ICT: Innovation and Computing

Road extraction from aerial imagery is not a trivial task. It plays a pivotal role in urban planning, navigation, disaster assessment and various other fields. It poses challenges due to complex scenarios and factors, including occlusion. Hence conventional methods prove to be inefficient for the purpose. Image segmentation and deep learning models are extensively employed in recent times to extract objects from images. In this paper, the performance of Unet architecture-based model has been improved by Resnet50, VGG16, DenseNet169, Xception and Efficientnet-b4. Further, to investigate the performance of Unet model, three other models FPN, PSPNet and PAN were implemented and evaluated on Massachusetts road dataset. The work presents the comparative analyses of the performance of models.

As per data requirement on huge, all moves to take data from here to there and in that case misusing of data take place and actual data get lost or get altered and then get misused, so privacy of data is required on large scale and to do it we all should be aware of the process, methods, approaches, and techniques. So main focus in this paper is on different methods, approaches, and techniques in very easy and understandable form with best examples.

This paper discusses methods, tools, approaches, and techniques used for the penetration testing on the cloud-based web application on Amazon AWS platform. The findings of a penetration test could be used to fix weaknesses and vulnerabilities and significantly improve security. The testing is implemented by undertaking a malicious attack aiming to breach system networks and thereby confirm the presence of cloud infrastructure. The research focuses on cloud-based web applications’ high-risk vulnerabilities such as unrestricted file upload, command injection, and cross-site scripting. The outcomes expose and approved some vulnerabilities, flaws, and mistakes in the utilised cloud-based web application. It is concluded that some vulnerabilities have to be considered before architecting the cloud system. Recommendations are proposing solutions to testing results.

In recent years, electric vehicle charging stations have improved as EVs have grown incredibly popular and attracted a lot of attention. By providing efficient platforms for the exchange of energy between power grids and EVs, EVCSs are unquestionably essential for the emergence of energy. Additionally, the temporary and geographic distribution of electric charges is influenced by the requirements of EV charging. The effective management of energy among EVCSs, the power grid and EVs depends on EV charging. The primary objective of this study that is being proposed is to create and modify an artificial intelligence-based energy management system for electric vehicles. In this proposed work, fuel cells, ultracapacitors, batteries and EVs are considered. The ultracapacitor may provide peak power and recover braking energy by connecting to the DC bus in parallel via a bidirectional DC-DC converter, which reduces the load on the fuel cell system and battery. This lengthens battery life by requiring less energy to charge and discharge the battery. The RNN model, which uses the fuel cell, UC, battery and EV factors, works in tandem with energy management. The appropriate speed, battery power, UC's SOC, UC current, UC voltage and fuel cell power are taken by altering the FTP 75 driving cycle. The suggested RNN model uses these regulating factors as input and uses them to predict the phase delay for better energy management. The experimental evaluation and analysis of the proposed model are carried out in MATLAB. Lastly, the superiority of the suggested technique is demonstrated with respect to error measures.

Preeclampsia is a condition that only occurs during a woman's pregnancy and is identified by a rise in the expecting patient's blood pressure, often after the 20th week of pregnancy. It is one of the top three causes of mortality among pregnant women worldwide. Accurate preeclampsia risk prediction would allow more effective, risk-based maternal care pathways. Delivering accurate preeclampsia risk assessment ranging from high to low requires feasible biomarkers. The maternal health risk public dataset provided by Oslo University Hospital, Oslo, Norway was used in this work. The data was collected from different hospitals, community clinics, and maternal health cares at Oslo University Hospital, (Oslo, Norway) through the IoT-based risk monitoring system. The dataset includes biomarkers/indicators such as heart rate, blood glucose levels, diastolic and systolic blood pressure, body temperature, and others. These five most important biomarkers should be kept under their respective normal levels as they play a vital role in predicting risks during pregnancy. The machine learning techniques for predicting various risk levels, including Naïve Bayes (NB), logistic regression (LR), Ada boost (AB), support vector models (SVM), decision tree models, the k-nearest-neighbor algorithm (KNN), and random forest (RF) are used in this work. These supervised machine learning tools gave an accurate prediction of the preeclampsia risk level, with the experimental results giving the highest accuracy to random forest (RF) of 96.39%, among the used machine learning tools.

A form of navigation problem called path planning can be resolved using a variety of techniques. This paper presents an overview of path planning techniques, specifically focusing on finding the shortest and most efficient path in a static environment. Self-driving autonomous vehicles can identify the safest, most practical and economically advantageous routes from source to destination using appropriate path planning and decision-making in real-world urban contexts. The proposed work first utilizes an open-source CARLA Simulator to implement path planning using the A star algorithm in its inbuilt town map. It makes use of CARLA library modules such as Waypoint API, CARLA Townmap, and PID controllers for its functionality. Secondly, the local real-world map is exported from the osm.org website and consists of local geographic data required to demonstrate the path planning of autonomous vehicle in a real-world environment. The results are demonstrated using the simulator. With several path planning algorithms present, this work utilizes A* algorithm and gives out the shortest path between start and end locations. The major advantage of using the CARLA simulator is that we can use the inbuilt Python API to convert a given exported .osm file to a .xodr file, which can be integrated into the simulator, thus allowing our algorithms to be tested in real-world scenarios.

