOG) of the Temporal Difference Map (TDMap) of the frames of each shot, we recognize the action by performing a comparison between the generated HOG and the existed HOGs in the training phase which represents all the HOGs of many actions using a set of videos for training. Also, using the TDMap images we recognize the action using a proposed CNN model. Action summarization is performed for each detected person. The efficiency of the proposed approach is shown through the obtained results for mainly multi-action detection and recognition.

Abstract Codebook‐based writer characterization is an effective technique that has been investigated in a number of recent studies on identification and verification of writers. These methods divide a set of writing samples into small units (fragments or graphemes) and cluster these patterns to produce a codebook. Writer of a handwritten sample is then characterized by the probability (distribution) of producing the codebook patterns. In most cases, a small subset of the database under study is employed to produce the codebook while the rest of the database is used in evaluations. This work aims to validate the hypothesis that the codebook simply serves as a representation space to compare different writings and, in most cases, the patterns in the codebook do not significantly influence the identification and verification performance. The hypothesis is validated by generating a number of codebooks using Greek, Arabic and Chinese handwritten samples. Moreover, codebooks using fragments of handwritten music scores, printed text and synthetic data are also investigated. Evaluations on three well‐known handwriting databases (CVL, BFL and IAM) validate the idea that, in general, the codebook patterns do not have a significant impact on characterizing writer from handwriting.

Abstract People with blindness or low vision utilize mobile assistive tools for various applications such as object recognition, text recognition, etc. Most of the available applications are focused on recognizing generic objects. And they have not addressed the recognition of food dishes and fruit varieties. In this paper, we propose a smartphone-based system for recognizing the food dishes as well as fruits for children with visual impairments. The Smartphone application utilizes a trained deep CNN model for recognizing the food item from the real-time images. Furthermore, we develop a new deep convolutional neural network (CNN) model for food recognition using the fusion of two CNN architectures. The new deep CNN model is developed using the ensemble learning approach. The deep CNN food recognition model is trained on a customized food recognition dataset.The customized food recognition dataset consists of 29 varieties of food dishes and fruits. Moreover, we analyze the performance of multiple state of art deep CNN models for food recognition using the transfer learning approach. The ensemble model performed better than state of art CNN models and achieved a food recognition accuracy of 95.55 % in the customized food dataset. In addition to that, the proposed deep CNN model is evaluated in two publicly available food datasets to display its efficacy for food recognition tasks.

Abstract The automatic generation of realistic images directly from a story text is a very challenging problem, as it cannot be addressed using a single image generation approach due mainly to the semantic complexity of the story text constituents. In this work, we propose a new approach that decomposes the task of story visualization into three phases: semantic text understanding, object layout prediction, and image generation and refinement. We start by simplifying the text using a scene graph triple notation that encodes semantic relationships between the story objects. We then introduce an object layout module to capture the features of these objects from the corresponding scene graph. Specifically, the object layout module aggregates individual object features from the scene graph as well as averaged or likelihood object features generated by a graph convolutional neural network. All these features are concatenated to form semantic triples that are then provided to the image generation framework. For the image generation phase, we adopt a scene graph image generation framework as stage-I, which is refined using a StackGAN as stage-II conditioned on the object layout module and the generated output image from stage-I. Our approach renders object details in high-resolution images while keeping the image structure consistent with the input text. To evaluate the performance of our approach, we use the COCO dataset and compare it with three baseline approaches, namely, sg2im, StackGAN and AttnGAN, in terms of image quality and user evaluation. According to the obtained assessment results, our object layout guidance-based approach significantly outperforms the abovementioned baseline approaches in terms of the accuracy of semantic matching and realism of the generated images representing the story text sentences.

