VOLUME 14, ISSUE 5, MAY 2025

Abstract: Emotion recognition from speech has gained significant attention in the field of human-computer interaction, where it plays a crucial role in creating empathetic and responsive systems. Traditional speech recognition systems focus on transcribing words, while Speech Emotion Detection (SED) aims to identify underlying emotional states from speech signals. In this research, we propose a machine learning-based SED system utilizing both classical and deep learning approaches for emotion classification. The system processes audio samples from the RAVDESS dataset, extracting features like MFCCs, Chroma, and Spectral Contrast using the Librosa library. The classification task is performed using Support Vector Machine (SVM) and Convolutional Neural Network (CNN) models. Experimental results indicate that the CNN model outperforms the SVM model, achieving a classification accuracy of 91.45% compared to SVM’s 85.60%. The CNN's superior performance is attributed to its ability to learn high-level features from spectrogram representations. The system demonstrates its applicability in various domains such as virtual assistants, educational tools, and adaptive entertainment platforms. This study underscores the potential of deep learning techniques in improving emotion detection accuracy and suggests future directions, including multilingual datasets and real-time applications on edge devices.

Keywords: Speech Emotion Detection, Machine Learning, Convolutional Neural Network (CNN), Support Vector Machine (SVM), Feature Extraction, RAVDESS Dataset, MFCC, Emotional Classification.

Abstract: AI-based predictive battery health monitoring system to address challenges associated with lithium-ion battery failures and degradation in electric vehicles and renewable energy systems. By employing machine learning and deep learning algorithms, including CNNs, LSTMs, Logistic Regression, KNN, and SVM, the system accurately predicts key parameters such as State of Health, State of Charge, and Remaining Useful Life. Comparative analysis using datasets like NASA’s highlights the superior performance of CNN and LSTM models over traditional rule-based methods. MATLAB Simulink simulations enhance data quality for training and testing, while novel feature extraction techniques ensure robust model performance across diverse conditions. The system achieves a high accuracy of 0.986 in predicting battery metrics, demonstrating strong noise resilience and dynamic adaptability. These results emphasize the potential of AI-driven battery management systems to improve maintenance strategies, reduce operational costs, and promote the sustainable use of lithium-ion batteries.

Keywords: State of Health, State of Charge, Remaining Useful Life, CNN, LSTM, MATLAB, Logistic Regression, KNN, SVM.

Abstract: Urban traffic management is getting more and more difficult as the city expands and the number of vehicles increases. To address this, we put forward an AI-Based Dynamic Traffic Management System with Real-Time Detection & Priority Signal Optimization that, through computer vision and object identification, can effectively monitor and direct traffic flow. Conventional systems usually rely on pre-programmed timers or physical sensors, which can lead to bad timing of signals and slow reaction to real road conditions. Our system eliminates the use of these external sensors since it uses live video feeds to identify vehicles and pedestrians in real-time. It constantly monitors traffic density and flow patterns, enabling traffic signals to adjust dynamically instead of adhering to a fixed schedule. One of the key advantages of this system is its capacity to give priority to emergency vehicles and increase pedestrian safety at crossings. This makes emergency responses quicker and walkways safer for the public. By shifting from static, sensor-based techniques and embracing an AI-driven, vision-based solution, the system provides a more intelligent, scalable, and affordable solution for traffic management in today's world.

Keywords: Dynamic Traffic Management, Real-Time Object Detection, AI-Based Traffic Control, Traffic Signal Optimization, Emergency Vehicle Prioritization, Pedestrian Safety, Computer Vision in Traffic Systems, Smart City Traffic Solutions, Adaptive Traffic Signal Timing

Abstract: The rising popularity of digital image sharing requires verification systems for visual content authenticity. Digital media reliability suffers because of image tampering that takes place on social platforms. The current detection approaches fail with degraded images while unable to restore lost content. This research presents PFDNet as a deep learning-based framework which detects photo tampering and restores authentic content through its framework. The Cyber Vaccinator module generates a tamper-proof updated image through integration of the actual content with edge information. The Invertible Neural Network (INN) performs alteration detection in its forward process and restores original content in its backward process. The accuracy verification function of Run-Length Encoding (RLE) exists for restoration purpose. The experiment results demonstrate PFDNet successfully recognizes tampered images while restoring them faithfully and maintaining their authenticity.

Keywords: Photo Forgery Detection, PFDNet, Image Tampering, Deep Learning, Cyber Vaccinator, Tamper Resistance

Abstract: Sign Speak is an AI-driven system designed to enable real-time, bidirectional communication between hearing individuals and those with speech or hearing impairments. It translates speech into animated sign language and recognizes hand gestures to generate spoken output in multiple languages, including English, Kannada, Tamil, and Hindi. The system integrates Speech Recognition, Natural Language Processing, and Computer Vision using OpenCV and MediaPipe. It leverages Google Translate for multilingual support and gTTS for voice synthesis. Built on a Flask backend with a responsive HTML, CSS, and JavaScript frontend, Sign Speak performs reliably under varied conditions.Designed for scalability, the system allows easy integration of updates like dynamic gesture recognition and regional sign language support. Testing has shown high accuracy and seamless module coordination. Future enhancements include mobile and wearable versions, continuous gesture recognition, emotion detection, and AR/VR integration—advancing its mission of inclusive, accessible communication.

Keywords: Sign Language, Speech Translation, Computer Vision,, Accessibility, Multilingual Translation Speech Recognition, and Natural Language Processing.

Abstract: This report outlines a deep learning-based method for monitoring social distancing by analyzing the spacing between individuals in shared environments. The aim is to reduce the spread of infectious viruses like HMPV, and other potential future threats. The system uses video input, applying the YOLOv3 model—a pre-trained neural network—to detect pedestrians. To measure distances between people accurately, the footage is converted into a top-down, two-dimensional perspective. Individuals who don't maintain the required safe distance are flagged with red boxes and connecting lines. The model was tested with pre-recorded pedestrian footage and showed strong performance in identifying distancing violations. This approach can be effectively used in hospitals and other public settings to help control disease transmission.

Keywords: Social distancing, Deep learning, YOLOv3, Object detection, Computer vision.

Abstract: This project proposes an intelligent toll gate security system integrating machine learning and IoT-based sensors for enhanced vehicle monitoring and safety checks at toll gates. When a vehicle approaches the toll, a camera activates to capture and detect the vehicle's number plate, identifying it through image processing algorithms. Additionally, metal sensors analyses the vehicle to detect any unauthorized metallic objects, such as weapons. If any potential threat is detected, the system prevents the vehicle from passing by automatically controlling a gate mechanism powered by a DC motor connected to an ESP-32. The system also alerts higher authorities through a notification sent via a Telegram bot. In cases where police personnel allow vehicles to pass without thorough inspection, notifications are sent to higher authorities, with vehicle tracking for further monitoring. This setup uses a laptop camera as the visual input for machine learning tasks, while ESP-32 manages sensors and gate operations through UART communication. The proposed system enhances toll gate security, ensuring strict vehicle checks and real-time alerts to prevent unauthorized or hazardous vehicle entry.