As the field of machine learning grows rapidly, a lot of attention has been paid to how training complex models affects the environment. Carbon emissions caused by the computing needs of machine learning algorithms are becoming a big concern. This is because these models need a lot of computing power and energy. The goal of this paper is to find out how training Neural Machine Translation models affects the environment in terms of carbon footprint and to look into ways to reduce that effect. Machine translation, which automatically translates from source to target, is an area of natural language processing where researchers have been actively working for a long time. The performance of Neural Machine Translation (NMT) is enhanced by exploiting Artificial Neural Networks (ANN) in its model implementation. However, NMT is highly data-hungry and it requires longer training time. In this paper, an attempt has been made to estimate carbon emission when the different NMT models are trained in low-resource language pairs such as English to Hindi and English to Bengali language pairs on different hardware configurations. Finally, different alternatives have also been suggested to reduce this carbon emission and thereby its adverse impact on the environment can be minimized.

Women are the epitome of strength, resilience, and compassion. Their contributions to society, in every sphere of life, are immeasurable and invaluable. Women possess an innate ability to nurture, inspire, and lead with grace and determination. Despite their immeasurable contributions, women continue to face alarming levels of insecurity and vulnerability. The issue of women’s safety remains a pressing concern in our society. From the streets to their own homes, women are exposed to various forms of harassment, violence, and discrimination. The fear of walking alone at night or being subjected to assault restricts their freedom and hampers their ability to fully participate in public life. Women’s safety has emerged as a highly sought-after research area in recent years. However, current solutions on wearable devices often rely on manual activation through buttons or physical intervention, such as buzzing devices, to alert others in times of danger. While these methods have provided some level of assistance, they often require the victim to take specific actions, which may not always be feasible or safe in high-risk situations. Utilizing state-of-the-art IoT and embedded systems, our research paper presents an innovative and convenient solution that eliminates the need for manual intervention. By seamlessly detecting vital physiological changes, such as heart rate and stress levels, our technology automatically triggers a swift notification to both the individual’s immediate family members and relevant authorities. The key innovation lies in the seamless integration of these features, eliminating the need for manual intervention and ensuring timely responses to potential threats. This hands-free method establishes a new benchmark for safety solutions, offering women in potentially hazardous circumstances ease, assurance, and improved safety precautions. Our innovative technology creates a world where fear gives way to unwavering confidence and serenity in the face of challenges. By prioritizing women’s safety, we are paving the way for empowerment and a profound sense of security. Together, we can build a future where women can navigate their lives with peace of mind and a renewed sense of strength.

As the volume of vehicles on our roads continues to surge, accurate detection and counting of vehicles have become critical for effective traffic management. Identifying vehicles precisely is challenging due to the wide range of sizes, shapes, and external factors influencing computer vision. To overcome these challenges, here propose a vehicle detection strategy based on the YOLOv5 algorithm. YOLOv5 is an advanced object detection algorithm leveraging convolutional neural networks (CNNs) for high-precision, high-speed detection in images and videos. Our strategy harnesses YOLOv5’s capabilities, optimizing it for both speed and accuracy. Comprising convolutional layers, pooling layers, and fully connected layers, YOLOv5 collaboratively detects and identifies vehicles in images or video frames. Extensive training on a diverse dataset empowers the algorithm to recognize vehicles with exceptional precision. An empirical study evaluated YOLOv5’s performance across diverse vehicle types and environmental conditions. Results unequivocally demonstrated substantial improvements in vehicle detection speed and precision. Even under challenging scenarios, the algorithm consistently achieved real-time identification and enumeration of vehicles.

This paper describes the detailed specification and design of a portfolio manager, called Sakhee, which can handle multiple projects over their requirements, human, financial and software resources, priority and their dynamic rank ordering, deliverables out of a project, a knowledge repository, feedback and testing reports, and other resources required by the project. In addition, Sakhee will allow multiple projects to be monitored simultaneously and make effective and efficient transfer of assets—both digital and personnel—across projects. In addition, Sakhee also allows the close monitoring of the entire process and progress of various projects so that inter-project knowledge and information management can also be tracked. Sakhee is not a project scheduling, project reporting or time maintenance or time management tool. Further, it is not a tool for monitoring the progress of project or project member, but Sakhee manages Portfolios and aids portfolio managers. This paper also presents the detailed requirements for the design of portfolio manager, Sakhee, its ViewSpaces and functionalities, besides its information architecture. Sakhee therefore belongs to a class of implementation concept, called the Total Portfolio Management (TPM) system.

