Learning through an Intelligent Tutoring System (ITS) lies in performing well in academics and improving the learner’s learning outcomes. The recent developments within an Intelligent Tutoring System are not focused on improving human–computer interactions. The best way to overcome this constraint is to develop ITS interfaces to provide learners with better learning experiences. In this study, Augmented Reality (AR) potential along with AI methodologies is utilized within the developed ITS interface to improve the learning experience of the learning-disabled learners. Augmented Reality is where virtual images overlay the physical world, and Mixed Reality is an emerging technology that presents an altered reality, which engages users to interact with an environment developed using virtual objects. With several tools and applications available, creating and providing immersive learning experiences for learners is rising. AR in educating the specially-abled is widespread, and its benefits are sufficiently explored. However, limited AR applications are designed specifically for supporting the education of individuals with learning disabilities. The available application, and their impact on learning-disabled learners, needs detailed investigation. This work presents an Intelligent Tutoring System (ITS) to educate learners with Augmented Reality (AR) based content. The ITS learner module implemented in this study was developed for learning disabilities identification and we have assessed total 105 participants (with or without Learning Disabilities) for the experiment. The ITS performance is compared (with or without) AR content-based learning and based on the findings, AR-based learning through ITS is effective. The benefits are manifold, increase in motivation, ease of interaction, development in cognitive skills, enhancement in short-term memory, and making lessons more enjoyable, turning the overall experience stimulating and engaging.

Detecting faults in software modules is important for reducing system failures and improving software quality. Traditional fault prediction methods often rely on failure history or statistical models, which may not work well when structural complexities exist within the code. This work introduces a new model called Enhanced Coupling and Cohesion Metrics-based Fault Detection (ECCMFD). It uses deep structural properties of code, such as Conceptual Lack of Cohesion in Methods (C-LCOM) and Conceptual Coupling Between Object Classes (CCBO), to capture how components interact and how focused each class remains on its purpose. These metrics are passed into a Deep Convolutional Neural Network (Deep CNN) that learns patterns in software design and predicts fault-prone modules. The model is evaluated on standard NASA MDP datasets including KC1, CM1 and PC3. It outperforms widely used models like Baseline CNN, Random Forest (RF) and XGBoost in all key evaluation metrics. ECCMFD achieved a 5% improvement in precision, a reduction in Root Mean Square Error (RMSE) by 0.3, and better performance in F1 score and accuracy. This improvement is due to the combination of well-defined structural metrics and the deeper feature learning capability of the deep CNN architecture.

Abstract The emergence of deepfake images created by GANs models for malevolent purposes presents a serious risk to society as well as great challenge to digital security and trust. Leveraging the power of ensembles and combining machine and deep learning approaches, this paper presents VOTSTACK, an innovative ensemble model designed to combat the proliferation of deepfake images on social media. VOTSTACK utilizes a blended approach that combines Voting and Stacking ensemble techniques. It leverages the collective intelligence of three different classifiers—Decision Tree, Logistic Regression, and SVM—executing a hybrid feature selection method with Principal Component Analysis (PCA) conditioning as the preprocessing framework. It refines the features using iterative feature resolution with cross-validation (RFECV) method. This model operates through a two-phase architecture, with the first phase consolidating results using a voting ensemble and the second phase aggregating collective knowledge into a final decision using a stacking ensemble. A majority vote method is used in the first phase to aggregate predictions from the three base classifiers (Decision Tree, Logistic Regression, and SVM). Utilizing strength of each classifier and results of voting method as meta classifier, a stacking ensemble further refines these predictions in the second phase. The effectiveness and reliability of this approach is validated on a substantial dataset known as Real and Fake Images reliability. The proposed model outperforms conventional methods, achieving an impressive accuracy rate of 91.6%, a high precision score of 90.3%, a substantial recall of 89.8%, and an outstanding F1-Score of 90%.

Artificial intelligence and machine learning have been transforming the health care industry in many areas such as disease diagnosis with medical imaging, surgical robots, and maximizing hospital efficiency. The Healthcare service market utilizing Artificial Intelligence is expected to reach 45.2 billion U. S. Dollars by 2026 from its current valuation, off $4.9 billion. Diabetic Retinopathy (DR) is a disease that results from complications of type one and Type two diabetes and affects patients' eyes. Diabetic retinopathy, if remains unaddressed, is one of the most serious complications of diabetes, resulting in permanent blindness. The disease has been affecting the lives of 347 million people worldwide. The paper aims to propose a residual network-based deep learning framework for the detection of diabetic retinopathy. The accuracy of our approach is 83% whereas the precision value for checking the absence of DR is 95%.

