Authors - Vandana R. Babrekar, Sandeep V. Gaikwad Abstract - This paper presents a literature review of current state of knowledge and innovation related to precision agriculture (PA) technologies leveraging Internet of Things (IoT), Artificial intelligence (AI), Computer vision (CV), Machine learning (ML), and Deep learning (DL). The review provides a thorough understanding of the various developments made in the above-mentioned technologies in precision farming. It also states several challenges after reviewing research publications, and it also recommends thrust areas for future studies.
Authors - Bimal Patel, Jalpesh Vasa, Ravi Patel, Martin Parmar, Krunal Maheriya Abstract - Software robustness, a vital component of software quality, encompasses key attributes such as reliability, usability, efficiency, maintainability, and portability. This paper offers a comprehensive overview of these attributes and examines the role of optimization tools in enhancing software robustness. Reliability, which ensures a system’s consistent dependability, is achieved through techniques like redundancy, error handling, and extensive testing. Usability, focusing on the user experience, is improved through user-centered design, usability testing, and heuristic evaluation. Efficiency targets the optimal use of system resources such as CPU and memory, with performance profiling, load testing, and code optimization techniques helping identify and resolve bottlenecks. Maintainability, ensuring that systems can be easily modified or updated, is enhanced through modularity, code readability, and design patterns that simplify future changes. Portability, which allows software to operate across diverse platforms, is achieved through cross-platform frameworks and containerization technologies such as Docker and Kubernetes. Optimization tools, including profilers, load testing tools, static code analyzers, and dependency management tools, play a critical role in maintaining software robustness. These tools help identify performance issues, ensure resource efficiency, and improve code quality. By leveraging these tools, developers and project managers can build more reliable, efficient, and maintainable software systems. This paper serves as a valuable resource for improving the overall quality, resilience, and portability of software products.
Authors - Sehaj Preet Kaur, Rahul Chaudhary, Rashmy Moray, Shikha Jain, Sridevi Chennamsetti, Harsha Thorve Abstract - This study aims to investigate the factors affecting the usage intent of Non-Fungible Tokens (NFTs) and its adoption among Gen Z and Millennials. Purposive sampling technique was used Using structured questionnaire, primary data was gathered and statistical tool SEM using SmartPLS was employed to assess the influence of factors under UTAUT2 model on usage intention. The findings reveal that users are inclined to adopt NFTs when perceived as easy-to-use and hassle-free. Likelihood of adopting is further facilitated with some conditions like adequate resources and support. Interestingly, habituality toward traditional digital assets had diverted effort from the steepness toward NFTs while hedonic motivation shows a lack of inclination for novelty over utility. Besides, performance expectancy and social influence play a big role, while the perceived cost acts as a showstopper. The study contributes to the existing body of knowledge and the stakeholder to estimates NFTs intent use.
Authors - Killol Pandya, Trushit Upadhyaya, Aneri Pandya, Upesh Patel, Poonam Thanki, Kanwarpreet Kaur, Jinesh Varma Abstract - The 2x2 Ultra Wide Band(UWB) MIMO antenna model with 31 x 26 x 1.56 mm3 are designed and presented. The proposed antenna system consists of two primary radiators over the surface layer and positioned with sufficient space between each other to receive adequate response. The conducting patches include U shaped and inverse U shaped slots to increase the current traveling surface. The bottom corners of the patches are having fractional geometry and supporting arms associated with microstrip line to offer resonances at desired frequencies. The Ultra Wide band characteristics are reported due to partial ground plane geometry. The additional strip with slots are deployed with ground plane to receive the satisfactory isolation as the radiators create mutual effect on the other ones which turns into performance degradation. For UWB, the structure exhibits gain of around 2 dB and around a bandwidth of 85% which shows the applicability of discussed research. The other diversity parameters such as Mean Effective Gain(MEG), Gain diversity, Channel Capacity Loss and Envelope Correlation parameter are analyzed and presented. The proposed antenna is well appropriate for WiMAX, WLAN and C band applications.
Authors - Siddhesh Joshi, Manoj Naidu, Athrva Kulkarni, Sahil Kadam, Nilesh P. Sable, Pranjal Pandit Abstract - Artificial Intelligence (AI) is a new type of experience that uses computer-generated content to augment the Real World (RW). An emerging type of experience known as augmented reality (AR) involves adding computer generated content to the real world (RW) that is connected to certain places and/or activities. AR is starting to show up in audio-visual media (news, entertainment, sports, etc.) and is starting to make a real and exciting appearance in other areas of lives (e-commerce, travel, marketing, etc.) [2]. This paper proposes the development of a marker-based AR system that overlays interactive 3D models of industrial machinery onto real-world views. The proposed project will be based on Unity software, which will make it feasible to present complex industrial equipment and move away from typical product manuals that are only text-laden. Users can intuitively see and interact with different parts of their machinery during maintenance, troubleshooting, or training through the AR-based system. This will probably ensure more user interaction with less learning time for a technical operation and is, therefore, most beneficial for the industries that heavily depend on machineries' setups and maintenance. This paper describes methodology, and the probable effects of the marker-based AR digitizing product manuals with first-hand observation in the future of digital documentation across industrial life.
Authors - Alka Beniwal, Trishna Paul, Mukesh Kumar Rohil Abstract - In the rapidly changing landscape of daily life, medical imaging stands out as a significant and novel component, significantly impacting healthcare practices. The efficiency of medical imaging processes is pivotal, and within this realm, accurate image registration emerges as a key contributor. Despite its recognized importance, existing pipelines lack a definitive and well-defined structure tailored to the specific requirements of medical imaging. This paper exclusively directs its focus toward addressing this gap by thoroughly examining and redefining the image registration pipeline within the context of medical imaging. The objective is to enhance the efficiency of medical imaging procedures by establishing a tailored and comprehensive pipeline that aligns seamlessly with the unique demands of this critical domain. We also analyzed the different application areas of image registration in medical imaging with their benefits, challenges, and future directions.
Authors - V V N Sai Rajeshwar, Gangalam Sumanth Abstract - Attributes like size, shape, color, quantity, quality, etc. are very hard to find in an image of a particular object. There are various ways to process an image and derive attributes namely – Faster R – CNN, SSD (Single Shot Detector), SPP (Spatial Pyramid Pooling) - Net. Many modern technologies use YOLOv8, the 8th version of YOLO (You Only Look Once), which is quite accurate and detects real - time objects. To enhance the proficiency and scalability of the model Deep Convolutional Neural Networks (DCNN) were used. This paper highlights a very unique strategy to enhance the abilities of YOLO with the help of dynamic masking and also advances the searching and findings of attributes in a unique way. The use of Compute Unified Device Architecture (CUDA) which uses GPUs rather than CPU, made it convenient to run on a higher loads of data. After masking an image or stream of images, detecting colors is easier, and expressing dominant colors is the motive of this research. Availing Coordinates of specified images in a particular image or stream of images is very helpful in various surveillance-related actions like military, and navy, and also useful in domestic purposes.
Authors - Ashita C. Kolla, Dattatray G. Takale, Parikshit N. Mahalle, Bipin Sule, Gopal Deshmukh Abstract - The research paper mainly focuses on algorithmic bias in facial recognition technology using parameters like race and hairstyle. It involves a CNN model following the pre-processing step of the data and custom annotation. It further talks about advanced methods for dataset balancing, such as normalization and sampling, along with detailed annotations involving characteristics of different races and hairstyles. Compared to other models, the CNN model contains powerful feature extraction methods and other bias mitigation methods such as adversarial training and annotation to enhance the chance of predictions. The results reveal that the model has made significant progress with good performance and lesser bias. This study helps the industry develop more reliable FRT systems with effective strategies for reducing bias and maintaining accuracy. These advancements are important for applications in various industries, where unbiased facial recognition is important for fairness and effectiveness.
Authors - Shashwat Avhad, Nikhil Chavan, Lalit Patil Abstract - An extensive overview of the surface electromyography (sEMG) methods for signal processing and their application to prosthetic hand control is given in this paper's abstract. Techniques for analyzing muscle activity to enable natural and accurate movement of the hands in prosthetic have advanced immensely as a result of growing interest in sEMG-based devices. This article discusses various techniques, such as wavelet transformation, machine learning-based algorithms, and time- and frequency-domain approaches, for feature extraction and classification from sEMG data. It also looks at how deep learning models have recently been included, and how it has helped to increase the precision and stability of sEMG signal classification. In addition, hybrid models that combine traditional statistical techniques with neural networks are investigated for their potential to improve prosthetic control precision and adaptability. The study tackles typical real-time signal recognition problems, like noise reduction and multi-degree freedom movement control management. The review's conclusion highlights the need for more study on multi-modal systems that use machine learning and sophisticated signal processing in order to enhance the usability and reliability of prosthetic devices.
Authors - Priya Deokar, Sandhya Arora Abstract - Multiparty Computation (MPC) allows users to use their private inputs to compute a function without revealing anything about the inputs. This feature is useful in computing the aggregation of smart meter readings. The smart grid provides bi-directional communication between the smart meter and the utility supplier. Periodic sharing of fine-grained information by a smart meter presents a greater danger to privacy. Secure multiparty computing is one of the most efficient ways to preserve privacy among users. This paper briefly overviews MPC techniques and their usage in Smart Grid communication.
