VOLUME 13, ISSUE 4, APRIL 2026

Abstract: This paper investigates the existence of odd-even congruence labelling for various classes of graphs, including a splitting graph of bistars, corona products, tensor products shadow graphs and tadpole graph. A graph G(V,E) is defined as an odd-even congruence graph if there exists a bijective mapping f:V(G)→{1,3,5,……,2|V|-1} and an injective mapping f^*:E(G)→{2,4,6,……,2|E|} such that for every edge uv∈E(G), the condition f^* (uv)| |f(u)-f(v) is satisfied. In addition, constructive proofs and labelling algorithms to demonstrate that these specific graph structures, arising from products and splitting operations, admit such a labelling scheme are provided.

Keywords: Graph labelling, odd-even congruence, splitting graph of bistar graph, corona product, tensor product, tadpole graph and shadow graph.

Abstract: The present study focuses on analysing the reasons for job change among employees despite receiving sustainable earnings, with special emphasis on career growth, learning opportunities and job satisfaction. In the current competitive work environment, employees increasingly prioritize professional development and job satisfaction over monetary benefits. The study is based on primary data collected from 150 respondents using a structured questionnaire, and statistical tools such as percentage analysis, correlation analysis, chi-square test and standard deviation have been used for analysis. The findings reveal that career progression, skill development and job satisfaction are the major factors influencing job change decisions, and a negative relationship exists between career growth, job satisfaction and job change intention. The study highlights that organizations must focus on providing growth opportunities, maintaining a positive work environment and improving employee satisfaction to reduce turnover and enhance long-term organizational performance.

Keywords: Job Change, Career Growth, Learning Opportunities, Job Satisfaction, Employee Retention Work Environment, Correlation Analysis

Abstract: This study investigates the impact of social media influencer credibility and authenticity on brand trust and purchase intention among active social media users in India. With the rapid growth of influencer marketing as a mainstream digital strategy, understanding how influencer-related attributes shape consumer behaviour has become increasingly important for both researchers and practitioners. A structured questionnaire-based survey was administered to 152 respondents, and the data were analysed using Pearson correlation analysis and multiple linear regression. The findings reveal that both influencer credibility and authenticity exhibit significant positive relationships with brand trust. Brand trust, in turn, emerges as the strongest predictor of purchase intention, underscoring its central role as a mediating mechanism in the influencer-consumer relationship. While influencer credibility shows a direct positive effect on purchase intention, authenticity appears to operate indirectly through brand trust rather than independently influencing purchase decisions. These results highlight the importance of selecting credible and authentic influencers to build lasting brand trust and drive consumer purchase intention in the digital marketplace.

Keywords: Influencer Credibility, Influencer Authenticity, Brand Trust, Purchase Intention, Social Media Marketing.

Abstract: This project focuses on the design and development of a six-degree-of-freedom (6DOF) robotic arm capable of performing versatile and precise automation tasks. The robotic arm is designed to mimic human-like movements using six independent joints, enabling operations such as pick-and-place, assembly, and material handling. Servo motors are used for accurate positioning, while an ESP32 microcontroller manages control, communication, and system coordination. The system supports both manual operation through a Bluetooth-based mobile application and autonomous control via a web interface. An ESP32-CAM module is integrated to provide real-time video streaming for remote monitoring. The design emphasizes affordability, modularity, and scalability using open-source components, making it suitable for academic and industrial use. Safety features such as torque control, emergency stop mechanisms, and secure communication protocols are also incorporated. Overall, the robotic arm finds applications in manufacturing, healthcare, hazardous environments, and research, bridging the gap between theoretical learning and practical implementation.

Keywords: 6DOF Robotic Arm, Automation, Pick-and-Place Operations, Servo Motors, ESP32 Microcontroller, ESP32-CAM, IoT-BasedControl, WirelessCommunication, BluetoothControl,WebInterface

Abstract: The increasing demand for enhanced vehicular safety has driven the development of intelligent braking systems capable of minimizing human error and preventing accidents. This paper presents the design and development of an advanced automatic braking system utilizing multiple ultrasonic sensors for real-time obstacle detection and embedded control. The proposed system employs a multi-sensor configuration to improve detection accuracy and coverage by monitoring obstacles in multiple directions. An embedded microcontroller-based architecture is used to continuously process sensor data and evaluate the distance between the vehicle and surrounding objects. Based on predefined threshold conditions, the system performs dynamic decision-making to initiate appropriate actions, including warning generation, speed reduction, and automatic braking. A motor driver interface is implemented to regulate the motion of the vehicle, ensuring smooth and controlled braking performance. Additionally, a wireless communication module is integrated to enable remote monitoring and user interaction through a mobile device. The developed prototype demonstrates reliable performance in detecting obstacles and responding with minimal delay, thereby reducing the risk of collision. The system offers a cost-effective, efficient, and scalable solution for implementation in autonomous vehicles, robotic systems, and intelligent transportation applications. The results validate the effectiveness of the proposed approach in enhancing safety through real-time sensing and automated control mechanisms.

Keywords: Automatic Braking System, Ultrasonic Sensors, Arduino Nano, Obstacle Detection, Embedded Systems, Collision Avoidance

Abstract: Inspection of pipelines and confined industrial environments is a challenging and critical task, especially in areas that are hazardous or inaccessible to humans. This project presents the design and development of a four-legged spider robot integrated with a camera system, controlled using an Arduino microcontroller, for efficient detection of damages and defects in pipelines and constrained spaces. The robot is designed with a quadruped locomotion mechanism, enabling it to navigate uneven surfaces, narrow passages, and complex terrains where traditional wheeled robots may fail. Each leg is actuated using servo motors, providing stability and flexibility for movement in confined environments. An onboard camera is mounted on the robot to capture real-time video and images, which are transmitted wirelessly to the operator for monitoring and analysis. The Arduino-based control system coordinates locomotion and camera operation, ensuring precise movement and data acquisition. The robot can be remotely operated, allowing users to inspect pipelines for cracks, corrosion, blockages, and structural defects without direct human intervention. This reduces risk, improves efficiency, and lowers maintenance costs.

Keywords: • Pipeline Inspection Robot • Four-Legged Spider Robot • Quadruped Locomotion • Confined Space Exploration • Industrial Inspection System • Arduino-Based Control • Servo Motor Actuation • Wireless Monitoring System • Real-Time Video Transmission

Abstract: Advanced finishing processes are essential in modern industry to achieve the surface integrity required for high-strength materials used in aerospace and medical applications. This project presents the design and fabrication of an Electro Magnetic Abrasive Finishing (EMAF) machine, a non-conventional finishing system capable of achieving "mirror-like" surface quality. The system utilizes a suspension tank where an electromagnetic field drives abrasive particles, across the workpiece to precisely remove material from complex geometries. To ensure precise control, the machine integrates an Arduino-based microcontroller platform to manage system operations and motor outputs. The performance of the EMAF process was evaluated by analyzing the influence of key parameters, including rotating speed, feed rate, depth of cut, and the number of passes, on the surface roughness of aluminum and wood workpieces. Experimental results demonstrate that systematic calibration of these parameters significantly improves fatigue strength and wear resistance by eliminating surface irregularities. This project aims to provide a high-precision, cost-effective, and scalable finishing solution for industries where traditional grinding methods fail to meet stringent design requirements. Hassan El-Hofny et al [1]focused on the material removal rate (MRR) and the physics of the abrasive-workpiece interface, establishing the "Flexible Magnetic Abrasive Brush" (FMAB) as a multipoint cutting tool P.C. Pandey and H.S. Shah et al [2] categorized EMAF as a vital super-finishing process. Recent reviews confirm its necessity for medical implants and aerospace components where "mirror-like" finishes are required to improve fatigue life.

Keywords: Abrasive Machining, Electromagnetic Machining, Surface roughness, non-conventional finishing process.

Abstract: Fire accidents pose a serious threat to human life, property, and the environment, especially in hazardous or inaccessible areas where manual firefighting becomes difficult and risky. To address this issue, an intelligent fire-fighting robotic system has been developed using an Arduino Mega platform. This project focuses on designing and implementing a multi-functional robot capable of detecting, navigating toward, and extinguishing fire autonomously, while also allowing manual control through Bluetooth communication. The robot is equipped with multiple flame sensors strategically placed around the system to detect fire from different directions. An ultrasonic sensor is integrated to measure distance and avoid obstacles, ensuring safe navigation. Additionally, environmental monitoring is achieved using temperature and smoke sensors, enabling the robot to respond effectively to critical conditions. When fire or abnormal temperature is detected, the system activates a water pump mechanism through a relay and directs a servo-controlled nozzle to spray water over the affected area. The robot operates in two modes: automatic and manual. In automatic mode, it independently detects fire, navigates toward it, and extinguishes it without human intervention. In manual mode, the robot can be controlled remotely via Bluetooth, allowing the user to maneuver the robot and control the firefighting mechanism. A siren system is also incorporated to provide an audible alert during emergency conditions. The motor driving system is implemented using dual L298N motor drivers, enabling precise movement control such as forward, backward, left, and right navigation. The integration of sensors, actuators, and control logic makes the system efficient, responsive, and reliable in fire detection and suppression tasks. This project demonstrates a cost-effective and efficient solution for fire safety applications, particularly in industrial environments, warehouses, and areas that are dangerous for human intervention. The developed system highlights the potential of robotics and embedded systems in enhancing safety measures and reducing the risks associated with firefighting operations.

Keywords: Fire-fighting robot, Arduino Mega, Autonomous navigation, Flame sensors, Ultrasonic sensor, Water pump system, Servo-controlled nozzle, Bluetooth control, Manual and automatic modes, L298N motor driver.

Abstract: Alzheimer's disease is an age-related neurological disorder that affects the brain, sequentially depriving people of their memory, thinking skills, and capabilities to handle everyday tasks. Early-stage detection is necessary for impactful clinical management and for prolonging disease progression. This study offers a new method using deep learning technology to recognize Alzheimer's in its early stages by evaluating structural brain scans taken with Magnetic Resonance Imaging (MRI). The offered system utilizes a 3D Convolution Neural Network (3D- CNN) to explicitly process MRI datasets and instantly extract meaningful morphological features connected with early cognitive dysfunction. To increase diagnostic validity, MRI data preparation techniques such as noise filtering and image standardization are applied before model training. Through in-depth validation, the model confirmed it can accurately tell the variation between healthy people and those in the beginning phase of Alzheimer's. This self-operating system can operate as a decision support system for clinicians, helping in early detection and supporting quick response approaches for Alzheimer's disease

Keywords: Neurodegenerative Disorders, Cerebral Magnetic Resonance Imaging (MRI), Artificial Neural Networks, Three- Dimensional Convolutional Neural Networks, Computational Medical Imaging, Early detection of Cognitive Decline.