This paper concerns the development of both theoretical and operational frameworks, called Panorama, for comprehending large scale, multi-lingual software systems. The foundation to the system is the development and deployment of instrumentation agents, which have the ability to burrow through the software at various levels of depth—both statically and dynamically and treating the software as Black box and/or White box—in order to collect valuable information. In our past research (Gallagher and Lakshmi Narasimhan in, IEEE Trans Softw Engg 23(8):473–484, 1997), we have developed both a theoretical framework and an implementation of a core set of instrumentation statements in order to optimize data generation for testability purposes. This framework is distributed and applicable to assembly language, procedural and OO paradigms, and combinations thereof. This basic framework has been augmented with additional theoretical foundation and instrumentation procedures so that processes for software visualization and comprehension can be achieved. The instrumentation is completely handled by a distributed agent-based system, which is also capable of detecting the underlying programming languages and capturing all book-keeping information needed for various application requirements. It is envisaged that robust optimization procedures are required (as was shown before for testability purposes) for dynamic system composition. The system has the capability for self-cataloguing and querying toward the above-mentioned purposes, in addition to providing view-tailoring and view maintenance for specific application domains. The Panorama system can be mounted over a Cloud environment also.

This paper presents a novel driving style recognition method with high accuracy, speed, and generalizable. The proposed approach addresses the limitations of existing unsupervised clustering algorithms and single convolutional neural network methods due to the lack of diverse driving data types. The method first collects driver’s operation time sequence information from imperfect driving data. Next, it extracts driver’s style features using a convolutional neural network. The temporal data is then processed using Long Short-Term Memory (LSTM) networks for driving style classification. Further improving this model, we have used advanced algorithm called CNN + LSTM + BILSTM. Experimental results demonstrate an impressive recognition accuracy exceeding 99.

In India Agriculture is the backbone of the economy and a source of employment. Agriculture contributes 20% to the GDP of India. There are many losses due to diseases that bring downcast efficiency and increase financial losses. Agricultural area needs to stand and progress from such problems to be highly gainful. This can be achieved by detecting diseases at appropriate stages of the fruit and plant lifespan. Suitable machine learning algorithms (linear and logistic regression, KNN, K-means, SVM, etc.) and deep learning algorithms (CNNs, LSTMs, RNNs, etc.) can be used with image processing techniques for identifying the various diseases in fruits and plants. Researchers developed various methods for the detection of diseases using the above techniques. This paper proposes to emphasize the review of research work in the agriculture field such as disease detection in pomegranate and checking algorithms, diseases, datasets, accuracy merits, demerits of each technique used. For other researchers working in the field of image processing for the detection and classification of leaf/fruit diseases, this review article will be crucial in understanding the state-of-the-art.

The pendulum-based drive system with live streaming and object recognition is the paradigm that is suggested in this study. It is intended to carry out activities for surveillance and rescue. The system is driven by an Arduino controller, which interfaces with sensors, motors, and Wi-Fi to regulate the movement. Due to its benefits over traditional models, the model’s body is a spherical shell. Two cameras are affixed to the spherical ball in order to enable a 360-degree view. The model is able to move by lowering its center of gravity when the weight is changed. Although various mechanisms can be employed to design robots, the pendulum-driven mechanism is frequently utilized due to its inherent simplicity. Consequently, numerous spherical robots incorporate solely internal sensors for locomotion, stabilization, propulsion, and other functions. Nevertheless, a select few are additionally outfitted with sensors capable of detecting and measuring environmental variables. The review examines the potential of various driving mechanisms and sensors utilized in spherical robots and introduces and discusses a distinct category of mobile robots. The sensors analyzed have the capacity to augment the sensing capabilities of robots that are generally quite limited.

Attendance management is an important task in educational institutes as it reflects in the academic performance of a student. Upon observation, it was found that a considerable amount of time is being spent in this process (which has a repetitive nature) and a need of automating this process is realized. The main idea discussed in this paper is to use deep learning techniques to automatically identify the students present in a class and mark their attendance. Convolutional neural network (CNN) was being applied to detect the faces visible in the frame which was then used to recognize students present and mark their attendance. The proposed architecture was applied in the online mode of teaching, and the best accuracy of 76% was being recorded using the FaceNet model, while the best precision of 88% was being recorded using the VGG-Face model. These best-performing CNN models can be considered as an effective tool for face recognition to enhance the current attendance recording system.