Authors - Shrivardhansinh Jadeja, Bimal Patel, Jalpesh Vasa Abstract - The evolution of software development methodologies has profoundly influenced the gaming industry, marking a transition from traditional approaches to agile frameworks that emphasize flexibility and responsiveness. Traditional methodologies often struggled to meet the rapidly changing demands of game design and player expectations, leading to extended development cycles and reduced competitiveness. In contrast, agile methodologies have emerged as a viable solution, focusing on iterative development, continuous feedback, and collaboration among cross-functional teams. This paradigm shift has enabled companies like Riot Games to significantly enhance operational efficiency and swiftly respond to market dynamics. This research paper investigates Riot Games' agile transformation as a case study, illustrating how the adoption of agile practices has fortified its competitive advantage over alternative players in the gaming sector. By examining Riot's innovative team structures, leadership models, and iterative development processes, this study elucidates the critical role of agility in fostering creativity, improving product quality, and accelerating time-to-market. The findings highlight the necessity of embracing agile methodologies not only for individual organizations but also for the broader gaming industry seeking sustainability and growth in an increasingly competitive landscape. Ultimately, this paper offers valuable insights into how agile transformation can act as a catalyst for success in game development, providing a framework for other companies aiming to enhance their competitive positioning.
Authors - Pranav Indurkar, Mansi Dangade, Apoorva Kumar, Harsh Thakar Abstract - This paper explores the chip implementation of a low-power RSA encryption system, optimizing resource and Power usage while maintaining the security. The RSA Algorithm modeled using Verilog, implemented on the XILINX SPARTAN-7 FPGA (XC7S50-CSGA324-2) with a comparative analysis of 4- bit to 8-bit algorithmic parameters (such as p, q, e, d, M, C, n, phi_n). Two approaches are studied: one with uniform bit sizes of algorithmic parameters and another with smaller p & q bit sizes. Results show that the second approach yields better efficiency. Future work with CADENCE CIC tools will further optimize power consumption. This work offers insights into designing low-power RSA encryption chip for modern digital systems.
Authors - Tamanna Kalariya, Bimal Patel, Martin Parmer, Mrugendra Rahevar Abstract - The Agile model has demonstrated clear superiority over traditional software process models by offering enhanced flexibility and adaptability to evolving project requirements. Unlike rigid, linear methodologies, Agile promotes iterative development, allowing teams to quickly incorporate feedback and adapt to changing needs. This iterative approach is particularly advantageous for software testers, enabling continuous testing throughout the development lifecycle, which improves code quality and reduces defects. GitHub, as a collaborative development platform, amplifies these Agile benefits by providing robust version control and seamless integration with Agile workflows. For testers, GitHub facilitates real-time collaboration, enabling immediate feedback and shared responsibility for quality assurance. Features such as pull requests, issue tracking, and continuous integration streamline the testing process and enhance communication between developers and testers. Additionally, GitHub’s Agile-driven workflow provides a competitive advantage over other code-sharing tools by optimizing development cycles and fostering a strong collaborative culture. The platform's rich ecosystem of tools, combined with extensive community support, further boosts productivity and innovation. This paper examines how Agile methodologies within GitHub enhance both the development and testing processes, positioning GitHub as a leader in the collaborative development landscape and driving greater efficiency in modern software engineering practices.
Authors - Rashmy Moray, Sejal Vaishav, Sangam Dey, Sridevi Chennamsetti, Harsha Thorve Abstract - This paper investigates the impact of behavioural biases, specifically Loss Aversion, Regret Aversion, Reference Dependence, and Risk Perception on algorithmic trading using the framework of Prospect Theory. Using structured questionnaire, the Primary data was collected from the traders who use algorithm. Statistical tool, SmartPLS was employed to assess the endogenous factors and the behavioural biases that influence the intention to trade using algorithms. The findings indicate that Risk Perception and Reference Dependence significantly impact trading intent, whereas Loss Aversion and Regret Aversion do not show a significant influence on trading intent. This advocates that the systematic and emotion-free nature of algorithmic trading minimizes the effects of certain emotional biases. The study contributes profound understanding of behavioural biases of traders adopting algorithm offer distinctive path for future scope of research.
Authors - Aditi Vanikar, Rana Vanikar, Mihir Sardesai, Rupesh Jaiswal Abstract - Preserving the security and performance of a network in today's networked environment depends upon monitoring and analyzing network traffic. The paper discusses our system overview that analyzes live network traffic from a mirrored port by means of DPI. It includes the provision for gaining insight into the usage of networked applications by sorting packets into particular categories, such as HTTP, video streaming, or other protocols. The technology employs state-of-the-art machine-learning algorithms and categorization in order to identify and alert potentially damaging packets in a timely manner. Upon detecting any threats, the system alerts the user so that he or she can take appropriate preventive measures. The hybrid strategy ensures improved visibility and security of network environments through concurrent, real-time threat detection and traffic classification.
Authors - Nishit Patel, Bimal Patel, Mansi Patel, Parth Shah, Nishat Shaikh Abstract - Agile model has demonstrated clear superiority over traditional software process models by offering enhanced flexibility and responsiveness to changing requirements, particularly essential in the fast-paced automotive industry. Prior to adopting Agile, Tesla encountered significant challenges, including prolonged development cycles and difficulties in scaling manufacturing processes to meet increasing demand. These constraints limited Tesla’s ability to innovate rapidly and deliver advanced features such as Full Self-Driving (FSD) capabilities. By implementing Agile methodologies, Tesla transformed its software development and manufacturing ramp-ups, enabling iterative development with continuous testing and integration. This shift significantly accelerated FSD development, allowing for faster adaptation to technological changes. The adoption of Agile led to measurable improvements, including reduced time-to-market for new features, improved cross-functional collaboration, and enhanced product quality. Compared to other automotive giants, Tesla’s Agile integration has provided a distinct competitive advantage in terms of operational efficiency and innovation. While competitors often rely on established, rigid processes, Tesla’s Agile approach enables rapid adaptation to market demands and continuous consumer feedback. This paper highlights the transformative impact of Agile methodologies on Tesla’s product development and operational strategies, demonstrating how Agile has positioned Tesla as a leader in automotive innovation, driving significant advancements in technology and customer satisfaction.
Authors - Padmanabh khunt, Martin Parmar, Het Khatusuriya, Mrugendra Rahevar, Bimal Patel, Krunal Maheriya Abstract - The smart garage system presented in this paper incorporates advanced security and remote-control functionalities to enhance the user experience and ensure secure access. The implementation of a One-Time Password (OTP) authentication mechanism provides an additional layer of security, effectively preventing unauthorized access to the garage. Central to the system are ESP32 microcontrollers, which facilitate reliable and efficient communication between the keypad, relay module, and the mobile application. Utilizing LoRa communication, the system achieves long-range wireless connectivity, enabling seamless interaction between ESP32 microcontrollers even in areas with limited network coverage. The mobile application, developed using React Native, offers a user-friendly interface for homeowners, featuring login/signup options, direct garage door control, and OTP generation for secure access. A robust server backend, built with Node.js and supported by a MongoDB database, ensures efficient management of user data, including login credentials and generated OTPs. Furthermore, an admin panel is integrated to enhance user administration and access control capabilities. This comprehensive smart garage system not only improves security but also provides convenience and reliability for modern homeowners.
Authors - Ritika Patki, Srushti Jamewar, Tanika Mathur, Mridula Korde Abstract - High temperatures accelerate the spoilage of dairy products, particularly milk, posing significant challenges for food safety and waste management. This study presents a novel sensor-based detection system designed to monitor milk spoilage by measuring real-time changes in pH levels and carbon monoxide (CO) concentrations, which are key indicators of microbial activity and biochemical shifts in milk. The system utilizes an Arduino UNO microcontroller integrated with a pH sensor and an MQ-7 gas sensor to assess milk freshness through a combination of pH and CO data. Results are displayed on an LCD screen with intuitive indicators of freshness status—categorized as "fresh," "not so fresh," or "spoiled"—ensuring ease of interpretation for users. Experimental validation across various milk samples demonstrates the system’s effectiveness in early spoilage detection. Future developments may focus on non-invasive methods and IoT-based miniaturization, incorporating machine learning algorithms for residual life prediction through blockchain integration. This approach promises to reduce food wastage and enhance food security, offering an accessible and sustain-able solution for milk spoilage detection.
Authors - Aarti Agarkar, Ayush Sasane, Amankumar Kumare, Aditya Kadlag, Gaurang Khanderay Abstract - This paper presents a highly accurate intelligent traffic sign recognition system, which has potential to greatly improve safety on roads and make autonomous driving possible. It uses Convolutional Neural Networks (CNN) to detect and classify traffic signs, accurately robustly in real-time systems. The process involves pre-processing a large number of images, training the model and performing real-time detection with OpenCV in Python. The Node MCU Microcontroller is integrated to make the communication and responses more stable and can be set automatically after recognizing traffic signs. The findings show improved precision and stability of the properties that are essential for consideration in autonomous vehicle systems, where harmonious driving is needed to upgrade latest architectures.