Abstract: Agricultural production greatly depends on crop yield, production and quality of the crops. For increasing the overall Agricultural Production we have to detect and identify the crop diseases early and manage the diseases using specific types of techniques which are effective and efficient for the development of crop health. If we can't identify the crop diseases at the early stage it effect the income of the crop production. By using the traditional methods we can detect the plant and leaf diseases but it is a time consuming process and not that much efficient. And the traditional methods are suitable for only the small areas of crops, but not suitable for larger area of crops. Traditional methods provide very low accuracy. In this project, we can easily detect the disease at earlier by using Artificial Intelligence especially Deep Learning. The system uses artificial intelligence model to grasp the leaf patterns and checks whether the plant is healthy or not. It classifies the image and detects the disease at the earlier and provides the solution for the disease. The developed model distinguishes the plant leaves that are healthy or not. Comparing to the traditional methods this modern methods like deep learning will pro ide faster and accurate results. And the modern methods are cost effective and reduce the dependent on experts, integration with mobile or web applications. Overall by utilizing these deep learning techniques it helps the farmers to take action immediately, and increases the crop yield and quality and reduces the crop loss and provides sustainable agriculture.

Abstract: This project focuses on the design and fabrication of an automated pneumatic hammering machine using nylon fiber to enhance performance, efficiency, and durability. The system operates on the principle of compressed air, where an air compressor supplies high-pressure air that is controlled by a solenoid valve and directed into a pneumatic cylinder. The reciprocating motion of the cylinder piston is converted into a continuous hammering action, suitable for various industrial operations such as metal forming, riveting, forging, and assembly work. A control unit, such as a timer or PLC, is incorporated to automate the hammering process by precisely controlling the stroke frequency and operation time. This ensures consistent output, reduces human error, and improves overall productivity. The integration of nylon fiber plays a crucial role in this system by acting as a shock absorber, minimizing vibrations, reducing noise levels, and preventing mechanical wear and tear of components. The machine is designed with a metal frame structure to withstand repeated impacts and ensure stability during operation. The use of PU pipes enables efficient air transmission with minimal leakage. This automated system significantly reduces manual labour, increases operational speed, and enhances safety compared to traditional hammering methods. Experimental results show that the machine delivers uniform hammering force with improved efficiency and reduced energy loss. The project demonstrates a reliable, cost-effective, and low maintenance solution for industrial automation. Furthermore, it has the potential for future enhancements such as sensor-based feedback systems, IoT integration, and advanced control mechanisms to further optimize performance and energy utilization.

Keywords: Pneumatic Hammering Machine, Industrial Automation, Pneumatic Cylinder, Solenoid Valve Control, Nylon Fiber Damping, Vibration and Noise Reduction, Energy Efficiency, PLC-Based Control System

Abstract: Detection and classification of skin cancer using dermoscopic images plays an important role in early diagnosis and proper treatment planning. Examining these skin images manually takes a lot of time and it mainly depends on the experience of the doctor. This drawback increases the importance of automated detection systems. This project presents a deep learning framework with the help of transfer learning techniques to detect skin cancer accurately and in a precise manner. The system learns important features from dermoscopic images and improves the overall performance. To improve the capability of the system, a Vision Transformer model is used which analyzes the complete image instead of small regions. The model is able to learn features such as color variation, irregular shapes, uneven texture, and unclear boundaries of skin lesions. Before giving the images to the model, they are prepared in a simple way so that the system can learn better. First, the skin images are made ready so that the model can understand them properly. We resize some images and add a few more so the system can understand the data properly. The model was developed using Python with PyTorch. After the training part is finished, the system is given a skin image and it tries to identify what type it is. It checks the image and gives the result as either normal or cancer. When we tested the system, it worked fine in most situations and did not take much time to give the output.

Keywords: Dermoscopy Images, Skin Cancer, Melanoma Detection, Transfer Learning Model, Vision Transformer, Pytorch.

Abstract: Despite advancements in vehicle safety design, road accidents remain unavoidable and road accidents are still happening in both rural and urban areas because the increasing number of vehicles and multiple contributing factors. While rash driving may cause life and death situations, factors such as weather conditions, road type, traffic density, and seasonal changes also significantly influence accident severity. This project mainly applies machine learning techniques to predict the extent of danger resulting from road accidents by analyzing these influencing parameters. Random Forest and XGBoost models are used due to their effectiveness in handling complex and large- scale accident data and it also gives exact accuracy results of the road accidents. The proposed system helps us to identify high-risk conditions and supports traffic authorities and gives emergency response units in taking proactive safety measures. Overall, this work presents the practical approach to enhancing the road accidents through machine learning-based accident severity prediction.

Keywords: Machine Learning, Road Accident Severity Prediction, Random Forest, XGBoost, Predictive Risk Analysis.

Abstract: Energy consumption forecasting is necessary for the planning process and predicting electric consumption. It plays a key role in building an efficient energy system for power generation, distribution, and sustainable consumption. An accurate forecast of electric load is essential for a power system to be planned properly, generation to be scheduled, and electricity to be delivered economically. The complexity and non-stationary of the electricity load caused by industrial growth, urbanization, and changes in lifestyle in households make applying the classical statistical and rule-based forecasting methods difficult. Customary methods are not able to capture the nonlinear fluctuation in the load, long-range dependencies in time and the specific patterns in the various sectors, leading to wastage and uncertainty in operations. To this end, the present work proposes a time- series based energy consumption forecasting system that utilizes historical load data to forecast the future electricity demand for industry and household sectors. The system follows time- series modeling techniques to analyze historical consumption data and seasonal trends of load variations. The model can also be used to forecast consumption for each sector because the consumption of industrial and residential sectors varies both in demand and in terms of frequency. Long Short-Term Memory (LSTM), a deep learning model, can be implemented to capture the time series dependencies. From the experimental results, the proposed model has been proven to perform customary statistical forecasting approaches in terms of prediction accuracy and stability. The proposed model is able to recognize peak demand periods and provide reliable demand forecasts for load management on time. The optimal prediction of demand plays an essential role in effective power generation and consumption management for power utilities and policymakers to ensure sustainable operation.

Keywords: Energy Forecasting, Time-Series Analysis, ARIMA, LSTM, Load Prediction, Power Systems.

Abstract: In the digital age, the proliferation of counterfeit goods has led to an increasing need for reliable methods to detect fake logos, which often signify counterfeit products. To address this challenge, this project attempts to develop a robust Fake Logo Detection System which exploits advanced machine learning. A convolutional neural network (CNN) is used to analyze and categorize logo pictures and differentiate genuine logos versus fraudulent ones with high accuracy. The approach involves collecting a diverse dataset of authentic and fake logos, preprocessing the images to enhance quality and consistency, and training the CNN model on these datasets. Key steps include data augmentation to improve model generalization, feature extraction to identify distinguishing characteristics of logos, and fine-tuning the network to optimize performance. The system's effectiveness is evaluated through rigorous testing and validation, ensuring it can handle various logo designs and counterfeiting techniques. The ultimate goal is to provide a scalable and efficient solution for businesses and consumers to verify logo authenticity, thereby reducing the impact of counterfeiting and protecting brand integrity. By integrating deep convolutional classification models into brand protection systems, organizations can significantly improve counterfeit detection accuracy. This approach not only saves time and cost but also strengthens intellectual property protection. The solution is adaptable and can be extended to support multiple brands, making it suitable for real-world deployment in e-commerce, supply chain inspection, and digital content monitoring.

Keywords: Fake Logo Detection, Brand Forgery, Counterfeit Products, Deep Learning, Convolutional Neural Networks (CNN), Image Classification, Computer Vision, Logo Recognition, Brand Authentication, Pattern Recognition

Abstract: Solar energy is becoming the most used renewable energy in the world. Because of its environmental benefits and sustainability. Solar energy gives the most radiation due to recent weather conditions. We should overcome this challenging task. This project aims to predict the solar radiation intensity of the future by researching past historical weather conditions and time series data using machine learning techniques. The project proposed system uses the parameters like temperature, humidity, wind speed, atmospheric pressure and cloud as input features for the prediction models. Machine learning algorithms including long short term memory and support vector regression are applied to the complex and non-linear relationships between weather conditions and no solar radiance. Time series analysis tools are employed to capture only the seasonal trends and temporal dependencies present in the data. Accurate prediction of solar energy is essential for proper planning and working of revenue energy systems. This system project focuses on scanning solar radiation intensity using time series data and machine learning techniques. Time series forecasting models like auto regressive integrated moving average (ARIMA) are implemented to capture the seasonal variations present in the data. The proposed system predicted solar radiation values can be used to improve energy generation, solar panel deployment and support smart grid operations. By providing the accurate solar irradiance forecast that the system helps to reduce uncertainty in solar power generation. This system enhances the reliability of renewable energy systems. This system shows the result of machine learning based traditional prediction and demonstrating their effectiveness in solar radiation forecasting and renewable energy management.

Keywords: Solar Radiation Prediction, Solar irradiance forecasting, ARIMA, Time series analysis.

Abstract: Air pollution has become a major concern due to its harmful impact on human health and the surrounding environment. Most existing air quality monitoring systems focus on city-level Ambient Air Quality Index (AQI) values, which often fail to reflect pollution differences within smaller regions of an urban area. As a result, sudden pollution events and localized emission sources may remain undetected. This project presents a Context-Based Small Area Air Quality Index Prediction and Air Pollution Event Detection System (CM-SAAQIDS) designed to overcome these limitations. The proposed system uses historical air quality data along with environmental factors such as temperature, humidity, and wind patterns to estimate AQI values for individual micro-zones. It also identifies real-time pollution spikes and analyses possible causes, including traffic density, industrial activity, and unfavourable weather conditions. To ensure reliable results, the system incorporates methods to manage missing or inconsistent sensor data. By offering localized air quality forecasts and early warning alerts, the proposed approach supports timely decision-making by citizens and authorities, contributing to improved air pollution control and public health management.