Abstract Person re-identification across multiple cameras is an essential task in computer vision applications, particularly tracking the same person in different scenes. Gait recognition, which is the recognition based on the walking style, is mostly used for this purpose due to that human gait has unique characteristics that allow recognizing a person from a distance. However, human recognition via gait technique could be limited with the position of captured images or videos. Hence, this paper proposes a gait recognition approach for person re-identification. The proposed approach starts with estimating the angle of the gait first, and this is then followed with the recognition process, which is performed using convolutional neural networks. Herein, multitask convolutional neural network models and extracted gait energy images (GEIs) are used to estimate the angle and recognize the gait. GEIs are extracted by first detecting the moving objects, using background subtraction techniques. Training and testing phases are applied to the following three recognized datasets: CASIA-(B), OU-ISIR, and OU-MVLP. The proposed method is evaluated for background modeling using the Scene Background Modeling and Initialization (SBI) dataset. The proposed gait recognition method showed an accuracy of more than 98% for almost all datasets. Results of the proposed approach showed higher accuracy compared to obtained results of other methods result for CASIA-(B) and OU-MVLP and form the best results for the OU-ISIR dataset.

Due to the poor condition of most of historical documents, binarization is difficult to separate document image background pixels from foreground pixels. This paper proposes Convolutional Neural Networks (CNNs) based on predicted two-channel images in which CNNs are trained to classify the foreground pixels. The promising results from the use of multispectral images for semantic segmentation inspired our efforts to create a novel prediction-based two-channel image. In our method, the original image is binarized by the structural symmetric pixels (SSPs) method, and the two-channel image is constructed from the original image and its binarized image. In order to explore impact of proposed two-channel images as network inputs, we use two popular CNNs architectures, namely SegNet and U-net. The results presented in this work show that our approach fully outperforms SegNet and U-net when trained by the original images and demonstrates competitiveness and robustness compared with state-of-the-art results using the DIBCO database.

In this paper a video summarization method based on motion detection has been proposed. Sensor noise (noise of acquisition and digitization) and the illumination changes in the scene are the most limitations of the background subtraction approaches. In order to handle these problems, this paper present an approach based on the combining of the background subtraction and the Structure-Texture-Noise Decomposition. Firstly, each gray-level image of the sequence will be decomposed on three components, Structure, Texture and Noise. The Structure and Texture components of each image of the sequence are taken to generate the background model. The absolute difference used to subtract the background before compute the binary image of moving objects. We, also, propose a video summarization based on the background subtraction results. The generated background model is used to compute the change during all time of the sequence. The experimental results demonstrate that our approach is effective and accurate for moving objects detection and yields a good summarization of the video sequence.

Moving object detection is a fundamental step in various computer vision applications. Robust Principal Component Analysis (RPCA) based methods have often been employed for this task. However, the performance of these methods deteriorates in the presence of dynamic background scenes, camera jitter, camouflaged moving objects, and/or variations in illumination. It is because of an underlying assumption that the elements in the sparse component are mutually independent, and thus the spatiotemporal structure of the moving objects is lost. To address this issue, we propose a spatiotemporal structured sparse RPCA algorithm for moving objects detection, where we impose spatial and temporal regularization on the sparse component in the form of graph Laplacians. Each Laplacian corresponds to a multi-feature graph constructed over superpixels in the input matrix. We enforce the sparse component to act as eigenvectors of the spatial and temporal graph Laplacians while minimizing the RPCA objective function. These constraints incorporate a spatiotemporal subspace structure within the sparse component. Thus, we obtain a novel objective function for separating moving objects in the presence of complex backgrounds. The proposed objective function is solved using a linearized alternating direction method of multipliers based batch optimization. Moreover, we also propose an online optimization algorithm for real-time applications. We evaluated both the batch and online solutions using six publicly available datasets that included most of the aforementioned challenges. Our experiments demonstrated the superior performance of the proposed algorithms compared with the current state-of-the-art methods.

Palmprint recognition is an important and widely used biometric modality with high reliability, stability and user acceptability. In this paper we propose a simple and effective ensemble learning method for palmprint identification based on Random Subspace Sampling (RSS). To achieve it, we rely on 2D-PCA to build the random subspaces. As 2D-PCA is an unsurpevised technique, features are extracted in each subspace using 2D-LDA. A simple 1-Nearest Neighbor classifier is associated to each subspace, the final decision rule being obtained by majority voting rule. The experimental results on multispectral and PolyU palmprint datasets show very encouraging performances compared to state-of-the-art techniques.