Keywords: Check post Security, Unauthorized Access, Alerts, Weight detector.

Abstract: The Smart School Bus Student Monitoring System enhances student safety and automates attendance using RFID technology. A microcontroller like NodeMCU or Arduino processes data, recording attendance and timestamps, which are sent to a central server accessible by school staff and parents. Unlike GPS-based systems, it focuses solely on student tracking. In case of delays, drivers can input reasons, triggering real-time notifications to parents via Telegram chat Bot or SMS. The system is cost-effective, scalable, and offers a modern alternative to traditional school bus monitoring methods.

Keywords: IoT, RFID, Microcontroller, Student Monitoring, School Bus Tracking and safety, Attendance, Real-time Alerts.

Abstract: This project presents a secure and efficient method for embedding audio files into digital images using Least Significant Bit (LSB) encoding, ensuring covert communication while preserving the visual integrity of the carrier image. To enhance security, private PIN encryption prevents unauthorized access, while self-destructive encryption ensures time-based deletion of hidden data. Additionally, an intelligent compatibility alert system detects potential encoding issues, ensuring seamless operation across different devices. A hidden data verification system determines whether an image contains embedded audio, reducing errors during extraction. The application supports both recorded and existing audio files for encoding, making it ideal for secure messaging, encrypted communication, and digital watermarking. Developed using Cordova for both web and Android, the system ensures cross-platform compatibility and an intuitive user experience. It plays a crucial role in cybersecurity, intellectual property protection, and covert data transfer, offering a reliable and user-friendly solution for modern digital communication.

Keywords: LSB Encoding, Secure Communication, Private PIN Encryption, Self-Destructive Encryption, Covert Data Transfer, Digital Watermarking, Cybersecurity, Cordova.

Abstract: In today's competitive environment, skill development and access to quality education are crucial for career advancement. Numerous online course recommendation systems have been created in the field of personalized course-learning services to meet the various needs of students. However, the abundance of options of offline classes can also create confusion and make it difficult for students to decide on classes to be taken, especially when considering their career aspirations. However, despite these advancements, there still exist three unsolved challenges: 1) how to effectively recommend the offline classes if students want to search offline classes. 2) how to identify the high-correlated classes in the class corpora. 3) Offline classes recommendation on various platforms may contain fake reviews & ratings. To address this problem, the proposed system introduces an integrated offline classes and jobs recommendation system that utilizes machine learning techniques like knowledge graph, Natural Language Processing, decision tree to suggest relevant courses and employment prospects according to the student's interests, abilities, and professional objectives. An offline course recommendation system uses user skills and class information content, ratings & reviews to suggest personalized classes. Our system uses MERN Stack technology & ML algorithms like Natural Language Processing for review analysis & Knowledge graph to make personalized recommendations. System can authenticate the ratings & reviews given by students to respective classes. The job and course recommendations are the two main components of the system. The course recommendation feature makes recommendations for appropriate classes to the student based on their interests and profile. The job recommendation component helps students find appropriate jobs based on their qualifications, experience, and desired careers. The suggested classes and job suggestion system may help students make well-informed choices regarding their educational and professional pathways. Furthermore, the system can be extended to recommend offline classes with data given by users like location i.e. if users want to search classes within specific areas.

Keywords: Offline Class Recommendation, Job Recommendation System, Natural Language Processing (NLP), Knowledge Graph, Student Interests, Class Ratings and Reviews

Abstract: In the Digital Health landscape, an Enterprise Security Strategy Framework (ESSF) is of paramount importance for Electronic Health Record(EHR) organizations handling Digital Healthcare information. This paper outlines a comprehensive framework to protect the organization's assets, data, and infrastructure from cyber threats. It includes risk assessment, security standards, policies, implementation/monitoring strategies, and audit/assessment procedures. The goal is to establish a framework which incorporates resilient security posture that ensures data protection, regulatory compliance, and business continuity.

Keywords: Enterprise Security Strategy Framework, Electronic Health Records, ESSF, EHR, Risk Assessment, Security Policy, Security Standards, Implementation Strategies, Security Audit, Security Assessments, Security Monitoring, Health Care Security, Digital Healthcare

Abstract: The disease known as cancer is typified by abnormal cell proliferation that spreads throughout the body. In the lungs, abnormal cell growth leads to lung cancer. The lungs, the body's main respiratory control system, ensure that oxygen reaches every part of the body. It purifies the air and prevents infections and unwanted substances from entering the body. According to our immune system, every organ can battle inflammation and infections. Sometimes, though, they fall short in the fight against these infections, inflammations, and even malignant cells. This will inevitably lead to the development of cancer. Stages 0 through 4 are used to classify lung cancer. Early detection of lung cancer, either stage 0 or stage 1, increases a patient's chances of survival. If the cancer is found in its advanced stages, the chances of survival are quite low. Early identification of breast cancer is therefore essential. Many medical diagnostic methods, including X-rays and lung cancer screening, are available for the prediction of lung cancer. However, there are instances in which these diagnostic methods result in false positives or false negatives, requiring patients to get needless medical care. To avoid these outcomes linked to lung cancer projections, alternative approaches are needed. Even while there are other computerised methods for predicting lung cancer, they are also not very accurate. Therefore, using the lung cancer patient databases from Kaggle, we have created three models—Kernel Optimised Neural Network (KONN), Hierarchical Optimisation Neural Network (HONN), and Neural Adaptive Transformer Optimiser classifier method (NATO)—to predict lung cancer in its early stages. Along with the suggested efforts, the dataset is pre-processed using SMOTE to address the issues of class imbalance. Together with the training time for each suggested model, the performance of these methods is evaluated using the following metrics: accuracy, precision, and recall. When compared to the other two suggested models, such as Kernal Optimised Neural Network and Hierarchical Optimisation Neural Network, the Neural Adaptive Transformer Optimiser classifier approach in this research gives better accuracy and requires less training time.

Keywords: HONN, KONN. Lung Cancer, NATO, SMOTE.

Abstract: As artificial intelligence redefines customer engagement in banking, AI-powered chatbots have evolved into intelligent assistants capable of executing secure transactions, providing real-time financial advice, and detecting fraudulent behavior. However, the increased complexity and autonomy of these systems introduce new challenges related to transparency, fairness, inclusivity, and regulatory compliance. This study presents a comprehensive framework for designing next-generation banking chatbots that are not only intelligent and adaptive but also explainable, ethically governed, and accessible to all user demographics. The research explores critical gaps in current chatbot deployments, including the absence of standardized performance metrics, the underutilization of explainable AI (XAI), and the lack of accessibility for users with disabilities. It further examines technical challenges in integrating AI systems with legacy banking infrastructure and highlights opportunities to enhance chatbot trust through voice interfaces, emotional intelligence, and blockchain-secured digital identity. Through a multidisciplinary approach, this study offers actionable design, governance, and deployment strategies to help developers and financial institutions build secure, inclusive, and regulation-ready chatbot ecosystems.