Authors - Varalatchoumy M, Syed Hayath, Dinesh D, Dhanush C P, Manu R, V Sadhana Abstract - This paper presents an advanced Generative AI-powered system for video-to-text summarization, leveraging state-of-the-art Computer Vision (CV) technologies and Natural Language Processing (NLP) techniques. The developed system addresses the growing need to extract key information efficiently from lengthy videos across diverse domains such as education, entertainment, sports, and instructional content. By integrating visual and textual data, it pinpoints essential moments and generates concise summaries that capture the core message of the video, reducing the time users spend understanding extensive media. At the heart of this system lies a robust, open-source large language model (LLM), finetuned to produce human-like summaries from video transcripts. The system processes visual cues using advanced CV techniques—such as keyframe extraction and scene segmentation—and textual cues via Automatic Speech Recognition (ASR), which converts audio into text. This dual approach facilitates a deep understanding of spoken and visual content, ensuring that summaries are precise, relevant, and contextually accurate. The system has been evaluated on a diverse dataset, comprising videos of various genres, qualities, and lengths, demonstrating its capability to generalize effectively across a wide spectrum of content. Applications of this video summarization tool include content management, video indexing, educational platforms, and beyond, offering significant time-saving benefits to users and organizations. By incorporating real-time feedback, the system continuously refines its summarization techniques, enhancing accuracy and ensuring that users quickly access the most relevant information, thereby promoting greater accessibility and usability of video content.
Authors - Rajlaxmi Sunil Sangve, Riya Jha, Bhagyashri Narale, Sakshi Hosamani Abstract - Kidney disease is an asymptomatic disease, which leads to severe complications or even mortality if not diagnosed early. Routine diagnostic methods, such as serum-based tests and biopsies, are either less effective in the early stages of the disease. This paper proposes an automatic detection of kidney disease using CNNs applied to medical imaging data. Our model is designed to analyze computed tomography (CT) images for the identification of kidney disease, classifying normal and tumors. The proposed CNN architecture leverages deep learning techniques to extract features from these images and classify them with high accuracy. This paper aims to build a system for detection of kidney disease using CNN, based on a public dataset sourced from Kaggle. The paper involves several key stages, initiated from raw data preprocessing and feature selection, followed by training and evaluating machine learning model using CNN. Our proposed model demonstrated superior performance in kidney disease detection, achieving an accuracy of 95%.
Authors - Vidisha Deshpande, Gauri Shelke, Bhakti Kadam Abstract - Advancements in deep learning are fundamentally transforming assistive technologies, providing visually impaired users with unprecedented access to information and enhanced interaction with their surroundings. This paper comprehensively surveys traditional and emerging assistive technologies, focusing on real-time image caption generation systems. The modern advancements that bridge sensory limitations and digital interaction by covering a range of technologies such as Optical Character Recognition (OCR)-based text readers, object detection systems, image captioning systems, and intelligent haptic feedback devices are highlighted. In particular, the critical role of vision-language models and multimodal systems, which enable real-time auditory descriptions of visual scenes is studied. The survey also identifies significant gaps in real-world applications, particularly in terms of adaptability, cost, and inclusivity. These findings emphasize the need for more accessible, affordable, and real-time solutions that cater to the diverse needs of visually impaired individuals.
Authors - Chetana Shravage, Shubhangi Vairagar, Priya Metri, Shreya C. Jaygude, Pradnya P. Sonawane, Pradnya D. Kudwe, Siddheshwari S. Patil Abstract - This project presents an innovative phishing detection system that addresses the limitations of traditional methods by combining URL-based and content- based features to accurately identify fraudulent websites. Unlike conventional approaches that rely heavily on blacklisting and heuristics, which struggle with zero-day attacks and frequent updates, this system employs machine learning algorithms to automatically extract and analyze critical features from URLs and webpage content. By leveraging a comprehensive dataset that consists of fraudulent (phishing) websites along with legitimate websites, the system aims to improve detection rates to optimize performance based on evaluation metrics like accuracy, precision, F-1 score, recall, and false-positive rates. The system makes use of selective machine learning models like Random Forest, Decision Tree, and Support Vector Machine (SVM), which provide the benefit of increased scalability, robustness and improved effectiveness in phishing detection. Ultimately, this project aims to deliver a scalable, real-time detection solution that effectively mitigates phishing threats in a rapidly evolving landscape.
Authors - Aman Bhimrao Kamble, Shafi Pathan Abstract - Deep Q-Learning (DQL) has emerged as a promising method for enhancing Network Intrusion Detection Systems (NIDS) by enabling dynamic and adaptive detection of evolving network threats. This review examines the strengths, limitations, and potential enhancements of using DQL in NIDS. Even while DQL increases the accuracy of anomaly detection and manages massive amounts of network data, it has drawbacks such slow convergence, high processing costs, and vulnerability to adversarial attacks. This study proposes improvements to overcome these problems, including efficient reward systems, hybrid architectures that combine DQL with other machine learning models, and continuous learning to adapt to changing threats. Recommendations for further research to enhance DQL's efficacy in real-time intrusion detection are included in the study's conclusion.
Authors - Renuka Sandeep Gound, Farhan Mujawar, Niraj Dhakulkar, Payal Rathod, Saish Bhise, Kavita Moholkar Abstract - With the growing interest in naturopathy and holistic health, it is more important to make information about medicinal plants easily accessible. The Virtual Healing Garden is designed to do just that, offering a dynamic platform where AYUSH students, professors, and plant enthusiasts can explore a variety of medicinal plants in a hands-on, engaging way. By blending traditional healing knowledge with modern scientific research, the platform brings these plants to life through interactive 3D models. Using cutting-edge algorithms like photogrammetry and 3D modeling, the garden creates realistic representations of the plants, giving users a detailed and immersive experience. The Virtual Healing Garden will feature a rich database of 3D plant models, each paired with detailed information about their medicinal properties. This will provide users with a visually immersive and informative resource, making it easy to explore and learn about various plants in a more engaging way, whether for education or research. This virtual garden not only raises awareness of the health benefits of medicinal plants but also makes learning about them interactive and fun.
Authors - Chetana Shravage, Shubhangi Vairagar, Priya Metri, Rohit Rajendra Kalaburgi, Harsh Anil Shah, Abhinandan Vaibhav Sharma, Shubham Shatrughun Godge Abstract - Heart sound categorization is critical for the early identification and detection of cardiovascular illness. Recently deep learning methods have resulted in promising improvements in the correctness of heart sound classification systems. This work introduces a unique transformer-based model for heart sound classification that uses powerful attention mechanism to capture both local as well as global dependencies in heart sound data. Transformers, in contrast to traditional models that rely on handmade features or recurrent networks, can dynamically focus on the most important characteristics in time-series data, making them perfect for dealing with the complexity and variability of phonocardiogram (PCG) signals
Authors - Shannon D’Souza, Rashmy Moray, Sridevi Chenammasetti, Shikha Jain Abstract - This study explores factors affecting investors' intentions to utilize Peer-to-Peer (P2P) lending platforms using the DeLone and McLean Information Systems Success (ISS) model and the Technology Acceptance Model (TAM). Data was collected through an online survey, yielding 283 valid responses from 350 distributed questionnaires, using snowball and convenience sampling. Structural Equation Modelling (SEM) with SmartPLS validated relationships between system quality, information quality, service quality, user satisfaction, social influence, and continuous intention to use. Findings indicate that system quality and service quality have a significant positive predictive effect on user satisfaction, while social influence positively affects ongoing usage intentions. Improved system performance, information accuracy, and service responsiveness can foster investor trust, platform adoption, and retention, guiding stakeholders and regulators toward an environment of stable and successful P2P lending.
Authors - Pankaj Chandre, Palash Sontakke, Rajkumar Patil, Bhagyashree D Shendkar, Viresh Vanarote, Dhanraj Dhotre Abstract - In today’s digital landscape, the prevalence of scams, phishing, and malicious attacks poses significant risks to both individuals and organizations. Mitigating these threats requires a comprehensive cybersecurity strategy that begins with user awareness and extends to robust protective measures and incident response protocols. By integrating education, proactive defenses, and responsive actions, personal and organizational cybersecurity can be greatly enhanced. Mitigating scams, phishing, and malicious attacks requires a comprehensive approach to cybersecurity and personal protection. This strategy begins with the User Environment, where devices connected to the internet become vulnerable to threats. Education and Awareness play a crucial role, providing training on recognizing phishing attempts and setting up reporting mechanisms to flag suspicious activities. Building on this, Protective Measures such as strong passwords, multi-factor authentication, regular software updates, and the use of security tools strengthen the defenses against cyber threats. Should an attack occur, Incident Response protocols are activated, including the detection and investigation of incidents, followed by recovery actions to restore security and prevent future attacks. By integrating these layers of defense, individuals and organizations can significantly reduce the risks of cyberattacks and safeguard sensitive information.
Authors - Shubhangi Vairagar, Chetana Shravage, Priya Merti, Nikhil M. Ingale, Sakshi N. Gaikwad, Dhruv G. Yaranalkar, Atharva R. Pimple Abstract - Artificial Intelligence, specifically large language models and generative AI, has dramatically changed the finance industry. According to a few research studies, the current article discusses some of those findings that provided insight into different applications of artificial intelligence in the sector of financial operations and improved predictive analytics, operational efficiency, and quality in decision-making processes. Most critical findings point towards the efficiency of LLMs, where it has achieved automation, financial analysis improvements, and does comply with enforcement standards in their usage of such technologies as towards most security concerns, privacy, and ethical perspectives. Specifically, the long-term implications for financial decision-making and the potential consequences arising from the use of such technologies in an ethical view stand out starkly as red flags of concern. Thus, the review brings new knowledge in the sphere of AI in finance and grounds further justification to be done with proper research motives toward complete responsible development of AI.