This paper describes the ICFHR 2018 Competition on Multi-script Writer Identification with details on the competition tasks, databases employed, submitted systems, evaluation protocol and the reported results. The competition was aimed at exploring the traditional writer identification problem in a more challenging scenario of a multi-script environment where training and test samples of writers come from different scripts. Three different databases with handwriting samples in Arabic, French, English, Chinese and Farsi were employed in the six competition tasks. The realized results indicate that while high identification rates are reported in the literature by traditional writer identification systems, identifying writers in a multiscript environment is a much more challenging problem that requires significant investigations to extract effective handwriting representations that are able to characterize the writer across different scripts.

This study addresses the problem of identifying the authorship of historical manuscripts, a challenging task that offers interesting applications for document examiners and paleographers. We exploit handwriting texture as the discriminative attribute characterizing the writer of a given document. The textural information in handwriting is captured using a combination of oriented Basic Image Features (oBIFs) at different scales. Classification is carried out using a number of distance metrics which are combined to arrive at a final decision. A comprehensive series of experiments is carried out using different configurations of the oBIFs and the realized classification rates are compared with the state-of-the-art techniques on this problem

This paper presents a study on assessing the effectiveness of machine learned features to predict gender of writers from images of handwriting. Pre-trained Convolutional Neural Networks have been employed as feature extractors to discriminate male and female handwriting while classification is carried out using a number of classifiers, Linear Discriminant Analysis (LDA) being the most effective. Feature extraction is carried out by changing the scale of observation using word, patch and page images. Experiments are carried out on English and Arabic handwriting samples of the QUWI database and the realized results demonstrate the effectiveness of machine learned features in predicting gender from handwriting.

For most people from all ages and genders, participation in sports becomes part of their life, especially participation in soccer matches, which are considered a symbol of healthy living and active attitudes of families. Analyzing soccer matches, similar to analyzing other sports, is a challenging job for the coaches and trainers, as well as for the audience, due to the fast motion of players in some situations during the match and occlusions. That is why computer vision techniques are used to tackle these problems. In this paper, we test an efficient, simple and available annotating tool to label and generate the ground truth data of the interested targets (players or the ball) on a soccer field, which helps to assess the performance of the tracking techniques and achieve their goals. This method is tested on two different sequences of soccer datasets. The annotation results are tested by using four tracking algorithms based on context-aware correlation filters. The tracking results on both sequences that were annotated by the experts and annotated by this method were very similar, which shows that this robust method outperforms the state-of-the-art annotating techniques.

Separating content from noise in historical manuscripts is a fundamental task in digital palaeography. This study presents a fully automated segmentation approach based on the response of Harris corner detectors. The strength and clustering efficiency of the detected corners in the manuscripts are evaluated and used to segment the content from the background and noise. In addition, a manuscript reconstruction technique is proposed from the gradient field using the Poisson method to guide the interpolation. This reconstruction is able to remove noise significantly and hence enhances the contrast of the content thus making it easier for users to read and process these documents. The proposed approaches are evaluated using various standard databases to highlight their effectiveness and robustness to a multitude of noise and writing styles. Subjective and objective evaluations of the experimental results show that these techniques are able to successfully segment and reconstruct a very diverse set of scanned documents. An analysis of the results has also shown that the proposed technique compares favourably against similar counterparts.

The age of a historical manuscript can be an invaluable source of information for paleographers and historians. The process of automatic manuscript age detection has inherent complexities, which are compounded by the lack of suitable datasets for algorithm testing. This paper presents a dataset of historical handwritten Arabic manuscripts designed specifically to test state-of-the-art authorship and age detection algorithms. Qatar National Library has been the main source of manuscripts for this dataset while the remaining manuscripts are open source. The dataset consists of over 2000 images taken from various handwritten Arabic manuscripts spanning fourteen centuries. In addition, a sparse representation-based approach for dating historical Arabic manuscript is also proposed. There is lack of existing datasets that provide reliable writing date and author identity as metadata. KERTAS is a new dataset of historical documents that can help researchers, historians and paleographers to automatically date Arabic manuscripts more accurately and efficiently.

Radio frequency identification (RFID), the process of identifying objects within a predefined protocol definition technologies serves its primary application for tracking objects. The three main approaches followed are estimating tag range information, RFID localization techniques and artificial intelligence approach. The objective of this paper is to summarize and compare the major methods used in different RFID stages. Various techniques for RFID as well as localizations protocols are reviewed here.