Keywords: Explainable AI, Conversational Banking, Inclusive Chatbot Design, AI Governance in Finance, Sentiment-Aware Systems, Ethical Artificial Intelligence, Legacy System Integration, Blockchain in Digital Identity

Abstract: The Interactive Digital Classroom is a platform designed to bridge the gap between students and teachers by providing a seamless, accessible, and innovative learning experience. As a digital application, it enables users to access the system anytime, ensuring most flexibility in education. This project also offers the curated list of suitable platforms for students to enhance their learning experience. With advancements in technology, smartphones, high-speed internet, and cost-effective data plans, virtual classrooms and online meetings have become an integral part of modern education. However, one common challenge faced during online classes is interruptions caused by connectivity issues. To address this, our project incorporates offline video access, allowing students to revisit lectures even in the absence of an internet connection. The virtual classroom introduces an interactive method of teaching where communication takes place through live video sessions, text chats, feedback mechanisms, online exams, and other digital tools. This approach provides numerous advantages over traditional classroom learning by offering flexibility, accessibility, and improved student engagement. Through this project, we aim to revolutionize the education system by making remote learning more effective and efficient.

Keywords: e-learning, learning, virtual environment, virtual platform, Web-based.

Abstract: Diabetic retinopathy (DR) is one of the leading causes of vision loss globally, particularly among individuals with prolonged diabetes. Early detection and timely intervention are critical to preventing severe vision impairment or blindness caused by this condition. However, traditional diagnostic methods rely heavily on manual examination of retinal fundus images by trained ophthalmologists, which is both time-consuming and resource-intensive. In many underserved or rural areas, access to skilled professionals and diagnostic tools is limited, resulting in delayed or missed diagnoses. The need for an automated, scalable, and accurate system for detecting DR stages is paramount. Such a system can significantly reduce the diagnostic burden on healthcare professionals while ensuring timely identification of at-risk patients. The DenseNet169 model, is fine-tuned for DR detection by adding custom classification layers, including a global average pooling layer, dropout for regularization, and a sigmoid-activated dense layer for multilabel classification. This architecture allows the model to capture intricate patterns in retinal images, crucial for detecting subtle variations in DR severity. By leveraging deep learning technologies like DenseNet169 and integrating them into user-friendly platforms like Flask applications, it becomes possible to democratize access to DR screening, improve diagnostic accuracy, and support healthcare providers in managing the growing burden of diabetes-related eye diseases. This study addresses these challenges by proposing a robust and accessible system for DR detection, bridging the gap between advanced technology and practical healthcare solutions.

Keywords: Diabetic Retinopathy (DR) , Deep Learning , DenseNet169 , Retinal Fundus Images,  Automated Diagnosis

Abstract: The growing emphasis on sustainable aquaculture has driven the need for advanced technological solutions in fish farming. This project introduces a low-cost, IoT-based fish farming management system that leverages Arduino and GSM technologies to enable real-time monitoring and control of key water quality parameters. Equipped with sensors for measuring temperature, pH, turbidity, and dissolved oxygen, the system processes environmental data using an Arduino microcontroller and communicates alerts and updates to the farmer via a GSM module. Additionally, it can automate critical operations such as aeration and water circulation using actuators. By facilitating remote access and data-driven decision-making, the proposed system reduces manual intervention, lowers fish mortality rates, and boosts farm productivity. Its scalability and affordability make it well-suited for both small-scale and commercial aquaculture operations, promoting smarter and more sustainable fish farming practices.

Keywords: IoT, Fish Farming, Sustainable Aquaculture, Arduino, GSM Module, Water Quality Monitoring, Remote Control, Automation, Smart Farming, Sensors, Dissolved Oxygen, pH, Temperature, Turbidity, Actuators.

Abstract: Gas leakage poses a serious safety concern, potentially leading to fire outbreaks, explosions, and severe health complications due to the buildup of toxic gases. Conventional gas detection mechanisms depend largely on manual observation or basic alert systems, which often lack the efficiency for timely action. This project introduces an IoT-based gas leakage detection and prevention system designed for real-time identification and immediate response to hazardous leaks. The system incorporates an MQ6 gas sensor connected to a NodeMCU ESP8266 microcontroller to constantly monitor gas concentration levels. When gas levels exceed a predefined safety threshold, the system triggers a buzzer and LED alerts, shuts off the gas supply via a solenoid valve through user mobile with remote access, and activates a servo motor to open nearby windows for ventilation. With real-time notifications delivered to users, this system offers an intelligent, automated approach to enhance safety and prevent accidents in domestic, commercial, and industrial environments.

Keywords: Gas leakage detection, IoT-based safety system, MQ6 gas sensor, NodeMCU ESP8266, real-time monitoring, automated response, solenoid valve control, servo motor ventilation, remote access, buzzer and LED alerts, toxic gas prevention, fire and explosion prevention, domestic and industrial safety, smart home safety system, hazard mitigation.

Abstract: Over the years, efficient renewable energy sources have become a dire need. Parallel, multilevel inverters have also been gaining traction for their applicability in solar applications as they provide transformer-less voltage control and allow even low voltage PV modules to be fed into the grid. But, PV integrated MLIs have low efficiency and Harmonic Distortion. To help tackle this issue, this project proposes the implementation of a reduced switch cascaded seven-level inverter with a grid-connected PV system. Using methods such as reducing the number of switches, RSCC and MSPWM. Reducing the number of switches enables us to reduce the switching losses as compared to a normal cascaded MLI. However, the reduced switch topology causes voltage division, to tackle this MSPWM is implemented to obtain optimum switching angles to minimize the harmonics and RSCC helps stabilise the voltage. This paper finds that reduced switch topology and MSPWM prove to be a better alternative for PV integrated systems, with higher efficiency and reduced harmonics.

Keywords: Seven level inverter, Photovoltaic system, Grid system, Multicarrier Sinusoidal Pulse Width Modulation (MSPWM), Multilevel inverter (MLI).