Healthcare IoT devices are increasingly integral to modern life, significantly enhancing health monitoring and living standards. These devices provide healthcare professionals and patients with new capabilities for continuous patient monitoring, enabling real-time analysis and intervention. However, the widespread adoption of wearable IoT devices brings both advantages and challenges. Data captured by these devices is transmitted to web applications for review by medical professionals and patients, where it can be analyzed using algorithms to suggest treatments or generate alerts. For instance, an IoT sensor that detects a patient’s abnormal heart rate can trigger an alert, prompting timely medical response. Despite these advancements, energy consumption remains a critical concern. Collecting and transmitting patient data through HIoT devices results in substantial energy use, particularly when computations are offloaded to the cloud, a common practice that increases energy demands. Real-time data analysis further complicates this issue, requiring low latency and minimal delays. To address these challenges, fog computing offers a viable solution by reducing multi-hop data connection delays, distributing resource demands, and enhancing service flexibility. This paper evaluates a Wireless Body Area Network (WBAN) model using the NS-3 simulation program on the Ubuntu platform, incorporating both externally worn and internally implanted sensors. The model’s accuracy improved with the number of hidden layers, reaching 0.9921 with sixteen layers, though at the cost of increased computational time. Additionally, the Adaptive Battery-Aware (ABA) algorithm was compared with the traditional Battery-Reliable Low-Energy (BRLE) method. Despite a slight increase in voltage usage, the ABA algorithm demonstrated superior charge recovery efficiency, offering better management of battery resources. This study provides insights into optimizing accuracy, battery management, and energy efficiency for WBANs, while also exploring energy-saving technologies like fog and edge computing in healthcare IoT architectures. The findings highlight the importance of balancing computational efficiency, energy consumption, and real-time data processing in enhancing the performance and sustainability of HIoT devices.

Abstract Histogram Equalization (HE) is one of the most popular techniques for this purpose. Most histogram equalization techniques, including Contrast Limited Adaptive Histogram Equalization (CLAHE) and Local Contrast Modification CLAHE (LCM CLAHE), use a fixed block size technique for feature enhancement. Due to this, all these state of art techniques are used to give poor denoising performance after feature enhancement. In this paper, a deep learning based new approach, namely Dynamic Block Size Technique (DBST), is used to improve image denoising. In this approach, we use the Categorical Subjective Image Quality (CSIQ) image set, an image database generally used for preprocessing of images. The results obtained from experiments show better performance for different important parameters (used by state of art techniques). The work is novel in the preprocessing of images because in this work, we classify the image depending upon the image features for selecting appropriate block sizes dynamically during preprocessing. Proposed work outperforms in terms of PSNR, MSE, NRMSE, SSIM and SYNTROPY. The average respective values are 18.92, 863.86, 0.25, 0.81 and 19.35 and are better in comparison of CLAHE and LCM CLAHE.

Abstract This study aims to improve the performance of organic to recyclable waste through deep learning techniques. Negative impacts on environmental and Social development have been observed relating to the poor waste segregation schemes. Separating organic waste from recyclable waste can lead to a faster and more effective recycling process. Manual waste classification is a time-consuming, costly, and less accurate recycling process. Automated segregation in the proposed work uses Improved Deep Convolutional Neural Network (DCNN). The dataset of 2 class category with 25077 images is divided into 70% training and 30% testing images. The performance metrics used are classification Accuracy, Missed Detection Rate (MDR), and False Detection Rate (FDR). The results of Improved DCNN are compared with VGG16, VGG19, MobileNetV2, DenseNet121, and EfficientNetB0 after transfer learning. Experimental results show that the image classification accuracy of the proposed model reaches 93.28%.

Abstract Global livelihoods are impacted by the novel coronavirus (COVID-19) disease, which mostly affects the respiratory system and spreads via airborne transmission. The disease has spread to almost every nation and is still widespread worldwide. Early and reliable diagnosis is essential to prevent the development of this highly risky disease. The computer-aided diagnostic model facilitates medical practitioners in obtaining a quick and accurate diagnosis. To address these limitations, this study develops an optimized Xception convolutional neural network, called "XCovNet," for recognizing COVID-19 from point-of-care ultrasound (POCUS) images. This model employs a stack of modules, each of which has a slew of feature extractors that enable it to learn richer representations with fewer parameters. The model identifies the presence of COVID-19 by classifying POCUS images containing Coronavirus samples, viral pneumonia samples, and healthy ultrasound images. We compare and evaluate the proposed network with state-of-the-art (SOTA) deep learning models such as VGG, DenseNet, Inception-V3, ResNet, and Xception Networks. By using the XCovNet model, the previous study's problems are cautiously addressed and overhauled by achieving 99.76% accuracy, 99.89% specificity, 99.87% sensitivity, and 99.75% F1-score. To understand the underlying behavior of the proposed network, different tests are performed on different shuffle patterns. Thus, the proposed "XCovNet" can, in regions where test kits are limited, be used to help radiologists detect COVID-19 patients through ultrasound images in the current COVID-19 situation.