More than any other source of energy, electricity today needs to be used efficiently. Power theft is a leading cause of nontechnical losses in distribution networks. Electricity providers around the world are facing a major financial burden due to this issue. New methods of electricity theft are made possible by the smart grid paradigm. To begin, cybercriminals can now steal electricity from a distance. Smart meters placed as part of the AMI collect data that may be used to track and bill clients for their energy usage in real time. A malicious client could launch a cyber-attack on the smart meters in an effort to reduce their electricity cost. Another perk of the smart grid model is that consumers can make money by producing their own electricity from renewable sources and selling it back to the grid operators. Methods for connecting renewable DG units to the grid, such as the net metering system and the Feed-in Tariffs policy, are discussed in this article. Through the net metering system, customers are able to “bank” the DG’s excess generation for later consumption. Consumers who sell all of the electricity they create back to the grid receive compensation under the FIT scheme, often known as clean energy pay back. The financial incentives provided by FIT schemes significantly increase the rate at which renewable energy sources are adopted. To take part in FIT, a client needs two meters: one to track energy generated by DG unit and injected into grid and another to track the energy consumed within the home. Both production and consumption-based energy pricing models benefit from this. Therefore, the aim of this research is to compare the performance of various deep learning algorithms include DNN, RNN-GRU, and CNN for spotting power-grid attacks.

The rapid growth of scientific literature has made it increasingly difficult for researchers to stay informed about the latest developments in their field. One way to make sense of this vast amount of information is through topic modeling, a method that automatically identifies themes in a collection of documents. In this study, we applied topic modeling algorithms to a dataset of scientific publications indexed in the Scopus database. The goal was to identify the main topics discussed in these publications and to understand the research trends and areas of focus within a particular field. We used Latent Dirichlet Allocation (LDA) and Latent Semantic Analysis (LSA) to analyze the Scopus data and found that topic modeling can be an effective tool for exploring research trends in science. The results of this study can help researchers stay informed about the latest developments in their field and can aid in identifying new areas of research.

Over the past years, there is a rapid expansion for handling bioinformatics data, particularly processing with gene expression levels via microarrays. Due to the characteristic of microarray data, which often entails with more features and less samples, the task of classifying this data becomes notably intricate. By using microarray technology, gene expression profiles may be produced in massive quantities. Currently, gene expression data are used to diagnose illness. The use of deep learning algorithms is one such method that aids in this process. These methods work well for classifying and identifying informative genes. The classes of testing samples may be predicted using these genes. Microarray data used to identify cancer often has a small number of samples and a large feature collection size derived from gene expression data. Use of deep learning algorithms is currently receiving a lot of interest in the field of artificial intelligence to address various problems. In this paper, we examined a deep learning system for microarray categorization based on the convolutional neural network (CNN) over other machine learning techniques. The effectiveness of CNN has been compared with existing system, and results are discussed.

Accurately classifying physical activity is a big undertaking in a variety of industries, from healthcare to sports analytics. Physical activity plays a critical role in maintaining health and well-being. This review article offers a thorough overview of the many approaches and procedures used to categorize physical activities. We look at how the classification of physical activity has changed over time, from conventional approaches to the most recent developments in machine learning and sensor technologies. The study discusses several algorithms and methods used in physical activity classification, including adaptive boosting, random forest, KNN, and artificial neural network.

Traffic accidents are a significant global concern, leading to numerous fatalities and injuries, especially in remote areas where rapid detection and reporting to emergency services are crucial. This paper presents a novel application for identifying traffic accidents using brain-inspired neural network approaches such as spiking neural networks (SNNs) and Legendre Memory Unit-based recurrent neural networks (RNNs). These brain-inspired systems operate on a neuromorphic computing architecture with energy-efficient abilities including spatiotemporal processing. These brain-based architectures are briefly discussed and contrasted with conventional Artificial Neural Networks (ANNs) to produce metrics on the performance of the models for an in-depth comparison, using an accident detection data set generated from CCTV footage. This research also recommends the use of Predictive Quality of Service (PQoS) using different machine learning models to effectively communicate the findings of an accident scenario to emergency services. This is done by evaluating the effectiveness of the methods to locate nodal points in a V2X data set. Object detection methods like YOLOv5 have been briefly investigated for vehicle detection in an accident scenario in order to relay extra information about an incident and identify potential improvement areas. The main motivation for this method of accident detection and communication is the decrease of fatalities, especially in distant locations where quick detection and reporting of emergency services is crucial. This paper illustrates the relevance of the work in 5G and future 6G use cases, the role of AI and ML in visual image processing, and the social significance of this research after demonstrating reliable performance utilising a variety of machine learning algorithms in both accident detection and predicting QoS parameters.

In the realm of data mining, the presence of missing values poses significant challenges that can undermine the accuracy and reliability of analytical outcomes. This study delves into the critical task of addressing missing values to mitigate the potential for ambiguous results in data mining processes. Recognizing the pivotal role of complete and accurate data in generating meaningful insights, this article explores various approaches for handling missing values, including omission, imputation, interpolation, and model-based techniques with valuable insights into selecting the most appropriate strategy based on contextual factors. Study also provides information about the potential of model-based imputation with their variants. The research article highlights the nuanced process of model selection and its pros and cons. The study provides a layman framework that integrates both traditional and innovative methodologies; this study contributes to a holistic understanding of mitigating the impact of missing values.