Abstract: Machine Learning (ML) plays a vital role in building intelligent systems that automate eligibility assessment, pattern recognition, and personalized service delivery. In public welfare, ML offers scalable solutions to address challenges such as limited awareness, complex eligibility rules, bureaucratic delays, and low digital literacy—particularly in rural and marginalized communities. This paper introduces Scheme Navigator, a progressive web application designed to simplify access to over 150 central and state welfare schemes in India. The system combines rule-based logic with ML algorithms to dynamically match user profiles with relevant schemes, conduct real-time eligibility checks, and provide geolocation-based guidance using GIS tools. Built using Flask, PostgreSQL, Leaflet.js, Flask-Mail, and Twilio, the platform ensures secure authentication, multilingual accessibility, and timely notifications. A mixed-methods evaluation—incorporating surveys, interviews, and system testing—demonstrates enhanced eligibility awareness, usability, and engagement among users. The study highlights the potential of ML-driven platforms to deliver inclusive, efficient, and transparent welfare services.

Keywords: Machine Learning, Welfare Schemes, Eligibility Prediction, Rule-Based Systems, Progressive Web Application, Geolocation Services, Flask, Government Services, Public Welfare, Digital Inclusion, Multilingual Interface, User Engagement

Abstract: Agri Assist, an AI-powered Web application that enhances agricultural decision-making through intelligent chatbot interaction and real-time environmental monitoring. Using Natural Language Processing (NLP), the app allows farmers to ask questions in English text format and receive instant responses in both text and speech formats via an integrated Text-to-Speech (TTS) engine. The chatbot provides guidance on crop selection, soil health, weather forecasts, pest management, and government schemes. Agri Assist is further supported by a hardware kit equipped with sensors that collect real-time data on soil pH, moisture, temperature, and humidity. This sensor data is processed to deliver personalized, data-driven farming recommendations. The system is designed with inclusivity in mind, intuitive voice-based interaction, and a user-friendly interface tailored for rural users. Agri Assist empowers farmers with accessible and accurate agricultural insights, promoting smart farming practices and sustainable productivity across diverse environments.

Keywords: Agricultural Technology, Web Application, AI Chatbot, Smart Farming, Text-to-Speech (TTS), Natural Language Processing (NLP), Precision Agriculture, Soil Monitoring, Farmer Assistance, Crop Recommendation System, IoT in Agriculture, Environmental Sensors, Sustainable Farming, Digital Farming Platform

Abstract: An era of rapid technological advancements, the demand for skilled professionals in various industries is growing significantly. However, traditional vocational training methods face challenges such as limited hands-on experience, high equipment costs, safety concerns, and a gap between academic learning and industry requirements. FutureCraft leverages Augmented Reality (AR) and Virtual Reality (VR) to provide immersive, interactive, and accessible training experiences that bridge the skills gap and modernize vocational education. This project aims to develop VR and AR-based tools that offer industry-relevant, hands-on training modules. Through realistic virtual simulations, students can engage with complex industrial processes in a risk-free environment, enhancing their technical and practical knowledge.

Keywords: Virtual Reality (VR), Augmented Reality (AR), Vocational Education, Immersive Learning, Adaptive Training, Industry-Relevant Skills.

Abstract: This research proposes an assistive platform aimed at improving educational accessibility for students with hearing and speech impairments. By leveraging real-time sign language translation and expressive human gestures such as lip movement and facial expressions, the system ensures a more immersive and inclusive learning experience. The Web Speech API is used to transcribe spoken language into text, which is then mapped to sign language actions performed by skilled human signers. This approach not only increases communication accuracy but also mimics natural human interaction, making content more relatable and easier to understand. The platform supports multiple sign languages and is designed for seamless integration into both online and offline educational environments. With a customizable user interface and focus on real-time responsiveness, this solution bridges communication gaps and promotes equitable participation for students with disabilities.

Keywords: Inclusive education, Hearing impairment, Speech impairment, Real-time sign language, Accessibility

Abstract: This study proposes a deep learning-based approach to classify liver histopathological images into four categories: ballooning, fibrosis, inflammation, and steatosis, using the VGG16 convolutional neural network (CNN). The VGG16 model, pre-trained on ImageNet and fine-tuned on a liver disease dataset, is used for feature extraction and classification. Data augmentation techniques address challenges of limited medical images. The model is evaluated using precision, recall, F1-score, and accuracy metrics. This approach demonstrates the potential of deep learning to support pathologists in diagnosing liver diseases, offering a reliable and automated tool for healthcare professionals.

Abstract: This project focuses on building a smart and sustainable agricultural system that assists farmers in reducing the overuse of chemical fertilizers and pesticides. Excessive chemical application has long-term negative effects on soil fertility, crop yield, and the surrounding ecosystem. To address these issues, the system utilizes Artificial Intelligence (AI) techniques to provide personalized recommendations for suitable crops, fertilizers, and pesticides based on the specific soil condition and plant health data. The application allows farmers to register using their basic details and input soil data either manually (by entering nutrient values such as Nitrogen, Phosphorus, Potassium, pH level, and moisture) or by uploading images. Based on this information, the AI model predicts the most suitable crop for cultivation. Additionally, the system features a disease detection module where farmers can upload images of infected leaves. Using machine learning and image processing techniques, the system diagnoses the disease and suggests appropriate remedies, including the type and quantity of fertilizers or pesticides required.

Keywords: Artificial Intelligence, Sustainable Agriculture, Crop Recommendation, Soil Analysis, Plant Disease Detection, Machine Learning, Fertilizer Suggestion, Pesticide Management, Smart Farming, Image Classification, Precision Agriculture, IoT Integration, Environmental Protection, Farmer Support, Agricultural Technology, Leaf Image Analysis, Soil Nutrient Detection, Decision Support System, Cloud-Based Farming, Multilingual Interface, Mobile Agriculture App, Digital Farming, Agricultural Data Analytics, Smart Irrigation, Deep Learning, Computer Vision, Crop Yield Optimization, Remote Sensing, Farm Management System, Agricultural Sustainability

Abstract: The Crowdsourcing Web Application for Temporary Daily Work is a platform designed to connect individuals seeking short- term, daily tasks with freelancers or workers available to take on temporary employment opportunities. This web application allows users to post or browse available tasks such as delivery, cleaning, event assistance, data entry, or other short-term projects. The system is built to facilitate quick job assignments and offer real-time communication between job posters and workers. By leveraging the power of crowdsourcing, this platform ensures a dynamic and flexible workforce that can efficiently respond to immediate and temporary work requirements. The application also includes features like rating and reviews, secure payment processing, and task tracking to ensure quality and transparency for both employers and employees.

Keywords: Crowd-Sourcing, Temporary Work, Freelancers, Gig Economy, Task Management.