Authors - Kumbhar Sanjivanee Rajan, Kulkarni Prachi Prashant, Pawar Prachi Baghwan, Patara Diya Milan, Abira Banik Abstract - “Holistic Heal” is an app developed using Flutter, tailored to address the growing interest in alternative healthcare options such as Ayurveda, Yoga, and Naturopathy. The app simplifies the process of discovering and accessing these specialized hospitals and wellness centers, making holistic healthcare more accessible to users. Harnessing the capabilities of geolocation services enables users to find the nearest facilities, offering detailed information for each. The app also ensures a user-friendly experience. Built with a focus on promoting holistic well-being, “Holistic Heal” showcases the potential of technology to enhance traditional healing practices and empower individuals in their journey towards a healthier, more balanced lifestyle. Depending on their demands, the patient can search the hospital. Upon the patient’s request, this application offers the hospital and physician details that are currently accessible. A suggested application has been designed to find the closest hospital with the requested medical specialty. Hospitals’ nearest locations are identified using the Global Positioning System (GPS), which provides real-time geographic data by triangulating signals from satellites. This data, integrated into smartphones, is combined with Google Maps Application Programming Interfaces (APIs) to determine optimal routes from the user’s current location to hospitals, accounting for road networks, traffic, and travel modes. A patient can use this application to discover the closest hospital based on the availability of expert consultants. This application that was built is easy to use and effectively gives patients the necessary information.
Authors - Swati Kiran Rajput, Sunil Gupta Abstract - Diabetes mellitus is the root cause of a disease known as diabetic retinopathy, which is a disorder that affects the retina. In every region of the globe, it is the leading cause of blindness. Early detection and treatment are very necessary in order to delay or avoid the deterioration of vision and the loss of eyesight. The scientific community has proposed a number of artificial intelligence algorithms for the aim of identifying and classifying diabetic retinopathy in fundus retina pictures due to the fact that this is the intended objective. Utilizing a Convolutional Neural Network (CNN), we suggested a method for the identification and early prediction of diabetic retinopathy in this particular piece of research. Using a wide variety of hyper parameters, such as epoch size, batch size, optimized, and so on, the Deep CNN has been used for both training and testing purposes. Examples of normal and abnormal retinal pictures have been included in the MRI dataset. When the findings of the experimental investigation were evaluated using machine learning and deep learning algorithms like SVM, ANN, and CNN, the results were shown to be accurate. In conclusion, the CNN achieves a detection and prediction accuracy of 96.60%, which is superior than that of the SVM and other artificial neural networks.
Authors - Ashutosh Patil, Gayatri Bhangle, Sejal Kadam, Poonam Vetal, Prajakta Shinde Abstract - Advances in deep learning have enabled effective applications in agriculture, including fruit disease detection. Accurate identification of diseases in fruits such as Annonaceae and Rutaceae families is crucial for yield and quality improvements. Many studies employ deep learning models like CNN, ResNet, VGG, and DenseNet for disease detection across fruits such as apple, orange, guava, and grapes. This article reviews recent research on deep learning for fruit disease detection and classification, focusing on model performance, data utilization, and visualization techniques. We analyze existing studies to identify optimal strategies for fruit species and other underrepresented crops, outlining challenges and areas for future research on various types of fruit species and their family.
Authors - Malay Shah, Sayal Goyal, Rashmi Rane, Ruhi Patankar, Sarika Bobde, Arnav Jain Abstract - This study investigates the impact of risk-taking on football match outcomes, focusing on player substitutions. The analysis reveals that risk-taking propensity peaks when a team is trailing by 2-3 goals and diminishes when leading by the same margin. Younger managers outperform middle-aged ones in risky decisions, while older managers excel in later substitutions. Additionally, a manager's tenure with the team increases the effectiveness of risk-taking, particularly in earlier substitutions and stronger teams. This study also emphasizes the importance of mental state in player performance, proposing a framework combining Match Score Analysis (Kaplan-Meier Fitter) and Score Analysis to evaluate players' mental stability and survival rates during the game. By integrating these models, teams can make better-informed decisions regarding substitutions, considering both past performance and mental health, ultimately enhancing match outcomes. This research underscores the synergistic potential of combining black-box causal machine learning with interpretable models, offering valuable insights for football management and beyond.
Authors - Nilesh Deotale, Nafiz Shaikh, Ashwin Katela Abstract - WhatsApp chats consist of various kinds of conversation held among two people or a group of people. This chat consists of various topics. This information can provide a lot of data for the latest technologies such as Machine Learning. The most important things for Machine Learning models are to provide the right learning experience which is indirectly affected by the data we provide to the model. This tool aims to provide in depth analysis of the data which is provided by WhatsApp. Irrespective of whichever topic the conversation is based on, our developed code can be applied to obtain a better understanding of the data. The advantage of this tool is that it is implemented using simple python modules such as pandas, matplotlib, seaborn, streamlit, NumPy, re, emojis and a technique sentiment analysis which are used to create data frames and plot different graphs, where then it is displayed in the streamlit web application which is efficient and less resources consuming algorithms, therefore it can be easily applied to larger dataset. The Accuracy of this project is 75%. In summary, this project makes use of state-of-the-art data analysis technologies such as scikit-learn, Topic Modeling, Named Entity Recognition, Clustering, Word Embeddings, Natural Language Toolkit, and more. Sequence-to-Sequence Models, Text Classification and so forth. Users can better understand how others communicate by using Language Model Fine-Tuning to extract relevant information from WhatsApp discussions.
Authors - Susheela Vishnoi, Monika Roopak, Prashant Vats Abstract - Pathological tissue image categorization is essential in medical diagnostics, offering insights into disease types, progression, and treatment alternatives. The significant variability in tissue morphology and the overlapping visual patterns across different classes complicate accurate categorization. This study introduces an improved categorization model utilizing a bag-of-features (BoF) methodology integrated with the Roulette Wheel Whale Optimization Algorithm (RWWOA) to enhance classification accuracy and optimize feature selection efficiency. The proposed model utilizes the Bag of Features (BoF) technique to extract discriminative features from tissue images, thereby generating a feature-rich dictionary that represents various pathological structures. The RWWOA is employed to optimize feature selection, thereby reducing dimensionality and concentrating on the most pertinent features for precise categorization. Our method integrates the exploration capabilities of the Whale Optimization Algorithm (WOA) with the probabilistic selection mechanism of the roulette wheel, thereby dynamically balancing exploitation and exploration, which enhances convergence speed and categorization accuracy. Experimental results indicate that the RWWOA-BoF method outperforms traditional methods across various datasets, showing enhancements in classification precision, recall, and F1-score. This method offers a reliable resource for aiding pathologists in diagnostic imaging, which may expedite diagnostic processes and improve consistency in clinical practice.
Authors - Sanjana, Sukanya Sharma, Dipty Tripathi Abstract - Handwritten digit recognition is a key application in image processing and pattern recognition, with wide usage in areas such as postal services, banking, and mobile applications. This research paper presents a performance comparison between traditional machine learning models and deep learning models for accurate handwritten digit classification. The study focuses on developing a mobile application using Flutter integrated with TensorFlow Lite and Firebase to deliver real-time predictions. The app performs preprocessing on input images and employs model inference for efficient and accurate digit recognition. The objective is to determine the most effective model in terms of speed and accuracy for on-device predictions, emphasizing usability and real-time response
Authors - Yogini Prasanna Paturkar, Amol P. Bhagat, Priti A. Khodke Abstract - Social media platforms generate vast amounts of interaction data, offering valuable insights into user behavior, preferences, and trends. However, the sheer volume and velocity of this data pose significant challenges for real-time analysis and computational efficiency. This paper proposes a framework for random sampling of social media interaction data to address these challenges. By employing probabilistic sampling methods, we aim to reduce data volume while preserving key statistical properties and minimizing information loss. The proposed methodology leverages stratified and weighted random sampling techniques to ensure the representation of diverse user groups and interaction types. Applications of this approach include sentiment analysis, trend detection, and user behavior modeling. Preliminary experiments demonstrate that random sampling can achieve a significant reduction in computational overhead while maintaining analytical accuracy within an acceptable margin of error. This framework has the potential to enhance data processing pipelines in fields such as marketing, public opinion analysis, and event monitoring, enabling timely and resource-efficient decision-making.
Authors - Rowan Cowper, Grant Oosterwyk, Jean-Paul Van Belle Abstract - The rapid and widespread growth of AI use has brought about a number of important areas for research. This paper aims to examine the human factors that impact AI use: whether demographic attributes, trust in AI, and perceptions about AI influence whether someone will use AI or not. A survey among South African industry and academic respondents was used. The key findings of this study include that age and education have a significant impact on trust in AI. Domain knowledge and education levels were significant indicators of perception of AI, with higher levels of domain knowledge and education leading to lower perceptions of AI. Both AI trust and perception were found to have a significant positive impact on whether someone made use of AI or not. These findings may inform decision makers on targeted interventions, such as education, to increase the use of AI in industry and academic contexts. Hopefully further academic research will also validate our findings in other research contexts, such as India and/or different population segments.