Advancing technologies in the current era paved a lot to break the hurdles in medical diagnostic field. When cancer turned out to be the most common and dangerous disease of the age, novel diagnostic methodologies were introduced to enable early detection and hence save numerous lives. Accomplishment of various automatic and semi-automatic approaches in the diagnosis has proved its sufficient impetus to improve diagnostic speed and accuracy. A wide range of image processing based tools are currently available as a part of automatic cancer detection systems. Different imaging modalities have been utilized for extracting the suspected patient information, where the multispectral imaging has emerged as an efficient means for capturing the entire range of spectral and spatial data. In this paper, we review the current multispectral imaging based methods for automatic diagnosis of major types of cancer and discuss the limitations which are yet to be overcome, so as to improve the existing systems.

Automated grading of tumor cells proves to be a great way of enhancing the rapidity and accuracy of cancer diagnostic procedures. The application of image processing and machine learning techniques on the digitized biopsy slides enables the discrimination between various cell types. Apart from using the normal RGB/grayscale imaging, multispectral images tend to provide a wide range of information that can support the classification tasks. Besides using the visible range, wavelength bands in infrared ranges can be utilized in multispectral imaging. This paper presents our multispectral image acquisition system to develop a database for the colorectal biopsy slides. The dataset comprise images in both visible and near infrared spectrum for the 4 major categories of colon cells. A preprocessing algorithm with automatic estimation of parameters for adaptive histogram equalization is also presented. With the acquired database, our preliminary experiment involving a basic feature extraction and classification algorithm yielded satisfactory results.

Many real world complex systems naturally map to network data structures instead of geometric spaces because the only available information is the presence or absence of a link between two entities in the system. To enable data mining techniques to solve problems in the network domain, the nodes need to be mapped to a geometric space. We propose this mapping by representing each network node with its geodesic distances from all other nodes. The space spanned by the geodesic distance vectors is the geodesic space of that network. The position of different nodes in the geodesic space encode the network structure. In this space, considering a continuous density field induced by each node, density at a specific point is the summation of density fields induced by all nodes. We drift each node in the direction of positive density gradient using an iterative algorithm till each node reaches a local maximum. Due to the network structure captured by this space, the nodes that drift to the same region of space belong to the same communities in the original network. We use the direction of movement and final position of each node as important clues for community membership assignment. The proposed algorithm is compared with more than 10 state-of-the-art community detection techniques on two benchmark networks with known communities using Normalized Mutual Information criterion. The proposed algorithm outperformed these methods by a significant margin. Moreover, the proposed algorithm has also shown excellent performance on many real-world networks.

This competition is aimed at classification of writer demographics from offline handwritten documents using the QUWI database. QUWI is a bilingual database comprising writing samples of same individuals in Arabic and English. This allows evaluating the performance of different systems in a more challenging multi-script environment. This paper presents the details of the competition tasks, the datasets used in each of the tasks, a brief description of the participating systems, experimental protocol and evaluation criteria and finally the overall rankings of the participants.

Writer identification is an important field in forensic document examination. Typically, a writer identification system consists of two main steps: feature extraction and matching and the performance depends significantly on the feature extraction step. In this paper, we propose a set of novel geometrical features that are able to characterize different writers. These features include direction, curvature, and tortuosity. We also propose an improvement of the edge-based directional and chain code-based features. The proposed methods are applicable to Arabic and English handwriting. We have also studied several methods for computing the distance between feature vectors when comparing two writers. Evaluation of the methods is performed using both the IAM handwriting database and the QUWI database for each individual feature reaching Top1 identification rates of 82 and 87 % in those two datasets, respectively. The accuracies achieved by Kernel Discriminant Analysis (KDA) are significantly higher than those observed before feature-level writer identification was implemented. The results demonstrate the effectiveness of the improved versions of both chain-code features and edge-based directional features.