Abstract: The increasing need for transparent and efficient carbon credit markets has led to the exploration of Artificial Intelligence (AI) to address fraud, double-counting, and inefficiency in traditional carbon credit trading systems. This paper proposes an AI-based carbon credit trading system that leverages machine learning and predictive analytics to provide a transparent, secure, and automated marketplace for buying and selling carbon credits. Through the use of AI-driven smart algorithms, the system automates the issuance, verification, and trading of carbon credits, reducing transaction costs and eliminating intermediaries. The system ensures that each carbon credit represents a verified and actual reduction in emissions by employing AI-powered data validation and anomaly detection, offering real-time tracking and traceability. Furthermore, the advanced analytical capabilities of AI promote inclusivity, enabling a broader range of participants to engage in emission reduction efforts. This system not only fosters environmental accountability but also drives global sustainability by providing an accessible, cost-effective, and scalable solution to meet emission reduction targets and combat climate change.

Keywords: Artificial Intelligence, Carbon Credit Trading, Smart Contracts, Transparency, Machine Learning.

Abstract: This project introduces a smart web-based healthcare assistant focused on predicting Chronic Kidney Disease (CKD) and Cardiovascular Disease (CVD) using machine learning models. The system fetches the user's IP-based location to suggest nearby hospitals and allows users to book appointments seamlessly. It ensures user data security by implementing SHA-256 encryption. Designed with user convenience and early diagnosis in mind, the platform bridges the gap between patients and healthcare providers through intelligent recommendations and secure interactions.

Keywords: Heart Disease Prediction, Kidney Disease Prediction, Machine Learning, Hospital Recommendation, User Inputs, Personalized Care, Platform Independence, Data Security

Abstract: Internet of Things (IoT) is transforming environmental monitoring by enabling real-time, remote, and automated data collection across diverse settings. It introduces an IoT-based air quality monitoring system designed for early detection and proactive management of air pollution. The system integrates low-cost gas sensors (MQ2, MQ7, MQ135) with a NodeMCU ESP8266 microcontroller to measure harmful gases such as CO, CO₂, and smoke. Sensor data is transmitted wirelessly to the ThingSpeak cloud for real-time visualization and analysis. Automated air purification is triggered when pollutant levels exceed safety thresholds, reducing the need for manual intervention. The system also incorporates machine learning models to predict pollution trends, enabling smarter, data-driven responses to environmental changes. Developed using Python, ThingSpeak, and TensorFlow, the solution achieves high accuracy in pollutant detection and fast response times. Designed for deployment in urban areas, hospitals, and schools, the project highlights the role of IoT in building smarter, healthier environments. Future enhancements include AI-based forecasting, multi-sensor expansion, and solar-powered deployment for improved scalability and sustainability. The proposed system addresses limitations of traditional air monitoring by offering a cost-effective, scalable alternative. Its modular design allows easy integration with smart city infrastructure. This project exemplifies how IoT can support environmental sustainability and public health initiatives

Abstract: In today’s digital era, password security is a major concern as cyber threats are becoming more sophisticated. The Enhanced Password Generator Using Cloud Computing provides a secure, scalable, and efficient solution for generating and managing strong passwords. This project utilizes AI-driven password generation, secure cloud storage, multi-factor authentication (MFA), and real-time breach detection to ensure data safety. By integrating AES-256 and RSA encryption, role-based access control, and user-friendly interfaces, the system enhances password security and accessibility across devices. Cloud infrastructure ensures seamless synchronization, providing a modern solution for both individual and enterprise-level password management.

Keywords: Password Generator, Cloud Computing, Encryption, Multi-Factor Authentication, Cybersecurity.

Abstract: In the context of airport logistics, "turnaround time" refers to the interval between an aircraft's landing and its subsequent takeoff. Unfortunately, inefficiencies within these turnaround operations are a major factor behind flight delays. To achieve optimal profitability, airlines must strive to minimize the duration an aircraft remains grounded. Nevertheless, this objective is hindered by the necessity to comply with manufacturer-mandated maintenance procedures, which are vital to ensuring aircraft safety. These activities, outlined in detailed checklists and scheduled by the manufacturer, owner, or operator—under the oversight of certified airworthiness authorities—create significant constraints in reducing on-ground time. Consequently, streamlining turnaround procedures remains the only controllable aspect through which airlines can improve efficiency and profitability. As air travel serves as a cornerstone of global connectivity, maintaining strict standards for safety and security is indispensable. However, the COVID-19 pandemic has deeply impacted ground handling protocols, prompting the urgent need to revise traditional practices to align with enhanced hygiene and health regulations. One prominent challenge lies in the passenger embarkation process, which now requires strict physical distancing and thorough disinfection of the cabin after each flight. In response, this study explores potential revisions to in-cabin procedures by comparing them to pre-pandemic turnaround operations. Through a detailed, process-level examination, we identify individual touchpoints and suggest strategic adjustments aimed at improving operational efficiency. Our findings indicate that boarding durations have increased significantly—more than twice the usual time—due to social distancing mandates. Despite introducing various procedural changes, sustaining previous turnaround benchmarks while maintaining full passenger capacity remains problematic. Nevertheless, adopting alternative strategies—such as maintaining vacant middle seats (reducing capacity to approximately 67%) and boarding from apron stands using both front and rear doors—can help mitigate delays and support smoother aircraft turnaround operations.

Keywords: Aircraft, Turnaround Operations, Air Travel, Turnaround times, Cabin, Post-Pandemic World

Abstract: Deep learning has greatly enhanced computer vision by enabling models to extract complex features from large datasets. Utilizing Convolutional Neural Networks (CNNs) and modern architectures, significant progress has been made in object detection and human attribute recognition. This paper presents a real-time Flask-based web system that detects persons using YOLOv8, estimates their age and gender via pre-trained Caffe models, identifies clothing color through K-Means clustering, and calculates distance in steps using geometric estimation based on object height. The system processes live webcam video streams and provides verbal feedback through a text-to-speech engine, enhancing accessibility for visually impaired users. By integrating computer vision with audio feedback, the solution offers a practical and intelligent assistant for real-world scenarios. The system achieves reliable performance with an overall accuracy of 94.44%.

Keywords: Deep Learning, Computer Vision, YOLOv8, Flask, Face Detection, Age and Gender Prediction, Clothing Color Detection, Distance Estimation, Text-to-Speech, Accessibility, Real-Time System.

Abstract: The DaanSethu Donation App is a comprehensive digital platform designed to reduce resource wastage and address unmet community needs through sustainable redistribution. It allows for donations of food, clothing, toys, bedding, blood, and monetary contributions, promoting responsible consumption. Built using React Native for cross-platform compatibility and Firebase for secure backend operations, the app features a real-time geolocation-based matching system that minimizes logistical emissions and ensures efficient distribution. By reducing waste and supporting circular economy practices, DaanSethu not only strengthens community engagement but also contributes to environmental preservation. The platform's volunteer coordination tools, secure transactions, and data analytics further enhance its scalability and potential to transform community support systems sustainably.

Keywords: Resource Redistribution, Circular Economy, Sustainability, Firebase, React Native, Geolocation Matching, Community Engagement, Waste Reduction, Secure Transactions, Volunteer Coordination.