In the digital age, the significance of social media networking cannot be overstated. Individuals, businesses, and organizations have embraced these platforms as essential tools for communication, brand promotion, and community building. Social media analytics involve the systematic collection and analysis of data to measure the performance of one’s presence on platforms like Facebook, Instagram, Twitter, and LinkedIn. Key performance indicators (KPIs) such as engagement, reach, impressions, and conversion rates are pivotal in assessing the effectiveness of social media strategies. Additionally, understanding audience demographics, content performance, and competitor analysis informs data-driven decision-making. Growth perspectives in social media networking encompass strategies to expand reach, engage audiences, and foster sustainable growth. Consistency in posting, delivering high-quality content, and engaging with users are fundamental principles. Paid advertising, collaborations, and community building are strategies that extend one’s social media reach. Furthermore, data-driven optimization, trend monitoring, and user experience enhancements are indispensable for staying relevant in a dynamic digital landscape. This article is an attempt to provide a concise overview of the critical components related to social media networking along with analytics and growth perspectives. The article encapsulates the essence of social media networking analytics and growth perspectives, offering a foundation for individuals, businesses, and organizations to harness the full potential of these platforms, connect with their target audiences, and achieve enduring growth in the ever-evolving realm of social media.

The theory of network reliability has a lot of application in complex network structures, communication networks, cloud computing, traffic networks, and so on. This theory plays a crucial role in evaluating how efficiently networks and depicted as probabilistic graphs. Despite the challenge of network reliability evaluation being NP-hard, there exists a wealth of proposed solutions. However, a predominant number of these solutions have focused on the sequential computing, which is failing to fully leverage the advantages offered by multi-core processor architecture. This paper overcomes this limitation by proposing an efficient strategy that calculates the two-terminal reliability relied on parallel computing. The paper initially delves into a thorough analysis of existing methodologies, followed by the proposal of an efficient technique for computing terminal-pair reliability utilizing Logical-Probabilistic Calculus (LPC). Finally, the paper presents a parallelized iteration of the proposed algorithm designed for a multi-core processor architecture. Results obtained from experimentation confirm the superiority of our proposed algorithm in parallel version compared with other methods.

In modern wireless technology, optimum antenna structures are essential which response with high data rates, quality performance, and good reliability. A 4 × 4 MIMO antenna is designed, fabricated, analyzed, and proposed. FR4 material has been utilized for the commercialization. The distance between four elements is carefully designed to receive optimum isolation. Square ring-shaped defected ground structure (DFS) is introduced to achieve desire isolation. The radiation characteristics of the resonator have been carefully handled through iteration in the slot and patch dimensions. The proposed radiator exhibits dual-band response with moderate gain of 2.12 and 2.96 dBi. The computed efficiency of 80.37% and 83.48% has been recorded for 4.46–4.57 GHz frequency band and 4.95–5.03 GHz frequency band, respectively. The other output parameters such as MEG, CCL, and ECC are also having prominent values. There is a very close correlation between the software computed return loss values and actual values. The presented antenna is best suitable candidate for 5G sub-6 GHz and Wireless Local Area Network (WLAN) applications.

Diseases of the heart are now the top cause of mortality for individuals all over the globe. Numerous medical tests may be used to diagnose the many different types of heart disease; nevertheless, it is very difficult to forecast heart illness in the absence of such testing. The processing of large amounts of medical data may be aided by machine learning, which can also reveal previously concealed information that could not be discerned manually. By developing an improved model, this study's objective is to evaluate the many different machine learning approaches that are now available and the possible uses of such techniques in the area of cardiovascular disease prediction. The primary objective of the study is the development of an artificial intelligence-based system for the diagnosis of heart disease by using several machine learning methods. We demonstrate how machine learning may be used to assist determine the likelihood that a person will acquire heart disease. In this article, a Python-based application is constructed for the purpose of conducting research in the healthcare industry. This application is designed to be more dependable and to assist in the tracking and establishment of various sorts of health monitoring apps.

This study presents a novel solution in the form of a solar-powered water trash collector, designed specifically for deployment in backwaters and smaller water bodies, to address the critical issues of water pollution and waste management. Focusing on the challenges faced by national rivers, including hazardous polymers, plastic waste, and decaying organic matter, the collector operates under IoT-enabled Bluetooth control using mobile application, incorporating a conveyor belt mechanism for efficient waste collection, particularly targeting prevalent plastic debris in freshwater environments. A notable feature of this solution is its integration of solar power, ensuring self-sustained operation and reduced environmental impact, aligning with the imperative for eco-friendly waste management technologies. Bluetooth control enhances adaptability for smaller water bodies, crucial for precise navigation. The core objective is to provide an effective solar-powered water trash collector tailored for backwaters, addressing the escalating waste production due to population growth. Additionally, a comprehensive waste management approach is proposed, facilitated by an Android application for seamless navigation of integrated trash bins. In conclusion, this study offers an innovative approach to tackle water pollution and waste management, introducing a solar-powered collector uniquely equipped to address backwater ecosystems’ specific challenges while promoting sustainability and environmental well-being.