With the rapid development of Internet-of-Things (IoT), face scrambling has been proposed for privacy protection during IoT-targeted image/video distribution. Consequently, in these IoT applications, biometric verification needs to be carried out in the scrambled domain, presenting significant challenges in face recognition. Since face models become chaotic signals after scrambling/encryption, a typical solution is to utilize the traditional data-driven face recognition algorithms. While chaotic pattern recognition is still a challenging task, in this paper, we propose a new ensemble approach-many-kernel random discriminant analysis (MK-RDA)-to discover discriminative patterns from the chaotic signals. We also incorporate a salience-aware strategy into the proposed ensemble method to handle the chaotic facial patterns in the scrambled domain, where the random selections of features are made on semantic components via salience modeling. In our experiments, the proposed MK-RDA was tested rigorously on three human face data sets: the ORL face data set, the PIE face data set, and the PUBFIG wild face data set. The experimental results successfully demonstrate that the proposed scheme can effectively handle the chaotic signals and significantly improve the recognition accuracy, making our method a promising candidate for secure biometric verification in the emerging IoT applications.

This paper presents new class of time-frequency (T-F) features for automatic detection and classification of epileptic seizure activities in EEG signals. Most previous methods were based only on signal features derived from the instantaneous frequency and energies of EEG signals in different spectral sub-bands. The proposed features based on image descriptors are extracted from the T-F representation of EEG signals and are considered and processed as an image using T-F image processing techniques. The proposed features include shape and texture-based descriptors and are able to describe visually the normal and seizure activity patterns observed in T-F images. The results obtained on real EEG data show that T-F image descriptor-based features achieve an overall classification accuracy of up to 98% for 100 EEG segments using one-against-one SVM classifier. The results suggest that the proposed method outperforms those methods, which employ signal features only or combined signal-image features by about 3% for 100 EEG signals.

This competition targets writer identification and gender classification from offline handwritten documents using the QUWI database. The most interesting aspect of the competition is the use of a dataset with writing samples of the same individual in Arabic as well as English. The competition not only allows an objective comparison of different systems but also permits to investigate the performance of traditional script-dependent systems in a multi-script experimental setup. This paper describes the competition details including the competition tasks, the database employed, the methods used by the participating systems, evaluation and ranking criteria and the overall rankings of the participants. The competition received a total of 13 submissions from 8 different institutions. Writer identification tasks received 5 while the gender classification tasks received 8 submissions.

This paper presents novel time-frequency (t-f) feature extraction approach for the classification of EEG signals for Epileptic seizure activities detection. The proposed features are based on Local Binary Patterns (LBP) descriptor extracted from t-f representation of EEG signals processed as a textured image. Compared to most previous t-f approaches were based only on features derived from the instantaneous frequency and the energies of EEG signals generated from different spectral sub-bands, the proposed t-f features are capable to describe visually the epileptic seizure activity patterns observed in t-f image of EEG signals. The results obtained on real EEG data show that the use of t-f LBP descriptor-based features achieve an overall classification accuracy up to 99% for 150 EEG signals using 2-class SVM classifier. This is confirmed by ROC curve analysis.

Real-time traffic analysis using the road mounted surveillance cameras present multitude of benefits. This kind of traffic video processing has become an important means for intelligent traffic management and control. The estimation and analysis of road traffic motion is an involved task in computer vision and video processing. In our work, morphological operations and region growing method are used to perform salient motion detection of objects. In classical background extraction method, the background has to be learnt from large numbers of frames. In our method, no a prior knowledge about shape and size of object is acquired. Instead, sum of square difference is estimated via online learning for the calculation of the centroid distance. The test results indicate that the road vehicles and their statistics are determined through our algorithm with complete fidelity.