Authors - Rupan Panja, Rajani K. Mudi, Nikhil R. Pal Abstract - The Support Vector Machine (SVM) is a popular classification algorithm, however, it suffers from the drawback that the classification time for an unknown data point is proportional to the number of support vectors (SVs). Thus, its application for real-time decision-making, especially for complex class boundaries becomes problematic / impossible. In this article, we propose an algorithm, Self Organizing Support Vector Machine (SO-SVM), which can decrease the number of SVs without compromising the accuracy. In this algorithm, we first cluster the data points using self-organizing map to find potential class boundary points, which are crucial for determining the separating hyperplane in an SVM. The separating hyperplane is then learned from the selected boundary points, leading to a reduction in the number of SVs. Learning the SVM also becomes efficient because of the reduced number of data points. The proposed algorithm has been tested on a number of benchmark datasets and found to decrease the number of SVs without deteriorating the classification accuracy. The SO-SVM algorithm thus can be an efficient alternative to the SVM algorithm and thus can be applied instead of normal SVM without making a noticeable compromise with the performance.
Authors - Riya Reddy, Ayoush Singh, Keshav Tanwar, Rashmy Moray, Shikha Jain, Sridevi Chennamsetti Abstract - This study examines the underlying drivers of usage intention of chatbots by bank customers using UTAUT model. This model assesses four key factors determining the usage intent of chatbots: performance expectancy, effort expectancy, social influence, and facilitating conditions. Primary data was obtained through structured questionnaire from retail banking customers. Structural Equation Modelling method is used and statistical tool SmartPLS 4.0 was employed to analyze complex associations between variables. The results show that performance expectancy, facilitating conditions, and motivation have a significant association with the usage intention of chatbots; however, effort expectancy and social influence have no connection. In addition, the results show that perceived security and trust are essential criteria for adopting a chatbot in banking. The study contributes value to the existing body of knowledge by proving the factor influencing the intent usage of chatbots in the banking industry. It also highlights future work directions in the form of long-term impacts and insights that can guide banks in designing customer-centric AI systems and improving chatbot services.
Authors - Mihlali Mqoqi, Marita Turpin, Jean-Paul Van Belle Abstract - The manufacturing industry is undergoing significant changes due to the emergence of artificial intelligence (AI) technologies. This transition has implications, particularly with regards to the skills necessary to adopt and leverage these technologies effectively. This paper addresses the issue of the expanding gap between the skill set in the current workforce and those required for the future of work. A systematic literature review was conducted to investigate these skills, which resulted in the selection of 28 studies out of 216 initially identified that offered insights into the benefits and applications of AI in manufacturing. The findings show how AI has altered manufacturing processes through predictive maintenance, quality assurance, and product design. Further, there is a need for targeted upskilling and reskilling programs to bridge the current skill gap and equip the workforce to meet the changing demands of the industry. Initiatives that could be implemented for successful skills development are also discussed.
Authors - Nitesh Pradhan, Aryan Baghla Abstract - Electrocardiogram (ECG) signals are effective indicators for detecting obstructive sleep apnea (OSA) due to their ability to reflect physiological changes associated with apnea events. EfficientApneaNet is a deep learning-based model for the detection of OSA from a single-lead ECG. In general, traditional approaches are reliable but exhaustive and costly; therefore, ECG-based methods are being sought. Earlier machine-learning methods, including Support Vector Machines and Random Forests, are prone to real data noise. Based on these, EfficientApneaNet joins previous advances in deep learning for further improvement in accuracy and robustness. The proposed architecture is powered by three novelties, namely: the ENBlocks, inspired by EfficientNet, Squeeze-and-Excitation blocks, and attention. ENBlocks make use of depthwise separable convolutions that reduce the computational complexity and amplify the efficiency of spatial feature extraction. The Squeeze and Excitation blocks carry out channel recalibration to focus on relevant patterns, while the attention mechanism underlines the critical temporal events within the ECG sequences. On the Apnea-ECG dataset, EfficientApneaNet realizes state-of-the-art performance with 91.47% accuracy, 85.92% sensitivity, and 94.91% specificity outperforming those of leading existing CNN-LSTM hybrids. It adopts Adamax optimization for stability while implementing the technique of cosine annealing for LR scheduling. The residual connections avoid gradient vanishing and explosion. Through ablation studies, it is confirmed that SE blocks and the attention mechanism are both essential to achieving high sensitivity and high specificity. In this respect, EfficientApneaNet would be considered a significant improvement in OSA detection, as it has successfully handled spatial and temporal complexities in ECG data.
Authors - Soham Kulkarni, Suhani Thakur, Twsha Vyass, V. Yasaswini, Pooja Kamat Abstract - Dyslexia is a learning disorder that causes difficulty in reading and identifying relations between words and letters.[1] To improve reading accessibility within dyslexic patients, this study aimed to develop models for summarization of documents as well as provide a streamlined method to convert documents into dyslexia-friendly versions. Within this study, several summarization models were tested to generate effective text summaries, while defining Python functions to convert regular text into dyslexia friendly text. Models attempted with are: Term Frequency- Inverse Document Frequency Summarizer, Term Frequency-Inverse Document Frequency Summarizer with Support Vector Machine, and BART Transformer. After analysing the results, the BART Trained Summarization Model results are fruitful having a ROUGE R-1 F1 Score of 0.4510, a R-2 F1 Score of 0.2571 and a R-L F1 Score of 0.4177, ultimately successfully generating dyslexia-friendly summarized documents.
Authors - Anupama Pandey, Anubha Dwivedi, Rashmy Moray, Vivek Divekar, Shikha Jain, Sridevi Chennamsetti Abstract - The objective of the research is to identify the factors that affects the adoption of digital currencies by the Generation Z and the millennials, utilizing UTAUT model. It highlights the role of performance expectancy, effort expectancy, social influence, and facilitating conditions on the use of digital currency. A designed questionnaire was used to collect the primary data from Gen Z and millennials. Statistical techniques SmartPLS was used to analyse the correlation and impact of the various UTAUT constructs on the intent to use digital currency. Results highlight that performance expectancy and user influence have the significant influence on the adoption of innovative solutions. The paper also describes issues concerning the digital infrastructure and various regulations. Prospective strategies to increase the acceptability of the digital currency with the young people are advanced to articulate the perceived usefulness and to make the use of digital currencies easier.
Authors - Arun Kumar S, Niveditha S, Vikas K B, Varshini C, Hemalatha H U Abstract - Improved patient outcomes and effective management of Chronic Kidney Disease (CKD) depend on early detection, making it a major worldwide health concern. This study presents a hybrid machine learning model that combines Random Forest and LightGBM classifiers in order to accurately predict CKD stages. The dataset from Kaggle was used to build the model, and SMOTE was used to handle class imbalances in addition to feature engineering and data preprocessing. With regard to test data, the model's accuracy was 97.7%. Enhancing its actual clinical utility, a web-based tool was also developed to enable real-time forecasts of CKD stage based on patient inputs. Using a mixed-method approach that included more than 1600 participants' medical examinations and surveys With the integration of real-time clinical data and strong predictive models among its potential enhancements, the proposed approach provides a robust and easily utilized tool for clinical applications.
Authors - Mihir Deshpande, Hrishikesh Patkar, Madhuri Wakode Abstract - Intelligent Document Processing (IDP) automates extraction and categorization of information from unstructured and semi-structured documents. While standard Optical Character Recognition (OCR) computerizes the parsing of printed and scanned documents, so-called parsing tools like PyPDF2 and React-PDF take on the extraction of text from digital files. Emerging developments in IDP apply Large Language Models (LLMs) to enhance Natural Language Processing (NLP) to ensure the efficient interpretation, classification, and analysis of the extracted data. This paper provides a survey of IDP technologies with possible applications in financial and retail sectors-invoice processing, purchase order matching, and fraud detection. It also focuses on accuracy, scalability, and outlook of LLM-based IDP on the cloud, towards next-generation automation in more detail.
Authors - G.G. Rajput, Sumitra M Mudda Abstract - This paper presents abnormality detection from segmentation techniques for leg fracture segmentation from animal x-ray images. Gaussian filtering is used to remove the noise from the x-ray images. fracture is segmented from x-ray image by performing thresholding segmentation operations. Experiments are performed on clinical data set to present the severity of the fracture in images for threshold segmentation methods studied. Extract the Features from segmented images using GLCM techniques. SVM algorithm is use for classify the given animal x-ray image is fractured or not. Using thresholding segmentation techniques, fractures are separated from x-ray images. Utilizing GLCM techniques, extract the features from segmented images. The SVM method is used to determine whether or not the provided animal x-ray image is broken.
Authors - Nilesh Korade, Swaraj Waykar, Archit Waghmode, Shweta Tate Abstract - Though solo travel can be an enriching and adventurous experience, staying alone (without social interactions) can sometimes enhance feelings of loneliness and this can be detrimental to a traveler's mental health. In this paper, we present an AI Travel Buddy. This intelligent system uses real-time facial expression recognition to determine the emotion of travelers at the moment and provides personalized conversational topics based on the detected emotion using the emotion-aware voice assistant. Using the latest CNN architecture for emotion detection, and Conversational model for dynamic answers, the AI Travel Buddy is a companion to travelers, delivering emotional support to users on solo trips. The paper describes the technologies used, the system architecture, and its methodology, showing substantial improvement in user engagement emotionally alongside providing overall higher travel satisfaction.