Abstract In this paper, we propose a novel hybrid transformer architecture for food cuisine detection and classification. The work carried out within this paper develops a combination of Vision Transformer ensemble architecture with hand-crafted features, thereby making a hybrid Vision Transformer food recognition system. Recently, Vision transformers have been introduced as an alternative means of classification to convolutional neural networks. It performs pattern detection and classification without convolutions and interprets an image as a sequence of patches. The combination of Vision Transformer and hand-crafted features like GIST, HoG (Histogram of Oriented Gradients), and LBP (Local Binary Pattern) were employed on the dataset. The dataset was specifically created (for this work) from the public logging system. It consisted of 13 food categories with 400 images of Indian food items like Ghevar, Idli, Dosa, and much more. It helped to capture a variety of images from every domain and culture. This work made use of the common and readily available food items, which can further be increased by adding on the specialties (dishes) from different regions. Various experiments were performed on CNN with various classifiers like Random forest, and SVM. Further, we compared our proposed approach with several ensembles of CNN architectures. The experiments proved that our proposed approach outperformed the state-of-the-art ensemble CNN architectures for detecting food cuisines. The proposed hybrid approach achieved an accuracy of 94.63%, sensitivity 84.42%, specificity 95.23%, and kappa coefficient 0.93, which was the best amongst all approaches.

Abstract Food is indispensable for humans as their growth and survival depend on it. But nowadays, crop is getting spoiled due to fungi and bacteria as soil temperature are changes very rapidly according to sudden climate changes. Due to fungi-bacterial crop, the quality of food is declining day by day and this is really not good for human health. The goal of this research paper is the advanced detection of fungi-bacterial diseases in plants using modified deep neural network approach and DSURF method in order to enhance the detection process. Proposed approach of this research is to use the artificial intelligence techniques like neural network model and dynamic SURF method in order to identify and classify the plant diseases for fungus and bacteria. Additionally, support dynamic feature extraction DSURF & classifier combinations for creating image clusters with the help of Clustering. Deep learning model is employed for training and testing the classifier. The quantitative experimental results of this research work are claimed that authors have achieved the 99.5% overall accuracy by implementing DNNM and DSURF which is much higher than other previous proposed methods in this field. This proposed work is a step towards finding the best practices to detect plant diseases from any bacterial and fungal infection so that humans can get healthy food.

Abstract The growing demand for applications based on Generative Adversarial Networks (GANs) has prompted substantial study and analysis in a variety of fields. GAN models have applications in NLP, architectural design, text-to-image, image-to-image, 3D object production, audio-to-image, and prediction. This technique is an important tool for both production and prediction, notably in identifying falsely created pictures, particularly in the context of face forgeries, to ensure visual integrity and security. GANs are critical in determining visual credibility in social media by identifying and assessing forgeries. As the field progresses, a variety of GAN variations arise, along with the development of diverse assessment techniques for assessing model efficacy and scope. The article provides a complete and exhaustive overview of the most recent advances in GAN model designs, the efficacy and breadth of GAN variations, GAN limits and potential solutions, and the blooming ecosystem of upcoming GAN tool domains. Additionally, it investigates key measures like as Inception Score (IS) and Fréchet Inception Distance (FID) as critical benchmarks for improving GAN performance in contrast to existing approaches.

Abstract Cancer is a common disease with an increasing mortality rate in recent years. Lung cancer is the most common cancer in men and women alike. It is caused by uncontrolled cell development in the lungs. These cells are divided into two types: benign and malignant. Benign tumours are usually harmless, do not spread to other cells, and have a smooth and regular shape, whereas malignant tumours can be dangerous and spread to other body cells to form a new cancerous nodule with an uneven shape. If lung cancer is detected early, it can be treated. Lung cancer symptoms typically appear in the human body when it is in its final stage, but advanced technology and computer-aided systems can detect it at an early stage. Currently, numerous conventional and machine learning techniques are used for such automated detection systems to detect lung cancer in its early stages, but such automated detection systems do not provide accurate detection and the processing of lung cancer detection takes a long time. As a result, a novel method for detecting lung cancer that employs deep learning techniques for accurate detection while requiring less computation time is proposed. CT images are used in this study because they have less noise disturbance than MRI and X-ray images. Median filtering and patch processing are used to improve image quality on such CT scans. These pre-processed images are then subjected to a clustering segmentation process, which segments the image and feeds it to a CNN classifier. For feature extraction and classification, CNN architecture is used. In the future extraction section, various low-level and high-level features are extracted. The classification layer is in charge of determining whether the provided image contains a malignant, benign, or normal tumour. Finally, statistical parameters like MSE, PSNR, Accuracy, Sensitivity, Specificity, and others were computed and combined with the existing system in this work.