In this paper, we present a novel architecture based on field-programmable gate arrays (FPGAs) for the reconstruction of compressively sensed signal using the orthogonal matching pursuit (OMP) algorithm. We have analyzed the computational complexities and data dependence between different stages of OMP algorithm to design its architecture that provides higher throughput with less area consumption. Since the solution of least square problem involves a large part of the overall computation time, we have suggested a parallel low-complexity architecture for the solution of the linear system. We have further modeled the proposed design using Simulink and carried out the implementation on FPGA using Xilinx system generator tool. We have presented here a methodology to optimize both area and execution time in Simulink environment. The execution time of the proposed design is reduced by maximizing parallelism by appropriate level of unfolding, while the FPGA resources are reduced by sharing the hardware for matrix-vector multiplication across the data-dependent sections of the algorithm. The hardware implementation on the Virtex6 FPGA provides significantly superior performance in terms of resource utilization measured in the number of occupied slices, and maximum usable frequency compared with the existing implementations. Compared with the existing similar design, the proposed structure involves 328 more DSP48s, but it involves 25802 less slices and 1.85 times less computation time for signal reconstruction with N = 1024, K = 256, and m = 36, where N is the number of samples, K is the size of the measurement vector, and m is the sparsity. It also provides a higher peak signal-to-noise ratio value of 38.9 dB with a reconstruction time of 0.34 μs, which is twice faster than the existing design. In addition, we have presented a performance metric to implement the OMP algorithm in resource constrained FPGA for the better quality of signal reconstruction.

In this paper, we propose a novel approach for writer identification using codebook generation based on text skeletonization.Unlike other schemes, the skeleton in this approach is segmented at its junction pixels into elementary graphic units called graphemes. The codebook is generated by clustering the graphemes according to their distributions into a predefined grid. This method has been evaluated using the benchmarking dataset of the International Conference on Document Analysis and Recognition (ICDAR 2011) writer identification contest and has shown promising results. We also studied the effect of the amount of handwriting on the identification accuracy of the method and demonstrated that the proposed method is valid for Latin and Greek languages.

This paper presents new time-frequency (T-F) features to improve the detection and classification of epileptic seizure activities in EEG signals. Most previous methods were based only on signal features derived from the instantaneous frequency and energies of EEG signals generated from different spectral sub-bands. The proposed features are based on T-F image descriptors, which are extracted from the T-F representation of EEG signals, are considered and processed as an image using image processing techniques. The idea of the proposed feature extraction method is based on the application of Otsu's thresholding algorithm on the T-F image in order to detect the regions of interest where the epileptic seizure activity appears. The proposed T-F image related-features are then defined to describe the statistical and geometrical characteristics of the detected regions. The results obtained on real EEG data suggest that the use of T-F image based-features with signal related-features improve significantly the performance of the EEG seizure detection and classification by up to 5% for 120 EEG signals, using a multi-class SVM classifier.

Segmentation is considered as a core step for any recognition or classification method and for the text within any document to be effectively recognized it must be segmented accurately. In this paper a text and writer independent algorithm for the segmentation of sub-words in Arabic words has been presented. The concept is based around the global binarization of an image at various thresholding levels. When each sub-word or Part of Arabic Word (PAW) within the image being investigated is processed at multiple threshold levels a cluster graph is obtained where each cluster represents the individual sub-words of that word. Once the clusters are obtained the task of segmentation is managed by simply selecting the respective cluster automatically which is achieved using the 95% confidence interval on the processed data generated by the accumulated graph. The presented algorithm was tested on 537 randomly selected words from the AHTID/MW database and the results showed that 95.3% of the sub-words or PAW were correctly segmented and extracted. The proposed method has shown considerable improvement over the projection profile method which is commonly used to segment sub-words or PAW.

Face hallucination can be a useful tool for visualizing a low quality face into a visually better quality, making it an attractive technology for many applications. While faces in surveillance videos are usually at very low resolution, in this paper, we propose to use face hallucination technology to visualize faces from visual surveillance systems, and develop a weighted scheme to enhance the quality of face visualization from surveillance videos. Our experiment validated that in comparison with the classic eigenspace based face hallucination, our proposed weighted face hallucination strategy can help improve the overall quality of a facial image extracted from surveillance footage.

This paper proposes new time-frequency features for detecting and classifying epileptic seizure activities in non-stationary EEG signals. These features are obtained by translating and combining the most relevant time-domain and frequency-domain features into a joint time-frequency domain in order to improve the performance of EEG seizure detection and classification of non-stationary EEG signals. The optimal relevant translated features are selected according maximum relevance and minimum redundancy criteria. The experiment results obtained on real EEG data, show that the use of the translated and the selected relevant time-frequency features improves significantly the EEG classification results compared against the use of both original time-domain and frequency-domain features.