Authors - Sahana Balajee, Sakhi Saswat Panda, Anushruth Gowda, Animesh Giri Abstract - The extension of conventional Internet of Things (IoT) technologies from basic applications to advanced industrial processes has revolutionised operational efficiency, giving rise to a new class of IoTs called the Industrial Internet of Things (IIoT) systems. Despite these advancements, critical IIoT infrastructures remain exposed to a growing range of cybersecurity threats due to open vulnerabilities, stemming from the fact that these IIoT networks try to interconnect physical machinery with digital systems. Cyber attacks targeting IIoT systems attempt to exploit these vulnerabilities in order to cause severe operational disruptions and costly outages. These intrusions pose significant risks to data integrity, operational continuity, and total production, highlighting the need for effective cybersecurity measures in the industrial setting. To help prevent such intrusions this work proposes a unified predictive framework combining cyber attack detection with outage prediction to enhance resilience in IIoT environments. By leveraging machine learning algorithms, this framework analyses two types of data - network traffic for cyber threats and historical sensor data for forecasting the Remaining Useful Life (RUL) of critical components. This approach aims to identify risks and send alerts to minimise operational downtime preemptively, integrating predictive techniques for both cybersecurity and system reliability to strengthen IIoT systems, thereby helping industries maintain continuous, secure, and safe operations.
Authors - Rakhi Bharadwaj, Nikhil Patil, Riddhi Patel, Raj Pagar, Suchita Padhye Abstract - Integrated AI within the context of mental health has continued to be an emergent area of interest as more and more mental health applications in mobile devices, include chatbots. Such chatbots are the focus of this paper, and their design based on NLP, machine learning, and sentiment analysis is illustrated in order to help people who suffer from anxiety and depression providing them with individual therapeutic support. Despite these applications equip users with means of tracking moods, sharing concerns and getting ideas about mental health, they are not a license to practice. The study also responds to fundamental questions about their usability, usefulness, privacy, security, and protection of data. Because the need for mental health services continues to increase in parallel with technological progress, one has to maintain the harmony between innovation and ethics. Discussing the current mental health platforms, their technological model, and case studies of their usage, this paper also emphasizes the related risks – the misuse of AI in mental health treatment. Leaving aside the beautiful language and the fascinating discussions with the machines, the goal of the study is to provide theoretical perspectives as well as pragmatic suggestions for way AI can be used to improve mental health care responsibly.
Authors - Sunil Sangve, Rutuparn Kakade, Saamya Gupta, Sahil Akalwadi, Soham Pawar Abstract - The paper introduces a novel and detailed approach for automated generation of fully customizable and user-centric food recipes. The system utilizes fine-tuned GPT2 and fine-tuned YOLOv9 to work efficiently. The user can input the available ingredients by uploading an image which is then processed by the fine-tuned YOLOv9 model to extract the ingredients present in the image. Alternatively, the user can also input ingredients manually. The user input also contains macro and micro nutrient levels, caloric values, cholesterol levels, taste and texture. Utilizing fine-tuned GPT2 LLM, the system then generates a unique recipe that follows the input. Web application developed by Next JS 14 with backend support from Flask, helps greatly in enhancing User experience. The system also utilizes Stable Diffusion to generate colorful recipe images for user reference of how the final recipe should look like.
Authors - Sakshi Prakash Masand, Sadhana Shashidhar, Animesh Giri Abstract - High-stakes industries like the aviation industry demand minimal downtime requiring unified solutions that address all maintenance and troubleshooting needs. Existing solutions often function in isolation, requiring multiple systems for predictive analytics and manual-based repairs. To bridge this gap, this article presents a novel integration of machine learning based predictive maintenance and conversational AI for routine servicing and troubleshooting operations in industrial IoT systems, tailored to critical sectors such as aviation. A robust predictive maintenance model is built to predict the Remaining Useful Life (RUL) for aircraft components. Multiple traditional machine learning models like Random Forest, Support Vector Regression (SVR), XGBoost, and deep learning techniques like Gated Recurrent Unit (GRU) and Long Short Term Memory (LSTM) are compared for performance and accuracy, ultimately focusing on a refined time series specific approach. When the predicted RUL of a component falls below a predefined threshold, operators are automatically alerted to schedule maintenance. For interactive support, Rasa, a customisable conversational AI framework and a fine tuned LLaMA model provide instant and enterprise specific guidance, offering step by step instructions and reducing reliance on lengthy manuals. This solution combines predictive maintenance with dynamic assistance, saving valuable time and resources for the industry.
Authors - Anushka Ashok Pote, Laukik Nitin Marathe, Suvarna Abhijit Patil, Sneha Kanawade, Deepali Samir Hajare, Varsha Pandagre, Arti Singh, Rasika Kachore Abstract - Due to the increase in vulnerability of different types of diseases, the use of artificial intelligence is seen to be rapidly increasing in the healthcare industry for creating systems that will provide diagnosis, treatment, and patient care. One of the major challenges that is faced in recent days is that most of the traditional healthcare systems are not transparent and comprehensible. This review explores the importance of Explainable Artificial Intelligence in order to make advancements in precision medicine, focusing on personalized treatment and disease prediction. Despite being powerful, traditional AI models function as "black boxes," which do not offer any insights into how decisions are made. This limits their application in critical sectors like healthcare where trust and accountability are crucial. Explainable AI makes systems more transparent and interpretable allowing healthcare professionals to understand and trust AI-driven insights. It exhibits significant enhancements in diagnostic accuracy and treatment personalization across various areas like oncology, cardiovascular disease, neurology, etc. The review performs comparisons between explainability driven models and traditional models. It reveals that XAI-based models offer better accuracy along with precision. It provides interpretable decision-making which makes them more suitable for clinical applications. Even though these systems exhibit certain challenges like computational complexity and need for standardized evaluation metrics. This paper highlights transformative potential of XAI in healthcare industry by fostering more ethical, transparent and patient-centered solutions. It is poised to revolutionize precision medicine by improving patient outcomes and exhibiting significant contributions in the healthcare industry.
Authors - Austine S Manuel, Alana Ance John, Hamna Rafeeq, Thalhah Anas, Manoj V. Thomas Abstract - Smart assistive technologies improve the independence and access to transportation of the vision-impaired people [1]. This paper presents AI-enhanced smart glasses designed to empower visually impaired by transforming their interaction with the environment. The system combines cutting-edge technologies to offer a comprehensive assistive solution. A path navigation algorithm, powered by YOLOv8, ensures safe mobility by detecting obstacles and guiding users with audio feedback. Integrated OCR capabilities enable real-time text reading, while an image captioning module provides detailed scene descriptions for enhanced situational awareness. The glasses incorporate a camera and an ultrasonic sensor, delivering robust performance in diverse scenarios, including detecting objects at close proximity where traditional algorithms fall short. With seamless audio output for user interaction, the proposed solution bridges the gap between technology and accessibility, promising independence and confidence for visually impaired individuals. This innovative fusion of AI and sensory inputs defines a new benchmark for assistive devices.
Authors - P C Gita, Arthi R, R Geetha, Syed Hayath Abstract - Employee layoffs have significant emotional and psychological impacts, often reflected in public discussions on social media platforms like Twitter. Employee layoffs not only affect the employee at stake but also impacts the brand image of the company. This study explores the sentiments and emotions surrounding layoffs through a comparative analysis using three natural language processing (NLP) tools: TextBlob, VADER, and the NRC Emotion Lexicon. A dataset of layoff-related tweets was collected over six months, pre-processed, and analyzed for sentiment polarity and emotional tone. The analysis revealed predominantly negative sentiments, with emotions like anger, sadness, and anticipation being prevalent. While TextBlob and VADER effectively gauged sentiment, VADER performed better in handling informal language, and the NRC Lexicon provided a more nuanced emotional profile. The study highlights the psychological toll of layoffs and the importance of employer transparency in mitigating anxiety. Future research should consider advanced NLP models like BERT for improved sentiment detection and track the evolution of layoff-related sentiments over time..
Authors - Pinakci Kathotia, Ridhima Rathore, Kush Gupta, Eshita Vijay, Ujwala Kshirsagar Abstract - The developing subject of contactless voltage sensing and its revolutionary effects on the electrical sector are explored here. The basics of contactless voltage detection are first described, along with how it differs from conventional techniques. It then goes into detail on the different uses for this technology, such as non-invasive electrical testing and its employment in high-voltage environments. This paper also discusses the advantages of contactless voltage detection, such as improved efficiency and safety, as well as the difficulties that must yet be overcome for this technology to reach its full potential. Examples from everyday life are utilized throughout the paper to highlight the useful uses for contactless voltage detection.
Authors - Rohith Aryan TG, Kashish Sharma, Sunil Kumar, Abhay Sharma Abstract - A Revolution in the Diagnosis and Treatment Landscape. This paper brings forward how AI technologies can be utilized in earlier detection of mental illnesses, and how robotics provide therapeutic support to patients. Examining the various models, results and challenges on these technologies is aimed at putting attention on their perceived benefits and limitations in augmenting the health care of the mind. Comparative studies that include traditional treatment modalities will also be featured, along with recommendations for future research directions.