Keywords: Cognitive Behavioral Therapy, Emotion Recognition, Cognitive Distortions, Mental‑Health Chatbot, Deep Learning, Natural Language Processing (NLP).

Abstract: The present study was conducted to examine parental expectations and academic achievement in relation to the education of children with special needs. Parental expectations refer to parents' aspirations regarding their children's educational outcomes, while academic achievement indicates the extent to which students attain their learning goals. Children with special needs require additional support for their overall development. The study aimed to examine the parental expectations of male and female parents and to assess the academic achievement of male and female children with special needs. The descriptive survey method was employed. The population comprised children with special needs in the Malda District of West Bengal. A sample of thirty children, including physically and visually impaired students, was selected using purposive sampling. Data were collected through interview schedules prepared and standardized by the researchers. The data were analyzed to examine differences in parental expectations and academic achievement based on gender. The findings revealed no significant difference in parental expectations between male and female parents and no significant difference in academic achievement between male and female children with special needs. Overall, the study indicates the absence of gender-based disparities in both variables.

Keywords: Parental Expectation, Academic Achievement, Children with Special Needs.

Abstract: Public transport systems in small cities face problems due to the unavailability of real-time tracking and communicating systems. This paper proposes a real-time public transport tracking system that utilizes GPS technology and a cloud backend for real-time bus tracking and alerting. The proposed system is implemented as a cross-platform mobile application for passengers, administrators, and drivers. The proposed system is efficient in terms of time and cost. The proposed system is implemented and tested, and the results show that it is accurate and efficient in terms of time. The proposed system is cost-effective for small cities.

Keywords: Real Time Tracking, GPS, Public Transport, Flutter, Firebase, Smart Transportation, Mobile Application

Keywords: • Compressed Air Conveyor • Air Jet Propulsion • Material Handling System • Energy Efficient Design • Beltless Conveyor • Fluid Dynamics • Laminar Flow • Industrial Automation • Low Maintenance System • Acrylic Platform Conveyor • Sustainable Engineering • Smart Material Transport

Abstract: The rapid growth of digital technology and social media platforms has significantly transformed the tourism industry and the way travelers search for information and make travel decisions. Social media influencers have emerged as important sources of travel-related information by sharing experiences, destination reviews, travel tips, and visual content that inspire potential tourists. This study examines the role of social media influencers in shaping tourists' travel decisions, particularly focusing on the level of trust tourists place in influencer recommendations and the influence of influencer content on travel destination choices. The study adopts a descriptive research design and uses primary data collected through a structured questionnaire distributed through Google Forms. A total of 76 valid responses were obtained using a convenience sampling method. The collected data were analyzed using frequency and percentage analysis to understand the demographic profile of respondents, while Chi-square analysis was applied to examine the relationship between variables. The findings reveal that social media influencers significantly influence tourists' travel destination choices. The results also indicate that there is a significant relationship between the frequency of following travel influencers and the likelihood of selecting destinations based on their recommendations. Furthermore, the analysis shows that individuals who frequently follow travel influencers tend to exhibit higher levels of trust in their travel recommendations. Overall, the study highlights that social media influencers play a crucial role in shaping tourists' perceptions, building trust, and influencing travel planning behavior. The findings suggest that tourism marketers and destination managers can effectively use influencer marketing strategies to promote destinations and attract potential travelers.

Keywords: Social Media Influencers, Travel Decision Making, Tourist Behavior, Influencer Marketing, Travel Destination Choice, Social Media Tourism.

Abstract: This project presents the design and development of an intelligent IoT-based smart wheelchair system that enhances mobility, safety, and communication for physically challenged individuals. The system integrates advanced technologies such as Wi-Fi-enabled control using ESP8266, real-time GPS tracking, emergency alert functionality, and multi-mode operation including joystick and voice control. A key feature of the system is its assistive communication capability. Dedicated push buttons are provided for the user, which when pressed, generate predefined voice outputs such as "food", "water", "washroom", etc. This feature is especially useful for users who are unable to speak or communicate easily, allowing them to convey their basic needs effectively. The wheelchair is equipped with a GPS module that continuously tracks its location and transmits coordinates to the Blynk IoT platform. In emergency situations, the user can trigger an alert that sends a notification along with a live Google Maps location link to caregivers. The system also allows remote monitoring and locating of the wheelchair through a mobile application. Additionally, the wheelchair supports dual-mode control: manual operation using a joystick and voice-based control via serial commands. A buzzer is included for alert and horn functions to improve safety. Overall, this project provides a cost-effective, reliable, and user-friendly assistive solution by combining mobility, safety, communication, and IoT-based monitoring. It significantly improves the independence and quality of life for differently-abled individual

Keywords: • IoT-based smart wheelchair • Assistive technology • Physically challenged individuals • Wi-Fi-enabled control (ESP8266) • Real-time GPS tracking • Emergency alert system • Voice control • Joystick control • Multi-mode operation • Communication aid • Remote monitoring • Blynk IoT platform

Abstract: The present study aimed to investigate the effect of weight training on static balance among handball players. A total of 100 male handball players aged between 14-19 years were selected and randomly divided into two groups: experimental (n=50) and control (n=50). The experimental group underwent a structured weight training program for 12 weeks, while the control group continued their regular routine without any specialized training. Static balance was assessed using the BASS Stick Test before and after the training period. The results revealed that the experimental group showed improvement in balance performance, while the control group demonstrated minimal or no change between pre and post-test scores. However, statistical analysis using Analysis of Covariance (ANCOVA) indicated that the difference between the groups was not significant at the 0.05 level. Although the findings were not statistically significant, practical improvement observed in the experimental group suggests that weight training may contribute positively to balance development. Balance is a key component in handball, influencing stability, coordination, and overall performance during dynamic movements. The study concludes that weight training alone may not be sufficient to significantly enhance balance among handball players and should be combined with balance-specific exercises. Future studies are recommended with longer duration and larger samples to establish stronger evidence.

Keywords: Weight Training, Static Balance, Handball Players, Physical Fitness, BASS Stick Test

Abstract: Elderly women residing in urban slums represent a highly vulnerable population due to the combined effects of aging, poverty, and limited access to healthcare. This systematic review and meta-analysis aims to synthesize existing evidence on the prevalence of addiction (primarily tobacco and alcohol use) and hypertension among elderly women in slum settings. A comprehensive literature search was conducted using databases such as PubMed, Scopus, and Google Scholar for studies published between 2000 and 2024. A total of 18-25 eligible studies were included, encompassing over 10,000 elderly women aged 60 years and above. The pooled prevalence of hypertension ranged between 40% and 60% across studies, with some slum-based studies reporting rates as high as 57.5% among women. Awareness and control of hypertension were found to be low, with less than one-third of affected individuals aware of their condition. Addiction-related behaviors, particularly tobacco use (smoking and smokeless), were prevalent due to stress, social isolation, and low health literacy. Mental health conditions such as depression (31-48%) were strongly associated with substance use and poor cardiovascular outcomes. The findings indicate a significant dual burden of addiction and hypertension among elderly slum-dwelling women, driven by socio-economic deprivation and limited healthcare access. Targeted public health interventions focusing on early screening, addiction control, and gender-sensitive healthcare delivery are urgently needed to improve health outcomes in this population.

Keywords: Elderly Women, Urban Slums, Hypertension, Addiction, Tobacco Use, Alcohol Use, Meta-analysis, Public Health, India

Abstract: The management and analysis of student academic results in diploma education systems often rely on manually processing gazette PDF documents, which is time-consuming and susceptible to human error. This paper presents the design and implementation of an automated system for extracting, structuring, and analyzing student results specifically based on MSBTE (Maharashtra State Board of Technical Education) standard formats and academic requirements. The proposed system efficiently processes structured PDF gazette files, extracts relevant student information, and converts it into a well-organized format suitable for analysis. The system utilizes PDF parsing techniques through PyMuPDF to accurately extract data from MSBTE-formatted documents, followed by data structuring and processing using Python libraries such as Pandas and OpenPyXL. It performs comprehensive result analysis, including subject-wise performance evaluation, overall result computation, and structured report generation in Excel format. By adhering to MSBTE standards, the system ensures compatibility, consistency, and reliability in handling academic records. Unlike existing approaches that rely heavily on OCR and machine learning techniques for unstructured data processing , the proposed system focuses on a domain-specific, rule-based approach optimized for structured academic documents. Experimental results demonstrate improved efficiency, reduced manual effort, and high accuracy in data extraction and analysis. The system provides a scalable solution for academic institutions and can be extended in the future to support unstructured formats using advanced techniques such as OCR and natural language processing.

Keywords: Student Performance Analysis, Document Parsing, PDF Data Extraction, Educational Data Mining, Automated Academic Systems, Data Structuring, Result Analytics, Python-Based System, MSBTE Gazette Processing, Information Extraction.

Abstract: The Smart Helmet System is an embedded and IoT-based safety solution designed to enhance two-wheeler rider protection and accident prevention. The system integrates multiple sensors and communication modules with a microcontroller-based architecture to ensure real-time monitoring and intelligent decision-making. The core of the system is a microcontroller (such as Arduino Uno), which interfaces with various modules including an alcohol sensor (MQ-3), a GPS module for location tracking, and for wireless communication. A helmet-wearing detection mechanism (using an IR sensor or pressure switch) ensures that the ignition system is enabled only when the rider wears the helmet. The alcohol detection subsystem continuously monitors the rider's breath alcohol concentration. If the measured value exceeds a predefined threshold, the control algorithm disables the ignition system, thereby preventing the vehicle from starting. The GPS module provides real-time geolocation data (latitude and longitude), which is transmitted for tracking and emergency response. The Blynk app displays parameters such as helmet status, alcohol detection level, and live GPS location (latitude and longitude). Additionally, a solar energy harvesting unit is incorporated to supplement power requirements, improving energy efficiency and system sustainability. The system operates through a rule-based control algorithm implemented in Embedded C, ensuring low latency and reliable performance. By combining sensor fusion, wireless communication, and automation, the Smart Helmet system offers a robust and scalable approach to improving road safety and reducing accident risks.