Wireless communication innovations opened the door for the creation of compact, less-power, lower-cost, and wireless sensor networks’ multipurpose sensor nodes. The idea of “3 Any”—anybody, anywhere, and anytime—has helped wireless networks gain popularity. Nodes in a mobile ad hoc network can interact without the assistance of a centralized administration or network architecture. They may exchange data with them via several hops and are connected via wireless technologies. An independent group of mobile nodes forms the mobile ad hoc network, a wireless network lacking any kind of permanent infrastructure (mobiles, laptops, iPads, PDAs, etc.). Ad hoc joining and departing of each node cause the network to automatically re-construct its topologies and routing table info for the transmission of information packets. The researchers can understand both the general idea of MANET and its applications. In this paper, we work on secure data transmission and management using ECC and MANET-based security standards for secure data transmission using mobile nodes. The use of modified-AES and ECC and this security standard as a result shows improved PDR of 8–10%, throughput of 10–12%, and latency of the node as compared to the existing system.

With advancement in the technology, from 2G to 5G [TS 38.413: 5G NG-RAN; NG Application Protocol (NGAP)], there is a dedicated effort by the operators and the vendors to reduce the outage caused due to various reasons in the field and ensure service availability. In this regard, many tools were developed to aid in analysis of the problems through logs and other OAM counters to prevent them in future. AI/ML (Linin arXiv:2305.05092, 2023) has been applied to solve many problems across different domains and in this whitepaper, we discuss on how this (AI/ML) could help us solve RAN outage related issues by predicting them based on pattern analysis in a cost optimal manner. The key aspect of the solution discussed in this whitepaper is about anomaly predictions based on different logging files (e.g., cell trace, subscriber trace, etc.,). The early prediction of traffic failure patterns would aid in traffic steering action from the operators thus avoiding outage and ensuring high QoE (quality of experience) to the end user of the service. We shall also analyze how the proposed solution could be cost effective, vendor agnostics framework to analyze the 3GPP standard defined Subscriber, Cell, and Equipment traces.

Traffic congestion has emerged as a pervasive challenge across global urban landscapes, inducing delays, productivity losses, and heightened air pollution. Conventional traffic signal systems often falter in adapting to evolving traffic dynamics, compromising road network efficiency. To address this a novel paradigm—a smart traffic control system leveraging advanced computer technology is proposed. This system employs real-time monitoring and analysis to dynamically adjust traffic signals, optimizing traffic flow, and mitigating congestion-related adversities. The integration of deep learning and computer vision technologies is used to enable a nuanced understanding of visual data and patterns. Remarkably, the YOLO tool is utilized to enhance the system’s capacity to swiftly identify emergency vehicles and give them priority. The proposed system is designed to efficiently handle traffic density and includes features for prioritizing emergency vehicles. Furthermore, it employs a non-uniform allocation of waiting times to lanes, which is contingent upon real-time traffic density and patterns. A working model demo was designed and it was found effective in making intelligent decisions.

Gas leaks and fires present dangers that require detection and decisive action to protect lives and property. This paper specifically proposes cutting-edge sensor technologies with intelligent threshold algorithms and real-time communication capabilities to enhance the safety and well-being of industrial workers by reducing the risk of accidents and injuries, which addresses Sustainable Development Goal (SDG) 3 (good health and well-being). It includes MQ5 and MQ135 gas sensors, a DHT11 temperature sensor, and an IR flame sensor to provide monitoring. The real-time data from the gas sensors is carefully analyzed to trigger alerts at two threshold levels directly linked to gas concentrations, ensuring warnings in case of any risks. If the second threshold is exceeded, indicating the presence of gas, the device activates a safety protocol. An exhaust fan swiftly dissipates any accumulated gas, while a servo motor promptly shuts off the gas supply, effectively preventing any escalation of the situation. An integrated Wi-Fi module will enable communication and notifications through a dedicated smartphone app or web interface to enhance functionality, which enables remote device control and real-time access to sensor data for users. This combination of sensor technologies, intelligent algorithms, and remote accessibility showcases how technology-driven solutions can significantly enhance safety measures.