Abstract: The backend architecture of an e-commerce platform plays a vital role in ensuring the robustness, scalability, and personalized user experience of the digital shopping environment. With modern consumers demanding real-time engagement, seamless transactions, and tailored product recommendations, the backend must evolve to manage high volumes of concurrent users, secure data flows, and intelligent decision-making processes. This paper offers a comprehensive theoretical review of backend systems for e-commerce, focusing on layered architecture, data modeling, distributed computing, recommendation engines, and security frameworks. Drawing upon advanced algorithms and backend technologies like Node.js and MongoDB, the paper details how theoretical constructs translate into practical capabilities that drive performance and consumer satisfaction in competitive online markets. In the expanding realm of digital commerce, the backend of an e-commerce platform serves as the computational and operational backbone that orchestrates dynamic user interactions, product management, and secure transactions. The abstracted functionality of a well-designed backend is not merely about data persistence or user authentication but encompasses a range of interrelated services that drive the intelligent behavior of the platform. From enabling asynchronous operations that facilitate real-time updates to integrating advanced analytics engines that interpret user behavior, the backend lays the groundwork for personalized and scalable online retail.

Keywords: Backend Architecture, RESTful APIs, Microservices, Database Design, Caching, Security,Scalability.

Abstract: Smart farming, enabled by the integration of Remote Sensing (RS) and Internet of Things (IoT) technologies, represents a revolutionary approach to modern agriculture. This paper explores the synergistic application of RS and IoT for real-time monitoring of irrigation practices in smart farming systems. Remote sensing provides a bird's-eye view of agricultural landscapes, capturing crucial data on crop health, soil moisture, and environmental conditions. Concurrently, IoT devices deployed in the field offer ground-level data, including real-time sensor readings and feedback from automated irrigation systems. The integration of these two technologies facilitates a comprehensive and dynamic understanding of the agricultural environment. This paper discusses the challenges, methodologies, and benefits of integrating RS and IoT in the context of irrigation management. The proposed framework enhances precision agriculture by enabling timely decision-making, optimizing water usage, and improving overall farm productivity.

Abstract: Artificial Intelligence (AI), which involves the simulation of human intelligence processes by machines, is increasingly being applied across various domains, with one of its most impactful uses in the medical field. It is revolutionizing healthcare by enabling faster, more accurate diagnoses and improving patient outcomes through data-driven decision-making. The proposed system presents an AI-based automated diagnostic framework for the early detection of melanoma (skin cancer) and diabetes—two of the most prevalent and critical health conditions. Leveraging deep learning and image processing techniques, the system enhances diagnostic precision and efficiency. Convolutional Neural Networks (CNNs), a core AI method in medical imaging, are utilized for image-based disease classification. For melanoma detection, dermoscopic images are analyzed using pre-trained CNN models to identify cancerous patterns. For diabetes, retinal image analysis is integrated with clinical parameters to assess disease risk. This AI-powered system automates feature extraction, reduces the need for human intervention, and provides real-time, accurate diagnostic results. Developed using MATLAB, the framework shows high classification accuracy and robustness under varying image conditions. This project underscores the expanding role of AI in healthcare and aims to make intelligent diagnostics more accessible to medical professionals. Future work will focus on expanding datasets, enhancing model generalization, and integrating additional clinical features for more comprehensive health assessments.

Abstract: This paper examines the critical factors influencing the structural stability of high-rise buildings. Through a comprehensive literature review, detailed methodology, and insightful case study, we delve into the complexities of ensuring these towering structures remain safe and resilient. Our analysis considers various load-bearing mechanisms, material properties, and environmental factors, providing a holistic perspective on high-rise building stability.

Keywords: High-rise buildings, structural stability, load-bearing, material properties, environmental factors, finite element analysis, wind load, seismic design

Abstract: The spread of fake news impacts public perception and decision-making. Traditional machine learning models lack contextual understanding and interpretability. We propose a deep learning approach using FastText for text representation and Explainable AI (XAI) for transparency. FastText captures word and subword information, improving fake news detection. Deep learning models like LSTMs or CNNs enhance classification accuracy. To address the "black box" issue, we integrate XAI techniques such as SHAP and LIME. These methods highlight key words influencing predictions, aiding journalists and fact-checkers. Experimental results on benchmark datasets show superior accuracy and interpretability. FastText ensures efficient feature extraction, while XAI enhances trust. Our approach provides a scalable, ethical, and effective solution for misinformation detection.

Keywords: FastText, LSTM, decision-making, black box.

Abstract: The Convolutional Neural Network (CNN) and Recurrent Neural Network (RNN) architectures represent significant advancements in deep learning, particularly in image recognition and sequential data processing. Traditional drowsiness detection methods primarily rely on biometric measurements such as heart rate, pulse waves, brain waves, and eye movements. . By analyzing real-time visual data from a driver’s face and eyes, the system can detect subtle signs of fatigue, such as changes in eyelid movement, eye closure rates, and facial expressions. Additionally, the system provides real-time voice alerts upon detecting signs of drowsiness, ensuring immediate intervention and enhancing driver safety. The integration of CNN and RNN thus offers a highly efficient, real-time, and scalable solution for preventing fatigue-related accidents on the road.

Abstract: The increasing sophistication of cyber threats targeting Industrial Control Systems (ICS) and Supervisory Control and Data Acquisition (SCADA) networks necessitates an advanced threat detection framework. This field focuses on developing a Quantum-Adaptive Threat Detection (QATD) model to enhance cybersecurity resilience, improve detection accuracy, and minimize false positives. Utilizing a dataset comprising real-world ICS/SCADA threat incidents, the system implements quantum-inspired anomaly detection techniques and graph-based threat correlation to identify malicious activities in real time. The QATD model is benchmarked against conventional detection systems, including signature-based Intrusion Detection Systems (IDS), anomaly-based AI models, and machine learning classifiers, using performance metrics such as Detection Accuracy, False Positive Rate (FPR), Precision, and Response Time Efficiency. The system integrates Quantum Graph-Based Threat Correlation (QGTC) and Quantum-Optimized Attack Response (QOAR) mechanisms, significantly improving attack pattern recognition and automated mitigation strategies. The proposed system achieves over 90% accuracy in zero-day attack detection, reduces false positives by 40%, and enhances response efficiency by 50% compared to traditional AI-based cybersecurity solutions.

Keywords: ICS Security, SCADA Threat Detection, Quantum-Adaptive Threat Detection (QATD), Cybersecurity Analytics, AI-Driven Threat Mitigation, Zero-Day Attack Detection.

Abstract: Automated Timetable Generators (ATGs) have emerged as essential tools in educational institutions to simplify scheduling and resource allocation. These systems replace manual planning with intelligent, algorithm-driven scheduling that minimizes human error and resource conflicts. This paper surveys the current technologies, algorithms, and challenges in automated timetable systems. It explores various approaches, from rule-based systems to AI-powered models, and identifies gaps in existing solutions such as scalability, user flexibility, and real-time conflict resolution. The survey aims to guide future enhancements for adaptive, scalable, and institution-friendly timetable solutions.