Keywords: • Smart helmet system, • Internet of Things (IoT), • embedded systems, • Arduino Uno, • MQ-3 alcohol sensor, • GPS tracking, • ignition interlock, • rider safety, • solar energy harvesting.

Abstract: Potholes are a major form of distress in flexible pavements, affecting road safety, ride quality, and maintenance costs. Due to weak bonding, traffic loading, and moisture damage, traditional pothole repair techniques frequently fail. The use of bio-based additives to improve the sustainability and performance of asphalt binder has been investigated recently [1]. It has been reported that lignin, a naturally occurring biopolymer derived from biomass, enhances the stiffness and high-temperature stability of asphalt binders [2]. Additionally, sugarcane bagasse ash has been studied as a filler that can improve the durability and strength of asphalt mixtures [3]. By assessing binder characteristics like penetration, softening point, ductility, and viscosity, this study investigates the possible application of bitumen modified with lignin and bagasse ash for pothole repair.

Keywords: Pothole Repair; Modified Bitumen; Lignin; Sugarcane Bagasse Ash; Sustainable Pavement.

Abstract: Concrete is one of the most extensively used construction material all over the world. Many scientists and researchers are in quest for developing alternate construction material that are environment friendly and contribute towards sustainable development. Huge amount of fly ash is generating day by day which creates the disposal problem and has many environmental issues Also, building are accidently subjected to fire hazards,during fire the temperature of concrete may go high. Elevated temperature have the potential to induce the formation of cracks in concrete .Similar to cracks in any other material , these cracks can propagate and , over time , result in the compromise of structural integrity , leading to a shortened service life. In this investigation,The effect of open fire on mechanical strength of fly ash added concrete is studied. The fly ash added concrete is prepared by replacing the cement in concrete by 10% of the cement. The concrete cube specimens of standard size are casted and tested after 28 days of curing. Each concrete cube specimen of normal and fly ash added concrete is exposed to open fire at elevated temperature of (100°C , 200°C, 300°C, 400°C, 500°C, 600°C, 700°C). After temperature regimes the compressive strength is determined for each specimen.

Abstract: Phishing attacks remain a threat to users and organizations in every country on the planet. One problem with centralized phishing detection systems is the need for protection of data privacy and the system is becoming more complex as new types of attack occur. This research develops an improved phish discovering method based mostly on federated studying by processing plenty of person datasets privately to implement superior outcomes. Moreover, this approach allows for the combining of multiple local models across all user devices, leading to improved phishing detection results in comparison to single models while maintaining the privacy of raw data. When tested against varying datasets our system is shown to be superior by having better scalability, better adaptability to new threats, and better ability to protect user credentials.

Keywords: Phishing Detection, Federated Learning, Cybersquatting, Privacy Preservation, Decentralized Machine Learning.

Abstract: 1. Introduction and Cloud Security Landscape 1.1 Overview of Cloud Computing Adoption and Security Imperatives Cloud computing has fundamentally transformed enterprise information technology infrastructure, enabling organizations to achieve unprecedented scalability, flexibility, and operational efficiency. However, this rapid adoption has concurrently expanded the attack surface and introduced novel security challenges that traditional security paradigms struggle to address. The expansion of cloud deployment across public, private, and hybrid models has created increasingly complex security management requirements, with organizations now responsible for protecting distributed infrastructure spanning multiple geographic regions and service providers . The critical challenge facing cloud security teams is the inherent complexity of cloud configurations. As organizations migrate to cloud-native architectures, they often lack visibility into their own infrastructure configurations, creating silent vulnerabilities that can remain undetected for extended periods. Research indicates that misconfiguration has emerged as one of the most significant threat vectors affecting cloud security. Major security breaches, including the 2019 Capital One incident and the Toyota data exposure, have been attributed primarily to cloud misconfiguration rather than sophisticated zero-day exploits . These incidents demonstrate that configuration errors pose threats comparable to or exceeding those from advanced persistent threats, yet misconfiguration detection remains inadequately addressed by traditional security tools.

Abstract: Groundwater is one of the most essential natural resources for drinking, agriculture, and industrial purposes. However, rapid urbanization, industrialization, and agricultural activities have significantly affected its quality. This study focuses on the analysis of groundwater quality in Jalgaon and Jamner municipal corporation areas. Various physicochemical parameters and Water Quality Index (WQI) methods are used to assess the suitability of groundwater for drinking and other uses. The study also evaluates seasonal variations and identifies contamination sources. The results highlight the need for proper groundwater management and pollution control strategies to ensure sustainable usage.

Abstract: The construction industry plays a major role in economic development, but it also consumes a large amount of natural resources and energy. Conventional construction materials such as cement, steel, and concrete require significant energy for production and contribute to environmental pollution. In recent years, researchers have focused on developing sustainable and energy efficient construction materials to reduce environmental impact. Natural fibres have emerged as an effective alternative due to their renewable nature, low cost, and environmentally friendly characteristics. Among various natural fibres, banana fibre is considered highly promising because it is widely available as agricultural waste in many tropical countries. This research paper presents an experimental investigation of banana fibre as an energy efficient material for construction. The main objective of this study is to evaluate the potential of banana fibre in cement-based construction materials and examine its effect on mechanical and thermal properties. Banana fibres were extracted from banana plant stems and processed before being mixed with cement mortar in different proportions. Laboratory tests such as compressive strength tests, tensile strength tests, and thermal insulation tests were conducted to evaluate the performance of fibre reinforced materials. This research focuses on the experimental investigation of an energy-efficient cavity wall using banana fibres as an insulation material. Two wall models were constructed for the experiment: a conventional solid wall model and a cavity wall model filled with banana fibres.

Keywords: Banana Fiber, Energy Efficient Materials, Cavity Wall, Thermal Insulation, Sustainable Construction, Heat Transfer

Abstract: Food safety and quality assurance are major concerns in modern food processing and packaging industries. Conventional inspection systems mainly rely on manual monitoring, which may lead to inconsistent detection of defective packages and increased operational costs. This project presents the design and development of a Cold Plasma Conveyor System with Leak Detection using Python-based image processing for automated inspection and sterilization of packed food products. In the proposed system, a conveyor mechanism driven by a DC motor transports the packaged products through different processing stages. A USB camera captures real-time images of the packages, which are analyzed using Python image processing algorithms to detect defects such as packet leakage, cracks, or improper sealing. If a defective package is identified, a servo motor-based rejection mechanism automatically removes it from the conveyor line. The accepted products then move to the cold plasma treatment zone, where non-thermal plasma is applied to sterilize the package surface and eliminate microorganisms without affecting food quality. The system improves inspection accuracy, reduces human intervention, and enhances hygiene standards in automated food processing environments

Keywords: Cold Plasma Sterilization, Conveyor Automation, Image Processing, Leak Detection, Python-Based Inspection, Food Packaging Safety, Servo-Based Rejection System, Automated Quality Control.

Abstract: This project presents the design and development of an AI Integrated Mechanical Drying System for Post Harvest Groundnut with Moisture Sensors, aimed at improving the efficiency, accuracy, and quality of the drying process. Traditional drying methods such as open sun drying and conventional hot-air drying are highly dependent on environmental conditions, require significant manual effort, and often result in uneven moisture removal, contamination, and quality degradation.To overcome these limitations, the proposed system integrates Artificial Intelligence (AI), Internet of Things (IoT), and sensor-based automation, with a special focus on moisture sensing technology for precise drying control. The system is built around an ESP32 microcontroller, which collects real-time data from temperature, humidity, load cell, and moisture sensors. The moisture sensor plays a crucial role by directly measuring the moisture content of groundnuts, while the load cell monitors weight reduction to estimate moisture loss.An AI-based control algorithm analyzes the collected data and predicts optimal drying conditions. Based on this analysis, the system automatically controls the heater and blower through relay modules to maintain proper temperature and airflow. The system also incorporates IoT connectivity, enabling remote monitoring and control through mobile applications such as Blynk, allowing users to track drying parameters and receive real-time updates.The integration of moisture sensors significantly enhances drying accuracy by ensuring that the groundnuts reach the desired safe moisture level without over-drying or under-drying. This improves product quality, shelf life, and reduces the risk of fungal contamination. Additionally, the system minimizes energy consumption and reduces human intervention, making it efficient, reliable, and suitable for both small-scale and commercial applications.

Keywords: Artificial intelligence (AI), Internet of Things (IoT), ESP32 microcontroller, mechanical drying system, groundnut drying, moisture monitoring, predictive control, energy efficiency, automation.

Abstract: IT Service Management (ITSM) has historically been a digitally bounded discipline, limited to data artefacts such as logs, tickets, and network telemetry. This paper introduces Physical-Digital ITSM Integration (PDII) as a new paradigm wherein AI-powered video analytics, deployed on edge computing infrastructure, becomes the primary intelligence layer connecting the physical operational environment to ITSM workflows. Drawing on empirical research, industry case studies, and emerging AI capabilities including Vision Large Language Models (VLLMs), Small Language Models (SLMs) on edge hardware, and agentic orchestration architectures, this paper demonstrates that PDII represents a category-defining evolution of ITSM. The paper contextualises this evolution within India's industrial and regulatory landscape, including DPDP Rules 2025 and Industry 4.0 manufacturing imperatives, and concludes with a practical implementation framework and directions for future research.