The prediction of gender from handwriting is a very interesting research field. However, no standard benchmark is available for researchers in this field. The aim of this competition is to gather researchers and compare recent advances in gender prediction from handwriting. This competition has been hosted on Kaggle, it has attracted 194 teams from both academia and industry. This paper gives details on this competition, including the dataset used, the evaluation procedure and description of participating methods and their performances.

Stroke recovery from offline handwriting is a very interesting research field. However, no standard benchmark is available for researchers in this field. The aim of this competition is to gather researchers and compare recent advances in stroke recovery from offline handwriting. This competition has been hosted on Kaggle, it has attracted 45 teams from both academia and industry. This paper gives details on this competition, including the dataset used, the evaluation procedure and description of participating methods and their performances.

This paper presents a new offline dataset called the Qatar University Writer Identification dataset (QUWI). This dataset contains both Arabic and English handwritings and can be used to evaluate the performance of offline writer identification systems. It consists of handwritten documents of 1017 volunteers of different ages, nationalities, genders and education levels. The writers were asked to copy a specific text and to generate a random text, which allows the dataset to be used for both text-dependent and text-independent writer identification tasks. We describe the gathering and processing steps and define several evaluation tasks regarding the use of this dataset.

Arabic writer identification is a very active research field. However, no standard benchmark is available for researchers in this field. The aim of this competition is to gather researchers and compare recent advances in Arabic writer identification. This competition has been hosted on Kaggle, it has attracted forty-three teams from both academia and industry. This paper gives details on this competition, including the dataset used, the evaluation procedure and description of participating methods and their performances.

Arabic writer identification is a very active research field. However, no standard benchmark is available for researchers in this field. The aim of this competition is to gather researchers and compare recent advances in Arabic writer identification. This competition was hosted by Kaggle, it has attracted thirty participants from both academia and industry. This paper gives details on this competition, including the evaluation procedure, description of participating methods and their performances.

A system for Arabic writer identification using grapheme features and k-nearest neighbor classifier is built using Matlab programming language. The results of our preliminary study reveal that unknown writers can be identified by using edge base directional features and text-dependent method; however the simple system approach needs improvement to satisfy the requirements of real data. This project works on the following improvements: First a database of text- independent Arabic handwritten pages from around 100 different writers is gathered and used as a test bed. Then, features will be extracted from writers' handwriting. Prior to feature extraction, preprocessing operations is applied to documents to remove the background. In this research, we build an interactive background removal interface. Then the multi-scale edge-hinge features and grapheme features will be extracted from the handwritten pages. The classification will be performed by a k-nearest neighbor classifier. The project studies the performance of the new features, and recognition operations on Arabic text, on the identification rate of writers. Matlab programming language is used to write the programs for this project. This project aims at building an Arabic writer identification system consisting of three main processes: an interactive preprocessing to remove documents background, a feature extraction process to extract feature vector, and a classification process. The three processes will be implemented on the training and testing phase of the system.

Forensic handwriting analysis is an important task in crime investigation. Writers of documents are sometimes unknown or disputed so it is difficult for the forensic detectives to identity of those writers. Currently, forensic professionals perform the analysis of these documents manually. Document management system for forensic document identification, verification and analysis are described. The system is designed to assist forensic document examiner with identifying and comparing Arabic handwriting and its characteristics with a graphical user interface. The same system also provides the examiner with functionalities for retrieving information about documents, region of interest, words, and writers. The system has been implemented using Oracle Database 6i. Our forensic document management system (FDMS) prompts its users with a graphical user interface (GUI) for forensic handwritten document database exploration and management.

A system for writer identification based on Arabic handwritten words was built. First a database of words was gathered and used as a test base. Then, features vectors were extracted from writers’ word images. Prior to feature extraction, normalization operations were applied to a word or text line. In this research, we studied the feature extraction and recognition operations on Arabic text, on the identification rate of writers. Since there is no well known database containing Arabic handwritten words for researchers to test, we built a new database of off-line Arabic handwriting text to be used for writer identification research. The proposed database is meant to provide training and testing sets for Arabic writer identification research. Arabic handwritten words were collected from 100 writers. We evaluated the performance of edge-based directional probability distributions as features and other features in Arabic writer identification.