Authors - Aarv Mankodi, Sanya Jain, Vedant Mundada, Dinesh Kumar Saini Abstract - Breast Cancer is a malignant tumor that occurs in the breast. It is a very serious threat to the health and well-being of women. Over the years the study of detection of this cancer using histopathology image recognition has become quite popular. Most of the methods, however, focus on creating new deep learning models or improving existing models like VGG or AlexNet. While these convolutional neural networks have been very successful in their implementation, it does not mean they do not have challenges, namely the problem of imbalance in datasets. It is for this reason that this paper instead tries to use graphs to solve the problem of breast cancer detection. For this paper, we use a graphical neural network that is trained to detect breast cancer in histopathology images. This is because the challenges faced by the deep learning models namely overcoming the imbalance in datasets may not be present in this graphical approach. It is also to find out if there are some previously unknown improvements in using a graph for detection instead of deep learning models.
Authors - Vishalsinh Bais, Mansi Pagdhune, Wamik Khan, Gaurav Maske, Aditya Umredkar, Amol P. Bhagat Abstract - The rise of deepfakes has raised questions about the veracity of digital content. This has led to a lot of research into reliable detection techniques. In this study, we introduce a new deepfakes detection approach based on the mesoNet architecture and the use of convolutional neural networks (CNNs). The proposed model has a multi-layer structure that includes convolutional layers, pooling layers, and dropout techniques to effectively extract and discriminate features. Training on a dataset that includes our own forged images and deepfakes, this model shows promising results in identifying manipulated content. With the activation function of leaky ReLU, our mesoNet model shows great promise in accurately distinguishing deepfakes from real images. Our experimental results demonstrate its effectiveness in distinguishing between forged and real images, demonstrating its value as a powerful tool in the fight against digitally manipulated content.
Authors - Vinayak Kanhegaonkar, Ayush Mahant, Ali Sayyed, Abha Shah, Aparna Junnarkar Abstract - India’s farmers battle many challenges before they are able to sell their produce to the consumer, one such challenge is related to middlemen and market access. The current scenario is such that middlemen hold significant influence over the market. Middlemen are the intermediaries that farmers sell their produce to because of reasons such as lack of infrastructure, price fluctuations, limited storage facilities, lack of information and more. This causes farmers to lose much of their profit and overall reduces their income. Middlemen take substantial profits from their work; they buy for less from farmers and sell for more to consumers. Through our research we aim to create an app, KrishiMitra, for farmers to facilitate direct access to the market. KrishiMitra will be able to reflect real-time market prices, provide multilingual support, and have a bidding system. With this app we hope to provide farmers a platform to get fair prices, some bargaining power, proper and complete information about market prices and current trends. In the future, the application can be integrated with more services that address other challenges faced by farmers such as limited storage space for produce and climate change impacts on produce.
Authors - Abhijeet G. Dhepe, Ashish R. Lahase, Shalini V. Sathe, Sagar Chitte, Pooja B. Abhang, Bharati W. Gawali, Sunil S. Nimbhore Abstract - This study introduces a novel crop diseases database design for Automatic Speech Recognition (ASR) systems. It aimed to predict crop diseases, addressing the critical role of speech technology in agriculture. The research bridges the gap between farmers and advanced diagnostic tools by creating a specialized voice corpus focused on agricultural terminology and disease names. The dataset, which includes various phrases related to crop health management, was collected in naturally noisy environments from farmers in the Marathwada region. Recordings captured the authentic speech patterns of both male and female participants, encompassing various dialects and accents. This approach ensures real-world applicability, enhancing the reliability and relevance of ASR systems. By including native and non-native speakers, the study promotes linguistic inclusivity. It aims to empower farmers with accessible, speech-based disease prediction tools, ultimately fostering greater efficiency and resilience in crop management.
Authors - Shruti Gandhi, Charmy Patel Abstract - This research presents an innovative framework for developing educational chatbots that redefine student support by integrating advanced natural language understanding (NLU), intent recognition, and pragmatic analysis. Leveraging machine learning techniques, including pre-trained models like BERT, the chatbot achieves state-of-the-art performance in recognizing user intents and delivering contextually relevant responses. By addressing the limitations of traditional systems, such as poor personalization and difficulty in handling nuanced queries, this framework enables dynamic, adaptive, and engaging interactions. The chatbot transcends conventional query handling through pragmatic analysis, allowing it to interpret subtle nuances, emotional states, and real-world contexts. This ensures personalized responses that align with individual learning needs, fostering deeper student engagement and comprehension. Finetuned with diverse datasets and instructional materials, the system is robust and scalable, making it suitable for a wide range of educational applications. This approach also emphasizes human-like interaction, combining emotional intelligence with context-aware capabilities to create a supportive learning environment. By enhancing response accuracy, adaptability, and user engagement, the chatbot sets a new benchmark in educational technology. Ultimately, this research demonstrates transformative potential in creating intelligent, scalable, and highly effective tools for modern education, paving the way for a more personalized and interactive learning experience.
Authors - Madhuri Wakode, Geetanjali Kale Abstract - Deep learning has many applications in healthcare especially for disease prediction from complex images. One such application is to predict diseases from chest X-rays (CXR). These models need huge amounts of data available for training. A single healthcare facility may struggle to collect sufficient data to build robust and efficient models. The apparent solution is collaboration among multiple healthcare facilities to use their data to build efficient models together. However, facilities may not want to share the patient’s sensitive data with other facilities or with the central server. Federated Learning (FL) allows multiple parties to build models without sharing their data with each other. FL allows parties to train the model locally on their private data and only share the trained model parameters to the server. Server averages the model parameters sent by all the parties to build a robust global model. Server sends this updated model to each party who then again trains the model locally. This process continues till the model convergence. We propose using federated learning average on a large CXR dataset for multi-label classification. Our results show that federated learning achieves the accuracy of ~82% as compared to ~91% of that of traditional centralized training method. FL with more robust algorithms and larger datasets, can achieve performance comparable to centralized approach with an added advantage of collaborative learning with privacy preservation.
Authors - Namrata Naikwade, Shafi Pathan Abstract - This paper presents a cost efficient and secure framework for data storage in cloud environments, It uses Discrete Cosine Transform (DCT) for data compression and Ciphertext-Policy Attribute-Based Encryption (CP-ABE) for fine-grained access control. The framework focuses on some of the main challenges in cloud storage, such as storage costs minimization, data privacy and security along with swift access revocation. Framework significantly reduces storage overhead by integrating DCT which compresses image data effectively while maintaining data quality. This storage optimization strategy enhances the cost-effectiveness of cloud storage making it suitable for large scale applications. CP-ABE provides a secure data access management by enforcing attribute based encryption which enables flexible and precise control over who can access the data ensuring privacy and security even in untrusted cloud environments. It also provides a rapid access revocation to safeguard data from risks associated with unauthorized access which is a critical risk in dynamic cloud environments. Experimental evaluations shows that the framework minimizes storage costs by up to 50% while significantly improving data security and efficiency demonstrating that proposed framework is a practical and scalable solution for secure and cost-efficient data management, addressing the needs of modern cloud-based systems.
Authors - Komal V. Papanwar Abstract - MorseMate is a user-friendly Morse code converter designed to simplify the process of converting Morse code to alphanumerical values by allowing input through three buttons for three separate functions involved in the transmission of the code. This approach eliminates the need for timing-based input, making Morse code more accessible and easier to use. The device, powered by an ESP8266 microcontroller and featuring a 128x64 Organic Light Emitting Diode (OLED) display, converts Morse code into readable text in real-time. Users can input their sequences with the press of a button, and a long press of the submit button clears the display, allowing for continuous use. Custom I2C configuration provides flexibility in hardware setup, while the compact design ensures portability. This paper elaborates on how the system combines simplicity, efficiency, and practicality, to make Morse code more accessible to a wider audience.
Authors - Hepin Gondaliya, Parth Monpara, Dhaval Shah Abstract - Multiplier is a crucial factor influencing processor performance, making its optimization vital for efficient computing. This paper introduces a multiplier design that achieves remarkable improvements, reducing resource utilization by more than half, logic delay to nearly a quarter and total power consumption to less than a quarter of compared to conventional radix-2 booth’s multiplied The proposed design was synthesized and implemented using Xilinx Vivado on a Zedboard FPGA, demonstrating its effectiveness and scalability for modern FPGA-based systems.. . .
Authors - Deep Dey, Hrushik Edher, Lalit M. Rao, Dinesh Kumar Saini Abstract - SSL is an important technique in machine learning which has high value for image classification tasks because acquiring labeled data is expensive. SSL trains on the labeled data and then we apply it predict the unlabeled dataset. Once it is trained, it can be further fine-tuned with the smaller labeled subsets for specific task. Recently, SimCLR, BYOL, and MoCo have shown impressive performances. SimCLR uses contrastive learning to train models by maximizing similarity between pairs classified under the same class and minimizing it between pairs belonging to different classes. BYOL, on the other hand, does not rely on negative pairs but implements a two-network architecture that eases training. MoCo develops a dynamic dictionary through which scaling SSL becomes efficient on massive data and performs well even with the smallest of batches. We can evaluate this models on CIFAR-10 and ImageNet, SSL approach is used in the areas where the labeled training data is scarce. In some research it is shown that SSL is competitive with SL when the amount of labeled training samples are small.