Keywords: AI Video Analytics, ITSM, Edge AI, Physical-Digital Integration, Predictive Service Management, EdgeBox, Indoai, SLM, AIOps, DPDP 2025

Abstract: The increasing demand for efficient communication in educational institutions has highlighted the limitations of traditional enquiry handling systems, which rely heavily on manual responses from administrative staff. These conventional approaches are often time-consuming, prone to delays, and unable to scale effectively with the growing number of student queries. With advancements in Artificial Intelligence (AI), Machine Learning (ML), and Natural Language Processing (NLP), intelligent chatbot systems have emerged as a viable solution to automate and enhance user interactions. This paper presents EduConnect AI, an AI-driven admission counselor chatbot designed to provide instant, accurate, and context-aware responses to student enquiries across multiple institutions. The proposed system utilizes a hybrid architecture that combines rule-based response handling with NLP-driven intent recognition to ensure both reliability and adaptability. User queries are processed through an intelligent pipeline where text inputs are analyzed, classified into predefined intents such as admissions, courses, fees, and facilities, and mapped to relevant responses. The backend is implemented using FastAPI, enabling high-performance request handling and seamless integration with frontend components developed using modern web technologies such as HTML, CSS, and JavaScript. To improve response accuracy and user trust, the chatbot incorporates explainable reasoning by highlighting key factors influencing its responses. Additionally, a multi-institution recommendation module allows users to explore and compare information across different educational providers without manual intervention. The system also supports scalable data storage and analysis through structured datasets, enabling efficient tracking and visualization of user interactions. Furthermore, the platform is designed with a user-friendly interface that ensures smooth navigation and real-time communication, enhancing overall user experience. Experimental evaluation demonstrates that the chatbot significantly reduces response time, improves accessibility to information, and minimizes the workload on administrative staff. The results highlight the effectiveness of AI-based chatbot systems in transforming traditional enquiry processes into intelligent, automated, and scalable solutions for modern educational environments.

Keywords: AI Chatbot; Educational Enquiry System; Natural Language Processing; Machine Learning; FastAPI; Admission Counseling; User Interaction; Decision Support System; Automation; Web-Based System.

Abstract: Agriculture plays a vital role in the economy, and early detection of crop diseases is essential for improving yield and reducing losses. Traditional methods of disease identification rely heavily on manual inspection by experts, which is time-consuming, costly, and often inaccessible to farmers in rural areas. This paper presents Crop Disease Detection, an AI-powered web-based system designed to detect crop diseases from leaf images and predict weatherbased disease risks. The system utilizes advanced Artificial Intelligence models, including GPT-4o-mini Vision for image-based disease detection and Gemini Flash for real-time weather risk analysis. The proposed solution allows users to upload leaf images, analyze crop health, and receive detailed information such as disease name, severity, symptoms, and treatment recommendations. In addition, the system integrates weather-based risk prediction using parameters such as temperature, humidity, and rainfall to forecast potential disease outbreaks. The system also provides multilingual support, voice assistance, and a structured database for storing and analyzing historical data. Experimental results show that the system improves accuracy, reduces manual effort, and provides real-time assistance to farmers. Crop Disease Detection offers a scalable and intelligent solution for modern precision agriculture and can be extended further with IoT integration and advanced predictive analytics.

Keywords: Crop Disease Detection, Artificial Intelligence, Image Processing, Precision Agriculture, Weather Risk Prediction, Machine Learning, Smart Farming, GPT Vision AI, Agricultural Analytics

Abstract: Recycling of concrete debris can make a contribution to reduce the total environmental impact of the building sector. To increase the scope for recycling in the future, aspects of recycling have to be included in the design phase. Besides, aggregate sources near Metro Manila are almost depleted, so aggregates have to be brought from far quarries. Consequently, reclaiming aggregates from concrete debris would lead to environmental and economic benefits. This experimental study aimed to use crushed concrete debris as alternative fine aggregate in a mortar mixture. A conventional mortar mixture will be compared to concrete debris mixture of the same proportions.

Keywords: Aggregate, concrete debris, construction material, mortar mixture, recycled waste

Abstract: The AI vision based weight estimation of crab is an innovative, contactless solution designed to estimate the weight of crabs without the use of conventional load cell sensors. This system integrates a Raspberry Pi 4 (4GB RAM) and an ESP32 microcontroller to create a smart, automated, and hygienic seafood grading platform.The system captures top-view images of crabs using a camera module connected to the Raspberry Pi. These images are processed using Python-based image processing techniques with the OpenCV library, where key features such as area, length, and width are extracted. A machine learning regression algorithm is applied to estimate the crab's weight based on these extracted features. The predicted weight is then displayed on a 16×2 LCD with I2C interface, providing real-time feedback to the user. Additionally, a push button interface is included to initiate image capture and processing, while a buzzer system provides audio alerts indicating successful measurement or error conditions. The ESP32 supports auxiliary control functions and enables future IoT integration for remote monitoring and data logging.Experimental analysis shows that the system achieves an accuracy of approximately 97-98%, with minimal error margins, making it a reliable alternative to traditional weighing methods. The contactless nature of the system ensures improved hygiene, making it particularly suitable for seafood processing industries.This project demonstrates the effective integration of embedded systems, computer vision, and machine learning, offering a low-cost, scalable, and automated solution for real-time crab weight estimation and grading.

Keywords: Crab weight detection, computer vision, machine learning, image processing, OpenCV, Raspberry Pi 4, ESP32, regression algorithm, feature extraction, contactless measurement, smart aquaculture, IoT integration, automation system, real-time monitoring, seafood grading system.

Abstract: The rapid advancement of artificial intelligence has significantly transformed human-computer interaction, enabling the development of intelligent virtual assistants capable of understanding and responding to natural language. This paper presents the design and implementation of TARS-D, a desktop-based AI voice assistant that facilitates seamless interaction through both voice and text-based commands. The system integrates speech recognition, natural language processing (NLP), and machine learning techniques to interpret user intent and execute a wide range of system-level operations. It utilizes speech-to-text conversion for capturing user input and text-to-speech synthesis for generating natural responses. The assistant is capable of performing tasks such as file and folder management, application control, web browsing, scheduling, and information retrieval. A key feature of TARS-D is its emphasis on privacy and offline functionality, as it processes user data locally rather than relying heavily on cloud services. The modular architecture of the system ensures scalability and ease of integration of new features. Additionally, the assistant improves accessibility by enabling hands-free interaction, making it beneficial for users with visual or physical impairments. The proposed system demonstrates an efficient, secure, and user-friendly solution for desktop automation, contributing to enhanced productivity and improved user experience in modern computing environments.

Keywords: Artificial Intelligence, Voice Assistant, Natural Language Processing, Speech Recognition, Desktop Automation, Human-Computer Interaction.

Abstract: Black cotton soil exhibits significant swelling and shrinkage characteristics, making it unsuitable for construction without stabilization [1]. This study evaluates the effectiveness of bio-enzyme stabilization using experimental data and compares its performance with lime stabilization based on published literature. Laboratory tests including California Bearing Ratio (CBR), Liquid Limit (LL), Plastic Limit (PL), and permeability were conducted. Results indicate that bio-enzyme treatment improves soil strength and reduces permeability. Comparative analysis shows that while lime stabilization yields higher strength improvement due to pozzolanic reactions, bio-enzymes provide a sustainable and eco-friendly alternative [2], [3], [11]. The study highlights the suitability of bio-enzymes for environmentally conscious soil stabilization practices

Keywords: Black Cotton Soil, Bio-Enzyme, Soil Stabilization, Lime Stabilization, CBR, Atterberg Limits, Sustainable Engineering.

Abstract: Plant pests and diseases are big problems for farmers because they often mean big loses in crops and food production. The project's goal is to use machine learning to create an easy and useful system that can predict early signs of plant diseases and pest attacks. The system can help the farmer predict early and take quick actions to protect the crops and also improve production. It collects information from pictures of plants and things surrounding it. It uses advanced tools like CNNs for the analysis of the pictures and signs of diseases and bugs. It has information regarding the environment so that the prediction becomes more precise and reliable. The system is an easy-to-use solution that helps farmers reduce losses, apply pesticides only where needed, and make the best decisions for their crops. This project protects the farmer's produce and allows the crops to be grown more environmentally friendly. It can work well with almost all type of crops.

Keywords: Plants Pests and diseases, Image Quality, Sustainable, crops, CNN, Machine Learning.

Abstract: The increasing demand for intelligent defense systems has driven the development of advanced threat prediction models. This paper presents SHIELD-A, a strategic AI-based air defense simulation system designed to analyze and predict aerial threats in real time. The proposed system utilizes machine learning techniques to improve prediction accuracy while reducing computational complexity compared to traditional methods such as inverse reinforcement learning. By employing simulated datasets, the system achieves faster response times and enhanced efficiency. Additionally, geospatial visualization and real-time alert mechanisms improve situational awareness and user interaction. The proposed approach is suitable for applications in defense, surveillance, and security systems.

Keywords: Artificial Intelligence, Machine Learning, Air Defense, Threat Prediction, Simulation System

Abstract: This study examines the geographical distribution of goats across the districts of Karnataka state. The main aim is to understand the spatial pattern of goat population and its relationship with physical and socio-economic conditions. Secondary data from official livestock sources are used and mapped using GIS techniques. The results show that higher goat populations are mainly concentrated in the rain shadow and semi-arid districts of the state. This pattern indicates that goat rearing is an important livelihood activity in areas with low rainfall and dry region. The study highlights the role of goat farming in supporting the rural economy and provides useful information for regional planning and livestock development and the scenario of drought condition.

Keywords: Goat Distribution, LiveStock, Goat and Climate

Keywords: Artificial Intelligence, Machine Learning, Air Defense, Threat Prediction, Simulation System

Abstract: Virtual gestures that control multimedia playback in order to allow users to interact with the media without having any direct contact with it. Computer vision techniques are applied to a laptop camera in order to detect and interpret hand gestures in real time. Based on Python with OpenCV and MediaPipe, it tracks the hand movements and translates them into actions like play/pause, volume adjustment. Since it does not use deep learning model and rather follows the rules, this process is quick and relatively easy to implement. The functionality is augmented by a trigger-based feature so that gesture recognition is only active when necessary, preventing unnecessary processing and increasing efficiency. It does well in its usual realm with speedy and correct answers. Our method is simple, low-cost and applicable in smart environments and assistive systems when touchless interaction is needed.

Keywords: Gesture Recognition, Human-Computer Interaction, Computer Vision, Multimedia Control, Touchless Interface

Abstract: This project presents an IoT-based AI system for continuous respiratory disorder monitoring, aimed at providing real-time health tracking and early identification of critical conditions. The system uses an Arduino Uno as the central controller, integrating sensors such as temperature, pressure, and heart rate to gather essential physiological data. In addition, the MPU6050 sensor is included to monitor body movement and detect fall events, improving patient safety. The collected data is displayed on an LCD screen and transmitted to a mobile device for remote monitoring and timely notifications. The system focuses on delivering accurate and continuous observation of patient health outside hospital environments. By combining IoT connectivity with intelligent data analysis, the proposed solution enhances healthcare accessibility and supports early medical intervention. It is especially useful for elderly individuals and patients with respiratory issues, helping to reduce health risks through efficient and smart monitoring.