As the digitization of cities and the proliferation of automobiles continue, vehicular ad hoc networks (VANETs) have emerged as a prominent solution to enhance road safety and traffic management. VANETs enable vehicles to establish communication channels with each other (V2V) and with infrastructure nodes (V2I), facilitating real-time information exchange for applications like collision avoidance, traffic management, and multimedia delivery. This paper offers an in-depth analysis of popular VANET protocols, including 802.11p, WAVE IP, DSRC, and WSMP, to provide a comprehensive understanding of their functionalities within the VANET ecosystem. Through simulation studies involving simultaneous broadcast and unicast systems between nodes and roadside units (RSUs), as well as WSMP-based neighbor systems, we investigate their effectiveness in realistic scenarios. Ensuring the reliability and security of VANETs remains a substantial challenge. To address this, the paper explores the potential advantages of integrating VANETs with blockchain technology. By leveraging blockchain’s inherent immutability and decentralized architecture, VANETs can be fortified against malicious attacks and unauthorized access, thereby enhancing the overall system security. Furthermore, this paper investigates the vast potential of utilizing data obtained from VANETs. We examine various applications that can effectively leverage this data to make informed decisions and improve transportation systems. A noteworthy example involves the incorporation of machine learning models, where data from VANETs is processed through a blockchain network to obtain valuable predictions for actionable insights. This demonstrates how VANETs can contribute significantly to intelligent transportation systems and urban planning.

This research delves into the multifaceted challenges of open defecation and the constrained adoption of sanitation technology in rural Raichur villages, with a particular focus on Mamanathodi, Karnataka, India. Against a global backdrop where 3.6 billion lack improved sanitation, the study employs Participatory Rural Appraisal (PRA) and Human-Centered Design (HCD) methodologies, emphasizing the pivotal role of community-led initiatives customized to local norms. Beyond presumptions that infrastructure alone eradicates open defecation, the study illuminates socio-economic factors such as financial constraints, water scarcity, and cultural beliefs contributing to its persistence in Mamanathodi. The interdisciplinary team’s immersive ten-day experience, deploying PRA tools and observational techniques, brings forth nuanced insights. The research integrates demographic data, field observations, PRA tools, and scenario analyses. Proposing an ICT architecture for social change, the study advocates for data-driven, stakeholder-engaged, and convergent approaches. The envisioned architecture promises efficient data collection, stakeholder empowerment, information convergence, and cost-effective, scalable interventions. In conclusion, the paper presents recommendations urging policymakers to enhance stakeholder engagement and ensure the sustained longevity of sanitation initiatives, contributing significantly to the discourse on rural public health and hygiene practices.

The thorough study of hair and scalp falls under trichology. This study is performed using techniques like trichogram and Trichoscopy, which involve evaluating microscopic images taken of the hair and scalp and diagnosing the different scalp diseases like alopecia and telogen effluvium. The evaluation of the images supplies us with various informative parameters such as hair length, hair thickness, number, and density of vellus and terminal hair, and white spots. Trichogram, that is particularly used for detecting scalp conditions, is developed using image processing, and we have created a system that helps extract these features and generate a report displaying those parameters as well as comparing previously produced reports. The system allows the professional to upload an image or take an image on the go, using a dermatoscope. After receiving the image recorded, the image passes through several enhancement techniques such as resizing, sharpening, denoising, blurring, gray scaling, binarization, and morphological changes. Different methods are further applied to the enhanced image to analyze various parameters. Lastly after the analysis, a comprehensive report is provided to the professional and simultaneously stored in the dataset. The features derived from this cost-effective system will assist the practitioner to diagnose the disease and take productive measures to prevent and cure it effectively. The comparison provided gives the ability to successively monitor the condition and conclude the necessary. Image processing, which is the foundation of the entire system, is effectively applied to get the desired results.

Hair-related diseases are pervasive and can significantly impact individuals’ confidence and emotional well-being. Accurate diagnosis of these conditions poses challenges even for experienced professionals. However, the integration of technology, particularly deep learning and artificial intelligence (AI), has been showing bright results in the field of Trichology. This paper presents a comprehensive review of the techniques and technologies developed in Trichology. We begin by delineating common hair diseases, trichoscopy image acquisition methods, and available datasets. We also examine existing frameworks and tools that facilitate the creation of trichoscopy-based algorithms, along with frequently used evaluation metrics. The techniques studied in this paper involve hair loss detection using Mask R-CNN, Kaggle network, DEX-IMDB-WIKI and DEX-ChaLearn networks, deep learning-based scalp image analysis using limited data through ResNet, ResNeXt, DenseNet, and XceptionNet, image processing using grid line selection method, machine learning using SVN and KNN, and hair and scalp disease recognition using deep learning and image processing. The different algorithms used in all the mentioned techniques are analyzed, giving us a brief knowledge of how Trichology is applied with the help of technology to benefit professionals in accurately diagnosing different hair loss parameters and diseases. This review offers an overview of recent advancements in hair disease diagnosis using trichoscopy and Trichogram, emphasizing opportunities for further enhancement in this rapidly evolving field.