Keywords: Timetable Generator, Conflict Detection, Scheduling Algorithms, Educational Technology, Automation.

Abstract: Ensuring the safety and mental well-being of children and women in the digital era requires innovative solutions. This paper explores the development of an online platform that integrates virtual counseling services, gamified activities, legal rights awareness, and essential resources. By creating a secure and engaging virtual environment, the platform aims to empower users with interactive tools, mental health support, and knowledge to enhance their overall well-being. The study assesses the needs, concerns, and preferences of the target users and provides recommendations for effective implementation, prioritizing security, accessibility, and inclusivity.

Abstract: A chatbot is a computer program that makes conversations with humans using Artificial Intelligence (AI) in messaging platforms. Every time the chatbot gets input from the user, it saves input and response which helps the chatbot with little initial knowledge to evolve using gathered responses. We can implement an online chatbot system to assist website users. Using this tool, we can access files easily instead of going through different modules. Artificial Intelligence methods such as Natural Language Processing (NLP). In order to achieve quality education as a defined one of the sustainable goals, it is necessary to provide information about the education system according to the stakeholders’ requirements. The process to obtain the information about university/institute is a critical stage in the academic journey of prospective Students who are seeking information about the specific courses which makes that university/institute unique. This process begins with exploration to general information about universities through websites, rankings,and brochures from various sources. Most of the time, information available on different sources leads to discrepancies and influences student’s decisions. By addressing inquiries promptly and providing valuable information, universities can guide individuals in making informed choices about their academic future.To address this, the chatbot application is the most effective tool to be implemented and make it functional on university’s functional website. A chatbot is an artificially intelligent tool which can interact with humans and can mimic a conversation. This tool can be implemented using advanced Natural Language Processing (NLP) models to provide the pre-defined answers to the student’s queries. Chatbot is very helpful for query resolution during the counseling process of the institute as it will provide official/uniform information and can be accessed 24 × 7. Therefore, the aim of this research work was to implement a chatbot using various NLP models and compare them to identify best one. .

Keywords: AI bot, Virtual assistant, Customer support bot, Auto reply, Live chat, Message bot, Online assistant

Abstract: The growing number of malicious applications on the Android platform has raised significant security concerns, necessitating the development of effective detection mechanisms. This project focuses on categorizing Android applications based on their potential malicious behavior using machine learning techniques, specifically Artificial Neural Networks (ANN) and Support Vector Machines (SVM). A dataset consisting of 410 unique permission combinations, sourced from Kaggle, serves as the foundation for identifying key features that distinguish between benign and harmful applications. The project employs two classification models: ANN, which leverages deep learning to analyze complex permission patterns, and SVM, a traditional yet highly effective machine learning algorithm known for its precise decision boundaries. Additionally, the system includes a real-time malware detection web application built using Flask. Users can upload APK files, after which the system extracts permission-based features and applies the trained models to determine whether the application is benign or malicious. By integrating ANN and SVM, this project highlights the effectiveness of permission-based machine learning models in Android malware detection. The proposed approach strengthens mobile cybersecurity by demonstrating how advanced machine learning techniques can be utilized to combat modern security threats.

Keywords: Android security, malware detection, machine learning, Artificial Neural Networks (ANN), Support Vector Machines (SVM), APK permissions, deep learning, cybersecurity, Flask web application, mobile threat analysis.

Abstract: The increasing adoption of online education has led to a growing need for secure and reliable examination systems. Traditional in-person examinations ensure integrity through physical invigilation, but with the rise of remote learning, maintaining exam security has become a significant challenge. Online exams are often vulnerable to cheating, impersonation, and unauthorized assistance, which undermines the credibility of assessments. Existing online examination systems employ manual proctor- ing, where human invigilators monitor students via webcams. However, this approach is labor-intensive, expensive, and sub- ject to human errors. To address these challenges, automated proctoring solutions have been introduced, integrating artificial intelligence (AI) and machine learning to detect suspicious behavior. In this research, we present Proctor, an AI-based online proctoring system designed to ensure exam integrity through automated monitoring and real-time analysis. The system uses computer vision techniques to detect face absence, multiple peo- ple in the frame, mobile phone usage, and unauthorized objects. Additionally, it prevents tab switching, keyboard shortcuts, and right-click actions, further securing the examination process. Proctor is designed to be efficient, scalable, and user-friendly, making online assessments more reliable without requiring hu- man invigilators. By leveraging advanced AI models, the system enhances security while reducing the administrative burden on educators. This research explores the need, implementation, and effectiveness of AI-based proctoring, offering a novel approach to securing online exams.

Abstract: Automated detection of epileptic seizures from EEG recordings is critical for patient monitoring and early intervention. We propose a hybrid Convolutional Neural Network (CNN)–Long Short-Term Memory (LSTM) architecture that ingests eight bipolar EEG channels (C3–P3, P3–O1, P4–O2, P7–O1, P7–T7, T8–P8-0, T8–P8-1, and FP1–F7) to detect seizure events. On the CHB-MIT scalp-EEG dataset (49 999 samples), our model achieves 97.92 % test accuracy, 93.29 % precision, 85.33 % recall, 89.14 % F1-score, and an AUC of 0.9807. Cross-validation yields comparable metrics. We also deliver an interactive Streamlit web app for real-time inference. Index Terms: EEG, seizure detection, deep learning, CNN, LSTM, CHB-MIT dataset, Streamlit

Abstract: Social Media Sentiment Analysis (SMSA) has become an essential tool for organizations aiming to understand public perception through digital expressions. This study investigates advanced sentiment analysis techniques using Natural Language Processing (NLP) and Machine Learning (ML) models from 2020 to 2025. It highlights real-time applications, modeling strategies, and the efficacy of various sentiment classification algorithms. The paper also discusses the impact of deep learning and transformer-based models in improving accuracy and reliability. Our analysis confirms the growing relevance of SMSA in strategic decision-making across sectors including marketing, politics, and crisis management.

Abstract: In the age of rapid digital transformation, Cybersecurity remains a cornerstone for sustaining trust in modern technologies. As organizations adopt cutting-edge innovations such as artificial intelligence (AI), cloud computing, the Internet of Things (IoT), and blockchain, they are exposed to increasingly complex cybersecurity threats. This paper explores the core cybersecurity challenges faced in today's tech-driven world and examines the emerging trends and solutions aimed at mitigating those threats. The study also highlights the role of adaptive security models, regulatory frameworks, and collaborative global efforts in reinforcing cyber resilience.

Keywords: Cybercrime, Cybersecurity, AI Security, Cyber Threats, Blockchain Security, Cyber Regulations.