Keywords: IoT, AI, Arduino uno, Sensors, Mobile notification, Respiratory Disorder Monitoring.

Abstract: This project introduces an intelligent, automated system designed for bird feeding and poultry management. Traditional poultry systems rely heavily on manual monitoring, which often results in significant food wastage and inefficient tracking of bird health. To overcome these challenges, we propose a "Smart Poultry System" that integrates automated temperature adjustment and real-time health monitoring. The system automates feeding schedules based on specific phases of bird growth. The architecture combines Arduino-based hardware with Machine Learning (ML) synergy, utilizing a multi-sensor array for real-time data acquisition and autonomous decision-making. By leveraging these technologies, the system ensures optimal environmental conditions and precise resource management, significantly improving the efficiency of poultry farming.

Keywords: Internet of Things (IoT), Machine Learning (ML), Data Acquisition (DAQ), Real-time Monitoring, Smart Poultry Management, Automatic Feeding System.

Abstract: The "Intelligent Vehicle Damage Assessment & Cost Estimator" project aims to assist vehicle owners by providing accurate and automated damage assessment using advanced deep learning techniques. Leveraging YOLOv5, a state-of- the-art object detection model, this system is designed to analyze and evaluate damage sustained by vehicles, including four-wheelers. The model is trained on a comprehensive dataset of vehicle damages to accurately identify and classify different types of damage. By integrating this technology, users can independently assess vehicle damage, reduce dependency on manual inspection, minimize errors, and obtain precise cost estimations for repairs, ensuring transparency and better decision making.

Keywords: Vehicle damage assessment, cost estimation, YOLOv5, user assistance, deep learning, object detection, two- wheeler, four-wheeler.

Abstract: Mental health is one of the most neglected areas in healthcare systems around the world. Issues like depression, anxiety, ADHD, and chronic stress affect millions of people, yet a large number never get screened or diagnosed. One core reason is that most current detection approaches rely heavily on self- reported information, which is often unreliable when people feel too embarrassed, are in denial, or simply lack awareness of their own condition. To address this, we built NeuroEye - a browser-based tool that tracks eye activity passively during a session to detect early signs of mental distress. Instead of asking users to fill out questionnaires, NeuroEye uses a regular webcam to observe natural eye behaviour in the background. It uses the MediaPipe Face Mesh library to detect facial landmarks and calculate the Eye Aspect Ratio (EAR) for monitoring blinks in real time. Gaze tracking is done by monitoring iris position across nine zones. These signals are then compared against known clinical thresholds to identify possible indicators of stress, low mood, fatigue, or attention issues. All processing happens locally on the user's device and no data is sent to any server, keeping everything completely private. NeuroEye is a screening support tool and is not a replacement for professional diagnosis.

Keywords: Affective Computing, Eye Aspect Ratio (EAR), MediaPipe Face Mesh, Mental Health Screening, Ocular Biometrics, Gaze Estimation, Non-invasive Assessment

Abstract: Sign language serves as a fundamental medium of communication for individuals with hearing and speech impairments. However, the lack of widespread understanding among the general population often creates a communication barrier, limiting effective interaction. This research presents the design and development of an intelligent sign language detection system capable of recognizing hand gestures corresponding to alphabets, as well as selected words and sentences, in real time. The proposed system leverages advanced computer vision techniques and deep learning algorithms to accurately interpret hand gestures captured through a camera interface. Image preprocessing methods are employed to enhance input quality, followed by feature extraction and classification using a trained model. The system is designed to achieve high accuracy and efficiency while maintaining robustness under varying lighting conditions and backgrounds. Furthermore, the model is trained on a diverse dataset of sign language gestures to ensure reliable performance across different users. The output is translated into readable text, enabling seamless communication between sign language users and non-signers. Experimental results demonstrate that the system performs effectively in recognizing both static and dynamic gestures. This work aims to bridge the communication gap between the deaf-mute community and the wider society, contributing to inclusivity and accessibility through the application of artificial intelligence and human-computer interaction technologies.

Keywords: Sign Language Detection, Computer Vision, Deep Learning, Gesture Recognition, Human-Computer Interaction, Image Processing, Real-Time Systems, Accessibility, Assistive Technology

Abstract: The proposed Disaster Response System is a hybrid IoT and Artificial Intelligence-based solution designed using an Arduino/ESP32 microcontroller and a laptop for advanced ML/DL processing. The system collects real-time environmental data such as soil moisture, rainfall, vibration, and water level using sensors connected to the microcontroller. This sensor data (text input) is transmitted to a laptop, where Machine Learning algorithms analyze patterns to predict disaster conditions. Additionally, Deep Learning models process image and video inputs to detect disasters such as floods, landslides, or structural damage. Based on predictions, the system generates alerts via Telegram cloud, and LCD display. This approach ensures faster response, improved accuracy, and reduced disaster impact.

Keywords: Arduino/ESP32 microcontroller, soil moisture sensor, rainfall detection sensor, Ultrasonic Sensor (Water Level) , MPU6050 (Vibration Sensor), Machine Learning, Deep Learning, Telegram cloud, LCD display

Abstract: This article explores the integration of databases to compare and evaluate the machine learning algorithms that suits multi-label classification. In this predictive modelling work, I have performed Logistic regression, Random Forest, XGBoost, Multi-Layer Perceptron. Logistic Regression is for linear model prediction of drug side effects, Random Forest is chosen for High-Dimensional Biomedical datasets, XGBoost was selected because it is one of the most powerful gradient boosting algorithms, Multi-Layer Perceptron (MLP) architecture was used to learn nonlinear relationships between drug features and side-effect labels. As this is a predictive modelling task evaluation of machine learning models is essential step to assess its predictive capability, reliability, and generalization performance using appropriate evaluation metrics. The computational analysis for this study was carried out using Python due to its flexibility and strong ecosystem of scientific computing libraries. In this study, multiple machine learning models were implemented and compared to evaluate their effectiveness in predicting drug side effects using molecular properties and pharmacokinetics features.

Keywords: Adverse drug reactions, Machine learning, Logistic regression, Random forest, XGBoost, Multi-layer perceptron, Neural network, precision, recall.

Abstract: This project presents an AI-enabled blogging platform designed to allow users to share blogs while ensuring safe and moderated content. The system includes three main users: public users, administrators, and government authorities. Public users can write and publish blogs along with images through a web interface. When a blog is submitted, the server processes the content using AI-based text moderation and image classification to detect abusive or inappropriate material. If the content is safe, it is automatically published on the public blog page. If the system detects harmful content, the blog is flagged and sent to the administrator for review and approval. The platform also includes an AI commenting feature that generates automated feedback on blog posts. Overall, the system improves content quality, user engagement, and responsible communication by integrating artificial intelligence with a modern blogging platform.

Keywords: Next.js, Blog Management System, User Authentication, PostgreSql, Server-Side Rendering, AI-Based Content Moderation, AI commenting, Image Classification, Web Security, Fast Api

Abstract: In this study, we present the mathematical design in Pineapple spirals and digital signals using field theory. The golden Angle and Fibonacci sequence explain the packing in natural spirals. Finite fields and recurrence relations are applied to describe these arrangements mathematically.

Keywords: Finite fields (Galois fields), Fibonacci sequence, Golden angle, Phyllotaxis, Pineapple spirals, Recurrence relations, Digital signals, Signal processing, Pseudo-random sequences, Cryptography, Packing efficiency, Phase mapping AMS Mathematics Classification: 92C80, 94A55, 11A07, 11T30, 11B39.

Abstract: The "Smart Market Intelligence & Opportunity Detection System" project aims to support small businesses and entrepreneurs by providing data-driven insights into market gaps, consumer behavior, and competitive dynamics. Leveraging Machine Learning and Data Science techniques, this system analyzes diverse data sources such as e¬commerce platforms, regional economic indicators, and online consumer sentiment to identify unmet product demand and emerging opportunities. It utilizes clustering algorithms, predictive models, sentiment analysis, and scoring mechanisms to evaluate feasibility, assess competition, and recommend strategic actions. By integrating features like market gap detection, price intelligence, and risk analysis, the system enables informed decision-making, reduces market uncertainty, and promotes sustainable business growth.

Keywords: Market intelligence, e-commerce analysis, sentiment analysis, business opportunity, small business, machine learning

Abstract: This report details a project focused on developing a sustainable, low-cost, and eco-friendly method for treating domestic kitchen wastewater using the principles of phytoremediation. Kitchen wastewater, a major component of household greywater, poses an environmental challenge due to its high concentration of organic matter, nutrients, and suspended solids, indicated by elevated Biochemical Oxygen Demand (BOD), Chemical Oxygen Demand (COD), and Total Suspended Solids (TSS). Conventional treatment systems are often energy-intensive and cost-prohibitive for decentralized application.The core objective of this study was to evaluate the efficiency and suitability of the wetland plant, Canna indica (Indian Shot), as the primary phytoremediation agent. The project sought to analyze changes in key water quality parameters ( PH, BOD, COD, and TSS) before and after treatment, and to develop a functional, low-cost treatment system.

Abstract: This paper presents the design and implementation of an IoT-enabled fire-fighting robot with integrated smoke detection for application in hazardous and fire-prone environments. The proposed system is intended to operate autonomously, enabling early identification and suppression of fire while minimizing human involvement. It incorporates flame and smoke sensors interfaced with a microcontroller to continuously monitor environmental conditions and detect fire-related anomalies with improved accuracy.Upon detection of fire or smoke, the robot navigates toward the affected region and activates an onboard extinguishing mechanism, such as a water-based spray system, to control and suppress the flames. The integration of IoT technology through Wi-Fi communication facilitates real-time data transmission . Users can access system parameters, including temperature, smoke concentration, and operational status, via a connected interface.The developed system enhances safety by reducing human exposure to hazardous conditions and improving response time during fire emergencies. It demonstrates the effective integration of IoT, and robotics in developing reliable and intelligent solutions for fire detection and control applications.

Keywords: : IoT, Fire Fighting Robot, Smoke Detection, Arduino, Flame Sensor, Temperature Sensor,Wireless Control.