Authors - Tanuja Zende, Ramachandra. Pujeri, Suvarna Pawar Abstract - Human Emotion Recognition (HER) has gained considerable attention in recent years, driven by advances in machine learning, acoustic signal analysis and natural language processing. The advancements in HER using speech as a principal modality is addressed in this survey systematically. The importance of emotional intelligence in HCI, social robotics and mental health assessment in this review comprehends a complete analysis of approaches including feature extraction techniques, classification algorithms and data representation used in the field of speech emotion recognition. Furthermore, in this survey, existing methods into traditional rule-based systems, machine learning algorithms and state-of-the-art deep learning frameworks, stressing on the strengths and limitations are discussed. Additionally, thoughtful challenges such as the density of human emotions, the influence of contextual factors, and the need of annotated datasets to train robust emotion recognition systems find their involvement in this work. Current trends in multimodal emotion recognition (MER) and the incorporation of speech with other modalities are also discussed to provide a complete view. This amalgamation of existing literature aims to notify future research directions in emotion recognition systems, enhancing their pertinency across varied fields.
Authors - R. R. Bhoge, Ranjit R. Keole, Pravin P. Karde Abstract - In the rapidly evolving landscape of network security, Intrusion Detection Systems (IDS) play an indispensable role in distinguishing normal network traffic from anomalies. While traditional machine learning models have achieved notable success, the introduction of quantum-assisted techniques opens new avenues for improving accuracy and reliability. This review delves into the application of quantum supervised learning for Intrusion Detection, emphasizing recent breakthroughs and comparing the advantages of quantum methods to classical counterparts. As digital networks become increasingly interconnected, the demand for robust Intrusion Detection Systems has grown exponentially. This paper investigates quantum-enhanced cybersecurity solutions, particularly the incorporation of quantum supervised learning, to improve the detection and classification of network intrusions. The review discusses the theoretical principles of quantum supervised learning, its unique attributes, and the advancement of quantum-enabled IDS. This paper provides a detailed overview of how quantum computing is revolutionizing machine learning for cybersecurity, with an emphasis on the enhanced capabilities of IDS through quantum-assisted methodologies.
Authors - Prateeksha Chouksey, Tanvi Rainak, Piyush Shastri, Jayshree Karve, Ritesen Dhar Abstract - Identifying fishing spots is vital for sustainable fisheries management, conservation, and optimizing fishing activities. This paper provides a comprehensive review of Artificial Intelligence, Machine Learning, Big Data, and Data Analytics in enhancing fishing spot detection, surpassing the limitations of traditional methods such as fisher experience and historical catch data. These advanced approaches improve accuracy in locating productive fishing areas by integrating remote sensing, Geographic Information Systems, and real-time data from sonar and satellite imagery. The study emphasizes the role of environmental factors—such as water temperature, salinity, and ocean currents—in influencing fish distribution and explores the impact of technological advancements on eco-friendly fishing practices. Current challenges include data availability, environmental variability, and the need for interdisciplinary collaboration. Additionally, the paper outlines the transformative potential of these technologies to optimize resource utilization and preserve marine ecosystems, suggesting future research directions to address existing gaps. As these methods continue to evolve, their wider adoption is expected to support the sustainability of fisheries and environmental conservation significantly.
Authors - Akshat Vashisht, Ishika Tekade, Juhi Shah, Aniket Sawarn, Deependra Singh Yadav, Palash Sontakke, Rajkumar Patil Abstract - Segmenting grape leaf stress condition accurately is a critical step in precision agriculture, as it enables early detection and treatment to mitigate crop losses. In this research, we propose a novel approach leveraging the Segment Anything Model 2 (SAM-2) for precise segmentation of stress condition regions on grape leaves. SAM-2 is a foundation model for promptable visual segmentation in images and videos. This model can generate high-quality masks with minimal user input, which makes it an ideal tool for such tasks. The SAM-2 model was tested on field images and we achieved an accuracy of nearly 70% without fine-tuning. Experimental results demonstrate that SAM-2 outperforms traditional segmentation models like U2Net and V-net. Data augmentation can improve the performance of SAM-2, especially in challenging tasks such as detecting early-stage leaf spots or stress condition symptoms in overlapping leaves. Techniques such as rotation, adjusting brightness, and scaling, can simulate different conditions, balance the dataset and improve generalization. This helps SAM-2 to adapt to different scenarios and improve its ability to detect complex patterns. This shows the potential of SAM-2 in agricultural applications and provides a framework that can be integrated into advanced plant monitoring systems. By automating the segmentation process with minimal user intervention, SAM-2 significantly reduces the labour-intensive task of manual stress condition detection, thus saving time and resources in agricultural operations. Integrating SAM-2 into the grape leaf stress condition detection pipeline enhances the precision and reliability of stress condition identification systems. Furthermore, its adaptability across different segmentation tasks provides a foundation for scalable and automated plant health monitoring systems. This study establishes SAM-2 as a transformative tool for advancing sustainable farming practices and precision agriculture.
Authors - Prasanna Tupe, Parikshit N. Mahalle Abstract - This study presents a machine learning-based method for using the K-Nearest Neighbors (KNN) algorithm to optimize swing trading strategies. The algorithm forecasts stock price fluctuations over the next seven days using historical stock market data and offers various levels of nuanced trading signals. By empowering traders to make more accurate and knowledgeable judgments, this method outperforms conventional binary buy-and-sell recommendations. The KNN technique was selected due to its instance-based learning, non-parametric nature, and simplicity, which make it interpretable and computationally efficient. The model's accuracy rate demonstrated how well it could forecast changes in stock prices. This study provides a workable approach for swing traders looking to optimize their profits while highlighting the important role that machine learning plays in tackling the difficulties associated with stock market prediction. This study lays the groundwork for using machine learning to enhance trading tactics and financial market decision-making.
Authors - Arpita Mahakalkar, Karan Deshmukh, Kruthika Agarwal, Sudhanshu Maurya, Firdous Sadaf M. Ismail, Rachit Garg Abstract - Breast cancer remains a driving cause of mortality among ladies around the world, emphasizing the requirement for progressed location strategies. This considers creating a novel Computer-Aided Conclusion (CAD) framework leveraging MATLAB to progress the precision and proficiency of breast cancer location. The framework utilizes mammographic pictures and applies progressed measurable extraction procedures to analyze key characteristics, such as mass shape and edges. These highlights assist in classifying utilizing machine learning calculations, counting neural systems, and back vector machines. A one of a kind integration of wavelet change and multilayer perceptron strategies illustrated critical enhancements in recognizing Incendiary Breast Cancer (IBC) from non-IBC cases. The proposed approach beats conventional strategies, advertising improved symptomatic unwavering quality, decreased execution time, and tall exactness in cancer classification. This work underscores the potential of joining progressed machine learning procedures and picture-preparing apparatuses within the early and exact location of breast cancer, eventually supporting radiologists and decreasing demonstrative challenges.
Authors - Safwan Ahmed, Rojas Binny, S N Velukutty, Animesh Giri Abstract - Industrial Internet of Things (IIoT) networks have started to play a crucial role in revolutionizing manufacturing and Industry 4.0. However, the distributed nature of IIoT and its relative infancy make it a prime target for cyberattacks. This paper proposes a new approach to address the threats faced by industries using a conversational Artificial Intelligence (AI)-interfaced deep neural network model for detecting attacks on an IIoT network. The proposed approach extends to threat mitigation and has been evaluated using an extensive IIoT network traffic data set, demonstrating 93% accuracy in detecting 14 of the most common threats plaguing the industry. This model introduces the integration of conversational AI with deep learning, offering a user-friendly interface for naive users and accurate threat detection. The broader impact of this work lies in its potential to significantly enhance access to a robust and accurate real-time cyber-threat detection and mitigation system, thus contributing to a more secure and resilient industrial landscape.
Authors - Atharva Madhukar Nimbalkar, Madhukar Nimbalkar, Madhura Gondhalekar, Yashvendra Singh Dhandal, Parikshit Mahalle, Pankaj Chandre Abstract - The digital divide continues to pose significant challenges to delivering quality education, particularly in remote and underserved regions with limited or no internet connectivity. This paper explores the innovative integration of three advanced technological solutions within the Spacelink: Educational System, designed to bridge this gap. The system incorporates an AI-driven transcription and translation model using Whisper, a web-based Learning Management System (LMS), and a real-time communication platform utilizing WebRTC. These tools are tailored for offline functionality, operating effectively through LAN-based communication to ensure uninterrupted educational services. The AI transcription system enhances accessibility by converting spoken content into text and supporting multiple languages, addressing linguistic diversity and inclusivity. The LMS provides a robust offline platform for content delivery, course management, and student assessment, synchronizing data seamlessly when connectivity is restored. Meanwhile, the WebRTC-based communication tool facilitates real-time audio and video interactions optimized for low-bandwidth environments, promoting interactive learning experiences. Together, these technologies offer a holistic solution to overcome educational barriers, enabling equitable access to high-quality learning resources and fostering collaborative education in disconnected settings.