The original publication of this article contains error in the author name. The author name "J. Chinnababu" should be "J. Chinna Babu." The original article has been corrected.

Abstract Computer vision technology for detecting objects in a complex environment often includes other key technologies, including pattern recognition, artificial intelligence, and digital image processing. It has been shown that Fast Convolutional Neural Networks (CNNs) with You Only Look Once (YOLO) is optimal for differentiating similar objects, constant motion, and low image quality. The proposed study aims to resolve these issues by implementing three different object detection algorithms—You Only Look Once (YOLO), Single Stage Detector (SSD), and Faster Region-Based Convolutional Neural Networks (R-CNN). This paper compares three different deep-learning object detection methods to find the best possible combination of feature and accuracy. The R-CNN object detection techniques are performed better than single-stage detectors like Yolo (You Only Look Once) and Single Shot Detector (SSD) in term of accuracy, recall, precision and loss.

The major issues and challenges of the Industrial Internet of Things (IIoT) include network resource management, self-organization; routing, mobility, scalability, security, and data aggregation. Resource management in IIoT is a challenging issue, starting from the deployment and design of sensor nodes, networking at cross-layer, networking software development, application types, environmental conditions, monitoring user decisions, querying process, etc. In this article, computational intelligence (CI) and its computing, such as neural networks and fuzzy logic, are used to tackle the challenges of resource management in the IIoT. The incorporation of the neuro-fuzzy technique into the IIoT contributes to the self-managing intelligence systems’ self-organizing and self-sustaining capabilities, offering real-time computations and services in a pervasive networking environment. Most of the problems in IIoT are real-time based; they require fast computation, real-time optimal solutions, and the need to be adaptive to the situation of the events and data traffic to achieve the desired goals. Hence, neural networks and fuzzy sets would form appropriate candidates for implementing most of the computations involved in the issues of resource management in IIoT networks. A real-time testbed network is simulated and implemented on the Crossbow mote (sensor node) using TinyOS.

Summary Recently, there has been a massive rise in the popularity of Instagram, which connects individuals globally and allows videos and images to be uploaded and exchanged, and communicated over social media. Instagram is also an online playground of deceit. The use of filters, lighting, and cunning angles transforms the mundane into something spectacular. Automated spam accounts and fake profiles use this to their malicious advantage for executing attacks targeting high‐profile executives. Creating fake Instagram identities is easy to reproduce the idea of being accepted by many fans on social media. Fake accounts are used in the marketing of fake services and products. This research focused on designing and training a unique neural network model and proposed a new algorithm for detecting automated spam and fake Instagram account profiles. The precision and accuracy of the proposed method were achieved at 93% and 91%, respectively.

Internet of Things (IoT) is a new technology with multiple smart connected sensors capable of processing, storing, and computing. Industrial IoT (IIoT) is used in industrial applications, such as infrastructure, medical, logistics, and energy efficiency in smart grids. The network lifetime will be extended when sensor node energy usage is effectively controlled. This article proposed a multilayer threshold cluster-based energy-efficient low power and lossy networks (MTCEE-LLN) protocol for IIoT devices to decrease the network data traffic, sensor node energy consumption (EC) and also extends the network lifetime. The proposed scheme works in three phases: 1) network creation; 2) intra clustering; and 3) intercluster routing. The MTCEE-LLN forms equal-sized cluster in each transmission and elects the cluster head (CH). It maintains the destination-oriented directed acyclic graph (DODAG) to performs data transmission from the downward layer to the${\mathrm{ DODAG}}_{\mathrm{ root}}$. Furthermore, it aggregates the data packets in the cluster node to increases the network lifetime by reducing the number of redundant data packet transmissions. The proposed routing protocol has been evaluated based on different performance parameters such as packet loss rate (PLR), EC, control packet rate (CPR), and Node Failure Ratio. The simulated result proves its effectiveness compared to other traditional routing protocols.

Abstract Face Recognition is a challenging task for recognizing and detecting the identity of an individual. Although, plethora of work has already been done in the field of pattern recognition still there has been lot which has not been addressed in any of the literature. In the current research, we have presented a comparative analysis using three popularly known techniques for face recognition namely, Principal Components Analysis (PCA) using Eigen Faces, Hidden Markov Model (HMM) using Singular Value Decomposition, and Artificial Neural Network (ANN) using Gabor filters. These techniques are implemented and evaluated using various measuring metrics such as false acceptance, false recognition rate, and so on. We used ORL and Yale Face dataset to test the robustness of implemented algorithms. Results show that ANN model for face recognition outperforms the other two techniques by achieving more accurate results and shows the highest recognition rate of 97.49% on ORL database. Moreover, it is also observed that ANN model shows the minimum error count of about 2.502% on ORL database while it is 3.5% on Yale Face dataset. To evaluate further, the implemented techniques are compared with best known techniques in class implemented by various researchers.