Abstract: This report covers the development and implementation of a Fake Currency Detection System that uses machine learning and image processing. The rise of counterfeit banknotes presents a serious threat to financial systems and public trust. This project tackles this issue by creating an automated software solution that accurately distinguishes genuine currency from counterfeit notes. The system uses Python, OpenCV, and Scikit-learn to process images of currency notes, extract meaningful features, and classify them with a Support Vector Machine (SVM) algorithm. A user-friendly Tkinter-based Graphical User Interface (GUI) allows users to upload an image and receive an instant prediction of authenticity. The project showcases how AI can improve financial security while providing a cost-effective and scalable alternative to traditional hardware detectors.

Keywords: Fake currency detection, machine learning, image processing, support vector machine, Tkinter GUI.

Abstract: The process of growing mushrooms requires a lot of attention because it depends on the precise temperature, humidity, and moisture conditions in the growing medium. Unfortunately, with the application of conventional methods, the majority of these parameters are adjusted by hand, leading to many inconsistencies, increased labor costs, and decreased efficiency. This paper describes an innovative mushroom polyhouse design that incorporates the IoT technology, artificial intelligence, and ESP32-based microcontroller as well as a number of sensors. The system constantly monitors several main variables such as temperature, humidity, and the moisture of substrates and adjusts them by turning on the required devices, e.g. exhaust fans, humidifier, and pumps, in a closed-loop mode. Furthermore, the system is integrated with the cloud service where one can monitor all these variables and even adjust some of them. Such an approach increases automation and allows avoiding numerous mistakes. Consequently, it enhances the effectiveness of growing mushrooms, reduces expenses, and eliminates the likelihood of any losses due to improper adjustment of growing conditions.

Keywords: IoT, Smart Agriculture, ESP32, Mushroom Cultivation, Automation

Abstract: The exponential growth of digital image repositories across enterprise systems and the internet demands intelligent, scalable retrieval mechanisms capable of operating with high accuracy and efficiency. This paper presents a comprehensive AI-based large-scale image retrieval system that leverages the Contrastive Language-Image Pretraining (CLIP) model, specifically its Vision Transformer ViT-B/32 backbone, to extract rich 512-dimensional visual embeddings from images. The proposed system executes image indexing offline through batch processing, stores L2-normalized feature vectors, and performs real-time cosine similarity computation at query time to retrieve the top-K most visually similar images. Additionally, a Support Vector Machine (SVM) classifier trained on CLIP embeddings achieves 98.76% accuracy with a macro-average F1-score of 0.9804 across 27 image categories. The system is deployed as a responsive web application using the Flask framework, enabling end-users to perform real-time image-based searches through a browser interface. Comparative evaluation demonstrates that the proposed approach substantially outperforms all baseline methods including Dummy classifiers and Logistic Regression. The results confirm that deep visual embeddings derived from large-scale multimodal pretraining are highly effective for content-based image retrieval at scale. Keywords --- CLIP Embeddings, Content-Based Image Retrieval, Cosine Similarity, Vision Transformer, SVM Classification, Flask Deployment, Deep Visual Features, ViT-B/32, L2 Normalization.

Abstract: This presents the design and implementation of Interview scheduling, a monolithic web platform that unifies candidate onboarding, interview scheduling, live audiovisual collaboration, and structured evaluation. The system is implemented using Spring Boot 3 on Java 17, with a MySQL database accessed through Spring Data JPA, and browser-based clients built from static HTML and JavaScript served by the same application. Candidates register, verify email, keep detailed profiles with document uploads, and reserve interview slots created by administrators. A WebRTC mesh topology carries peer media; signaling, chat, typing, and presence use STOMP over WebSocket (SockJS endpoint). Administrators control global meeting lifecycle through authenticated HTTP operations protected by an opaque meeting administrator token. Speech segments are transmitted to an external Whisper-compatible HTTP transcription service; transcripts feed communication analytics (e.g., speaking rate and filler-word statistics) combined with technical, behavioural, and profile-derived scores. Weighted ranking produces comparable outcomes across candidates. Optional Judge0 integration evaluates coding submissions in a sandbox. Results can be published as PDF reports with embedded charts. The work explicitly documents security trade-offs: Spring Security is configured to allow all HTTP requests in the current codebase, and user identity for REST calls relies on client-supplied identifiers-acceptable for a controlled academic deployment but requiring hardening for production. The contribution is an end-to-end, implementation-grounded architecture description suitable for academic evaluation and further industrial extension.

Abstract: This study examines surface acting behaviour and its impact on employee well-being among BPO employees in Chennai. Surface acting, a key dimension of emotional labour, refers to the modification of outward emotional expressions without corresponding internal feelings. The study aims to assess the prevalence and intensity of surface acting, analyse the influence of organizational factors such as performance targets and shift schedules, and evaluate its psychological and organizational consequences. Primary data were collected from 150 respondents using a structured questionnaire. Statistical tools including descriptive analysis, Chi-square test, correlation, and regression were applied for data analysis. The findings indicate that surface acting is moderately prevalent among BPO employees and is significantly influenced by shift type and exposure to difficult customers. However, no significant association was found between job role and surface acting behaviour. The study further reveals that sustained surface acting leads to emotional dissonance, stress, and burnout, which negatively affect job satisfaction, organizational commitment, and employee retention. The findings emphasize the need for organizations to adopt supportive practices to reduce emotional strain and improve employee well-being.

Keywords: Surface Acting, Emotional Labour, Employee Well-being, Burnout, BPO Sector, Organizational Factors, Customer Interaction, Job Satisfaction, Organizational Commitment

Abstract: A nonlinear tumor-immune interaction model is used to show the effect chemotherapy has on the qualitative aspects of cancer cell proliferation. The model uses three coupled ordinary differential equations for tumor cells (T), effector immune cells (E) and chemotherapy (C). Tumor growth is modeled using a logistic equation; immune mediated killing of tumor cells (via a saturated response); stimulation of immune activity via tumor antigens; immunological exhaustion (i.e., reduction in immune activity due to chronic antigen presentation); cytotoxic effects of chemotherapy directly on tumors and indirect effects on the immune system leading to suppression. The mathematical nature of this system is studied in terms of its fundamental characteristics (positivity, boundedness and existence of an invariant set which satisfies biological constraints) so as to establish that the system will be well posed. In addition to establishing the existence of tumor free and co-existence steady-states, a linearization about each steady-state using the Jacobian matrix and application of the Routh-Hurwitz criteria are used to examine the local stability of the steady-states. Using the chemotherapy input rate as a control parameter to induce bifurcations within the system it is demonstrated that this model can exhibit transcritical and Hopf bifurcations. These types of bifurcations are shown to explain transitions in tumor burden (persistent or eliminated) in relation to levels of chemotherapy use, transitions between an immune controlled state of co-existence, oscillatory behavior related to remission-relapse patterns and complete tumor eradication. Numerical simulations were performed to confirm the results obtained through analytical techniques and to determine a critical chemotherapy dose level at which persistent tumor burden changes to tumor elimination. Finally, sensitivity analyses were conducted to demonstrate that treatment efficacy was dependent upon several factors, including intensity of chemotherapy, efficiency of the immune response against tumors, aggressiveness of the tumor population and toxic effects of chemotherapy.

Keywords: Tumor-immune interaction, chemotherapy, bifurcation analysis, local stability, tumor-free equilibrium, coexistence equilibrium, Hopf bifurcation, transcritical bifurcation, mathematical oncology, numerical simulation.

Abstract: The metal matrix composite with Aluminium -5wt%Magnesium (Al-5Mg) alloy matrix and the industrial waste of micro sized Red mud particles were used as reinforcement material with different proportions (5, 10 and 15wt.%) and prepared by stir casting technique followed by Equal Channel Angular Pressing (ECAP). It was noticed that there is a significant improvement of mechanical properties with an increase in the proportion of reinforcement of red mud up to certain extent in the Al-5Mg matrix composite. Casted billets of composite material are subjected to an annealing treatment of 2250C for 30 minutes so as to homogenize the microstructure of the material. The specimens were prepared from these composites for Equal channel angular pressing (ECAP) operation. The effect of ECAP on the microstructure and mechanical properties of Al-5Mg Red mud composite were evaluated for various weight fractions of red mud. The distribution of reinforcement particles, porosity and grain growth was observed from the optical microscope. The mechanical properties of Al-5Mg base material such as hardness and density were compared with the as-cast Al-5Mg Red mud and ECAPed composites.

Keywords: Al-5Mg metal matrix composite, Red mud, ECAP

Abstract: Accurate people counting is essential for crowd management, smart surveillance analytics, and crowd monitoring. Manually counting people is a tedious task that is vulnerable to error. In this paper, a people counting system based on deep learning and computer vision is proposed. The proposed system is based on the use of pre-trained convolutional neural networks like YOLO for people detection. The proposed system is able to perform under various lighting conditions. Experimental results show that the proposed system is accurate and can perform in real-time. The results show that the proposed system is suitable for smart city applications.

Keywords: People Counting, Deep Learning, Computer Vision, Object Detection, YOLO.

Abstract: Customer churn prediction in Business-to-Business (B2B) Software-as-a-Service (SaaS) environments presents unique analytical challenges that differ fundamentally from consumer-facing churn contexts. Subscription-based enterprise software companies face heightened churn risk at contract renewal boundaries, where complex organizational buying decisions involve multiple stakeholders and extensive switching-cost evaluations. This study investigates and systematically compares five machine learning classification methodologies - Logistic Regression, Decision Trees, Random Forests, Gradient Boosting (XGBoost), and Multi-Layer Perceptron Neural Networks - applied to B2B SaaS enterprise client behavioral data for predicting customer churn with operational precision. The research employs a comprehensive preprocessing pipeline encompassing median imputation for missing values, Interquartile Range (IQR)-based Winsorization for outlier treatment, Min-Max normalization, and Synthetic Minority Over-sampling Technique (SMOTE) for class imbalance mitigation. Feature engineering and dimensionality reduction are performed using chi-square statistical testing and Random Forest importance scoring. Experimental evaluation across stratified 10-fold cross-validation demonstrates that ensemble methods, particularly Gradient Boosting, consistently achieve superior classification performance - attaining AUC-ROC of 0.934, Precision of 0.843, Recall of 0.962, and F1-Score of 0.899 on the importance-selected five-feature subset. Feature importance analysis identifies CustomerCount and Products as the primary churn drivers, collectively accounting for over 75% of cumulative predictive importance, revealing that operational dependency breadth and platform integration depth are the fundamental determinants of enterprise client retention. Statistical significance of performance differences is confirmed via the Friedman test and Nemenyi post-hoc analysis. Findings provide actionable guidance for customer success teams, enabling data-driven prioritization of retention interventions and proactive risk mitigation in enterprise SaaS environments.