Social media is a platform for sharing content and interacting with other people through multimedia data such as photos, videos, and documents accessed via computers or smartphones. One of the most dangerous consequences of social media is the rise of cyberbullying, which is more diabolical than traditional bullying and hard to control. The objective of the system is to determine the bully score of the users, and it helps to identify how much the person is using bully phrases in social media. The proposed work aims to recognize the cyberbullying phrase in a tweet using VADER sentimental analysis of the user. BERT model is used to classify the bully tweets which performs with a better accuracy of 0.9535. A conversation graph is constructed by a bully score of each user with the help of the PageRank algorithm to identify the cyberbullies.

An Artificial Intelligence (AI)-based scribe known as L &E Refiner for blind learners is a technology that utilizes natural language processing and machine learning techniques to automatically transcribe lectures, books, and other written materials into audio format. This system is designed to provide an accessible learning experience for blind students, allowing them to easily access and interact with educational content. The AI scribe is able to recognize and understand various forms of text, including handwriting, printed text, and digital documents, and convert them into speech output that blind learners easily comprehend. This technology has the potential to significantly improve the accessibility and inclusion of education for blind individuals.

The digital evolution has led journalists to become much more familiar with technological devices in their daily work. Adaptations to the digital world, together with the different cases of information vulnerability toward professionals of the fourth estate, have highlighted the importance of digital information security in journalistic media. This systematic review of the literature addresses the analysis of 13 academic articles collected from the Scopus database that address the topic of digital security and journalism. The results indicate that there is a Spanish-speaking journalistic sector that has implemented digital security tools; however, other companies are truncated by organizational and political influences.

Modern cars employ Lane Keep Assist (LKA) as a crucial safety feature, using sensors and cameras to warn drivers when the vehicle veers off the road and correct it. Hardware-in-the-Loop (HIL) simulation is a powerful testing method to model sensor behaviour in controlled environments. Developers can assess LKA systems in diverse conditions, including road geometry, climate, and traffic patterns. HIL validation via dSPACE Scalexio provides a realistic and reliable platform for virtual testing. This paper used image processing techniques to achieve 95% accuracy in lane detection.

Pneumonia is a common and potentially life-threatening respiratory infection affecting millions worldwide. Prompt and precise diagnosis is critical for timely treatment and reducing mortality. Nowadays, the application of artificial intelligence, particularly Convolutional Neural Networks (CNNs), has shown promise in automating pneumonia classification from medical images. This research centres on developing and evaluating a CNN-based system for pneumonia classification using chest X-ray images. The aim of this study is to overcome the limitations associated with traditional methods of diagnosing pneumonia, which heavily depend on the expertise of radiologists. To achieve this, we utilized a substantial dataset of chest X-ray images that encompassed both cases of pneumonia and non-pneumonia, obtained from various sources. The dataset underwent pre-processing to enhance image quality and standardization for consistent analysis. A custom-designed Convolutional Neural Network (CNN) architecture was employed, and it was trained using a portion of the dataset. This leveraged the CNN's ability to autonomously learn distinctive features from the images. Following training, the CNN model was subjected to validation and testing using a separate set of images to evaluate its classification performance. Key metrics such as accuracy, sensitivity, specificity, and others were calculated to gauge the model's effectiveness in distinguishing between cases of pneumonia and non-pneumonia. To highlight the superiority of the proposed CNN-based system, we compared its performance to existing methods for pneumonia classification in terms of accuracy and efficiency. The experimental results clearly demonstrate the system's effectiveness in accurately classifying pneumonia from chest X-ray images. The proposed approach exhibits high accuracy, sensitivity, and specificity, surpassing traditional methods and highlighting its potential to aid healthcare professionals in pneumonia diagnosis. Automating pneumonia classification using CNNs can significantly alleviate the workload of radiologists, expedite the diagnosis process, and enhance patient care. This model has achieved an impressive 98% accuracy rate.

Detection and prevention of intrusion for computer networks are essential components that contribute to an organization's success. Machine learning is becoming a preferred method for a variety of classification and analytical issues because of recent breakthroughs in the field. Many network datasets containing pertinent and irrelevant features are available in networking communications. This raises the false alarm rate while significantly lowers the rate of intrusion detection. IDSs and IPSs have been utilizing various methodologies, and implemented to secure the availability, security, and reliability of corporate computer networks. This article examines the potential for machine learning automation in network security, a crucial area of computer networking. In 2015, dataset UNSW-NB15 was created and is the current benchmark network dataset, which is used in this article. We have implemented linear regression machine learning approach using the reduced feature space. Multiclass and binary classification are included in this paper. To compare the classifiers deployed, we calculated all of the standard evaluation parameters. The results demonstrated that accuracy with the use of binary classification 98.00% and with multiclass classification 0.01 needs further improvement or alternative methodologies to enhance the accuracy.