Keywords: customer churn prediction, B2B SaaS, machine learning, Random Forest, Gradient Boosting, XGBoost, feature selection, enterprise software, SMOTE, imbalanced classification, predictive analytics.

Abstract: The contemporary job market has become increasingly complex, creating a powerful demand for smart systems that offer individualized career guidance. Common methods of career support are unsuccessful in adapting to personal requirements and rapidly changing skill needs. To address this challenge, this paper proposes AUTONIX, a modular career intelligence platform integrated with AI, aimed at providing scalable and flexible employability support. The proposed system leverages Large Language Models (LLMs) to perform tasks including resume optimization, interview simulation, cover letter generation, and industry trend analysis. The platform is based on a full-stack architecture incorporating modern technologies such as Next.js for the frontend, Prisma ORM and NeonDB for data management, Clerk for secure authentication, and Inngest for asynchronous workflow processing. Structured prompt engineering is a core feature of the system, ensuring that AI-generated outputs are context-aware, consistent, and aligned with user requirements. By utilizing the Gemini API, the system transforms unstructured inputs such as resumes and job descriptions into actionable and insightful information. Experimental results demonstrate that AUTONIX produces high-quality, role-specific, and ATS-compatible outputs while maintaining system stability and responsiveness. The platform advances the application of AI-based career advisory systems and contributes to the development of scalable, intelligent, and real-world-deployable career guidance solutions.

Keywords: Artificial Intelligence, Career Guidance System, Full-Stack Architecture, Large Language Models, Resume Optimization.

Abstract: TradeLens AI is a risk-aware and explainable trade recommendation system developed to improve transparency and reduce unnecessary exposure in algorithmic trading. The framework follows a multi-layer architecture consisting of a data ingestion module, a predictive modeling unit, and a decision engine designed to evaluate risk before execution. The predictive component uses an XGBoost model trained on high-frequency financial data obtained from sources such as yfinance and Finnhub. Instead of directly acting on model outputs, predictions are passed through a Risk-Aware Decision Engine (RADET), which applies a confidence threshold and evaluates volatility conditions before approving trade signals. Simulation results indicate that this layered approach significantly reduces low-quality trade entries while maintaining high prediction reliability. Additionally, the system provides interpretable outputs, allowing users to understand the factors influencing each decision. By combining predictive performance with transparent risk control, TradeLens AI contributes toward more reliable and accountable automated trading systems.

Keywords: Risk-Aware Trading, Decision Support System, Machine Learning, Decision Tree, XGBoost, Explainable AI, Financial Risk Management, Trade Recommendation.

Abstract: The IoT-Based Surveillance Robot is developed to provide safe and efficient monitoring in hazardous environments such as war fields, industrial areas, and public places. The main objective of this project is to reduce human risk by replacing direct human involvement with a remotely operated robotic system. The robot is controlled through an Android smartphone using Internet of Things (IoT) technology, allowing the user to operate it from a safe distance. The proposed system uses an ESP32-CAM module to capture and transmit live video through a web browser. This enables real-time surveillance of the surroundings and helps the user monitor the movement of the robot continuously. To improve navigation and safety, an ultrasonic sensor is used to detect obstacles in front of the robot. Whenever an obstacle is detected, the robot stops automatically to prevent collisions. A metal detector sensor is also integrated into the robot to identify metallic objects such as hidden bombs, weapons, or explosive materials. When a metallic object is detected, the system immediately activates a buzzer alert to warn the user. The entire system is controlled by an Arduino UNO microcontroller, which coordinates all sensors, motors, and communication modules. Thus, the project provides a low-cost, reliable, and efficient solution for surveillance and safety applications.

Keywords: IoT, Surveillance Robot, ESP32-CAM, Arduino UNO, Ultrasonic Sensor, Metal Detector Sensor, Obstacle Detection, Live Video Streaming, Remote Monitoring, Hazardous Environment.

Abstract: Agriculture is an important sector that supports the livelihood of many people, especially in developing countries. However, plant diseases can reduce crop productivity and cause losses to farmers. Identifying these diseases at an early stage is important so that proper treatment can be given on time. In this project, we developed FloraScan, a web-based system that detects plant diseases from leaf images using deep learning techniques. The system uses a Convolutional Neural Network (CNN) with transfer learning based on MobileNetV2 to classify diseases in tomato, potato, and bell pepper plants. Users can upload an image of a leaf through the web interface, and the system predicts the disease and also provides basic information such as possible treatments and preventive measures. The model achieved an accuracy of around 97.22%, which shows that deep learning can be useful for early plant disease detection.

Keywords: Plant Disease Detection, Deep Learning, CNN, Transfer Learning, MobileNetV2, Agriculture

Abstract: Debugging remains one of the most cognitively demanding tasks in software development, particularly for novice programmers, due to the limited interpretability of traditional compiler and runtime error messages. Existing solutions, including integrated development environments (IDEs), online forums, and generic AI assistants, fail to provide context-aware, code-specific explanations within a unified workflow. To address these limitations, this paper presents a RAG-augmented AI Code Debugger that integrates Retrieval-Augmented Generation (RAG) with locally deployed Large Language Models (LLMs) for explainable and context-aware error analysis. The proposed system combines a browser-based code editor with a FastAPI backend, a subprocess-based execution engine, and a multi-stage AI pipeline consisting of error parsing, semantic retrieval using ChromaDB, and inference via a locally hosted DeepSeek Coder model using Ollama. A composite query strategy, incorporating error type, error message, and faulty code snippets, is used to retrieve relevant debugging knowledge from a structured knowledge base, which is then injected into the LLM prompt to improve response grounding. The system generates structured outputs including error interpretation, root cause analysis, and corrected code, enhancing both usability and educational value. Experimental observations demonstrate that the integration of RAG significantly improves the relevance and specificity of debugging explanations compared to direct LLM prompting. Additionally, the use of on-device LLM inference ensures data privacy, eliminates API dependency, and enables cost-effective deployment. The system also introduces an interactive mechanism for handling input-dependent programs, improving robustness in real-world debugging scenarios. Overall, the proposed approach highlights the effectiveness of combining retrieval-based reasoning with local LLMs to build intelligent, explainable, and scalable debugging assistants.

Keywords: Retrieval-Augmented Generation (RAG), Large Language Models (LLMs), Code Debugging, Explainable AI, ChromaDB, Semantic Search, Ollama, DeepSeek Coder, Software Engineering, AI-Assisted Development

Abstract: Breast cancer is one of the most prevalent and life-threatening diseases affecting women globally, where early detection plays a crucial role in reducing mortality rates. In recent years, emerging Artificial Intelligence (AI) techniques, particularly Machine Learning (ML) and Deep Learning (DL), have shown remarkable potential in improving the accuracy and efficiency of breast cancer detection and diagnosis. This systematic review presents a comprehensive overview of global research efforts that leverage AI-based methodologies across various medical imaging modalities, including mammography, ultrasound, magnetic resonance imaging (MRI), and histopathological imaging. The study reviews traditional ML algorithms such as Support Vector Machines (SVM), Decision Trees, Random Forests, and k-Nearest Neighbors (k-NN), alongside advanced DL architectures including Convolutional Neural Networks (CNNs), Deep Neural Networks (DNNs), and transfer learning models. The analysis highlights that DL approaches, especially CNNs, significantly outperform conventional ML techniques due to their ability to automatically extract complex features from large-scale datasets. Additionally, the review discusses hybrid and ensemble models that combine ML and DL techniques to enhance predictive performance. Key challenges identified include limited availability of high-quality annotated datasets, class imbalance, overfitting, lack of interpretability, and issues related to generalization across diverse populations and imaging systems. The review also emphasizes the growing importance of explainable AI (XAI), data privacy, and ethical considerations in clinical deployment. Comparative insights from global studies reveal varying levels of accuracy and robustness depending on data sources, preprocessing techniques, and evaluation metrics. Overall, the findings suggest that AI-driven approaches hold significant promise in supporting radiologists, improving diagnostic accuracy, and enabling early-stage detection of breast cancer. Future research directions focus on developing standardized datasets, improving model transparency, and fostering interdisciplinary collaboration to ensure reliable and scalable real-world applications.

Keywords: Breast Cancer Detection, Artificial Intelligence (AI), Machine Learning (ML), Deep Learning (DL), Convolutional Neural Networks (CNN), Medical Imaging, Mammography, Ultrasound Imaging, MRI.

Abstract: Smart agricultural monitoring plays a crucial role in improving crop yield and maintaining soil health. In India, many farmers face challenges in selecting suitable crops and managing soil nutrients, often leading to excessive fertilizer usage and financial loss. This paper presents Smart Soil Sense, an Internet of Things (IoT) based system that analyzes soil nutrients including Nitrogen (N), Phosphorus (P), and Potassium (K) along with environmental factors such as temperature, humidity, pH, and rainfall to recommend appropriate crops. The system integrates real-time sensor data collection using ESP32 microcontroller, NPK sensors, and DHT11/DHT22 environmental sensors. Multiple machine learning algorithms were evaluated, including Decision Tree, Naive Bayes, Support Vector Machine, Logistic Regression, Random Forest, and XGBoost. Experimental results indicate that XGBoost achieves the highest accuracy of 99.31%, demonstrating its effectiveness for crop prediction. The proposed system also incorporates a disease detection module using convolutional neural networks (CNN) for image classification to identify plant leaf diseases and recommend appropriate pesticides. Overall, the approach offers a cost-effective and reliable solution for improving agricultural productivity and supporting data-driven decision-making in farming practices, with potential to reduce fertilizer misuse by up to 30-35% and improve crop yields through precision agriculture.

Keywords: Smart Agriculture, Internet of Things (IoT), Machine Learning, Soil Analysis, XGBoost, Crop Recommendation, Precision Farming, NPK Sensors, ESP32.