Abstract: The OralCareX is an innovative AI-powered diagnostic tool designed to enhance dental diagnostics through the application of Convolutional Neural Networks (CNNs). This project addresses the significant challenges faced in dental healthcare, particularly in underserved regions where access to professional dental care is limited. With oral diseases affecting billions globally, the need for early detection and accurate diagnostics is critical. OralCareX leverages advanced AI technology to analyse dental X-rays and images, enabling the identification of conditions such as cavities, gum diseases, and fractures with greater accuracy and speed than traditional methods. The system aims to empower visually impaired individuals by providing accessible and user-friendly diagnostics, promoting proactive oral health management. By integrating AI into dental practices, OralCareX improves the accuracy of diagnostics while simultaneously improves healthcare accessibility, ultimately contributing to enhance patient well-being and overall quality of life.

Keywords: AI, dental diagnostics, Convolutional Neural Networks, OralCareX, healthcare accessibility, early detection, oral health management, visually impaired.

Abstract: Automating the conversion of chess literature into digital formats has become essential for enhancing training, analysis, and archiving. This work presents an intelligent system that automates the process of converting chess puzzles and games from PDF documents into PGN (Portable Game Notation) files. The approach begins with extracting high-resolution images from PDF pages, followed by image processing techniques such as adaptive thresholding and perspective transformation to detect and align chess boards. A Convolutional Neural Network (CNN) is then employed to recognize chess pieces on the board and generate accurate FEN (Forsyth-Edwards Notation) strings. These FEN strings are subsequently converted into PGN format, capturing not only board positions but also move sequences and annotations. The system significantly reduces the manual effort and has 98% accuracy, time required for transcribing chess games from books, offering a scalable and efficient solution for converting both classical and puzzle-based chess content into usable digital formats for modern chess databases and training applications..

Keywords: Chessboard recognition, PGN conversion, FEN, CNN, OCR, Tkinter GUI, image processing

Abstract: Conventional fan speed control systems are typically limited by manual adjustments or simplistic on/off switching based on basic temperature thresholds. This project falls within the domain of intelligent embedded systems for environmental control, specifically focusing on automated fan speed regulation using temperature sensors. It utilizes a DHT11 sensor for real-time temperature monitoring, an Arduino microcontroller for data processing, and an LCD display to provide continuous user feedback. Traditional fan control systems typically rely on manual adjustments or simple on/off switching based on fixed temperature thresholds, leading to energy inefficiency and inconsistent comfort. To address these issues, the proposed system dynamically adjusts fan speed using Pulse Width Modulation (PWM), allowing smooth and proportional control in response to ambient temperature changes. This automated approach eliminates manual intervention, enhances energy efficiency, and improves user comfort. The system has demonstrated an accuracy of 90.47%, marking a significant advancement in intelligent climate control for domestic and small-scale industrial applications, and laying the groundwork for future IoT-based home automation solutions.

Keywords: Automated Fan Control, Temperature Sensor (DHT11), Arduino, Pulse Width Modulation (PWM), Energy Efficiency

Abstract: Large Language Models (LLMs) have significantly advanced natural language processing tasks but are heavily reliant on well-crafted prompts for optimal performance. However, manual prompt engineering is time-consuming, often sub-optimal, and lacks robustness to various perturbations in input prompts. Additionally, there is a growing need to address the security implications of prompt extraction attacks. Existing research has proposed automated methods for prompt engineering, ranging from rewriting under-optimized prompts to generating high-quality human-like prompts from scratch. Despite these advancements, challenges persist in achieving prompt effectiveness, robustness, and security.

Keywords: NLP, LLM, Prompt Engineering, Prompt Recovery, Zero shots, Few shots, Chain of Thoughts.

Abstract: Visual impairment continues to be a significant challenge worldwide, affecting over two billion individuals and often restricting their ability to navigate and interact with their environment independently. Traditional aids such as white canes and guide dogs, while invaluable, do not provide contextual awareness or access to dynamic information like printed text or sudden obstacles. To address these limitations, we present EyeGuide AI, an innovative assistive technology system that leverages the power of artificial intelligence and computer vision to deliver real-time object detection, text recognition, and emergency alert capabilities. Developed using Python, YOLOv8, and Microsoft Azure Cognitive Services, EyeGuide AI seamlessly integrates local and cloud-based AI functionalities to ensure robust and reliable performance. The system is designed to be cost-effective and compatible with widely available hardware, making it accessible to a broad user base. Our experiments demonstrate that EyeGuide AI significantly enhances navigation, reading accuracy, and emergency preparedness for visually impaired users. The system’s modular design, real-time feedback, and focus on user accessibility highlight its potential to redefine the landscape of inclusive assistive technologies.

Abstract: This project, Bconnect, is a web-based platform developed using the MERN stack (MongoDB, Express.js, React.js, and Node.js) aimed at facilitating seamless communication and collaboration within an academic or organizational environment. BConnect provides users with features such as real-time messaging, group discussions, announcements, and file sharing, making it an ideal solution for enhancing productivity and connectivity among students, faculty, and staff.

Abstract: The conversion of 2D images to 3D models has become a significant area of research in computer vision and graphics. This project explores the development of a web-based system that leverages the Flask framework for backend processing to convert 2D images into 3D representations. The core objective of the project is to implement a simple yet effective pipeline that processes input 2D images and generates a 3D model through various computer vision algorithms and machine learning techniques. Using Flask, a lightweight web framework for Python, the system receives 2D images from users, processes them through pre-trained models or algorithms, and then outputs a 3D model or visualization. The 3D model is constructed by inferring depth, texture, and geometric properties from the 2D image. This model can be further visualized in the browser using WebGL or exported into standard formats like STL or OBJ for use in 3D printing or digital modeling applications. The project aims to demonstrate the potential of combining web technologies with advanced image processing techniques to create accessible tools for 3D model generation from basic 2D inputs. This could be applied in various fields, including digital design, augmented reality, and game development, offering a convenient and scalable solution for converting 2D images into 3D assets.

Keywords: 2D to 3D conversion, computer vision, 3D reconstruction, Flask framework, machine learning, image processing, WebGL visualization, depth estimation, STL/OBJ export, web-based system, digital modeling, 3D printing, augmented reality, game development, geometric inference.

Abstract: The increasing volume of digital music content has led to a growing demand for personalized music recommendation systems that can understand and cater to individual preferences. This paper proposes an emotion-based music recommendation system leveraging machine learning techniques and implemented using Python technology. The system aims to enhance user satisfaction and engagement by recommending music tracks based on emotional context, providing a more immersive and personalized listening experience. Key components of the system include a robust data preprocessing pipeline, feature extraction from audio signals, and the development of machine learning models trained on emotion-labeled datasets. Python libraries such as Pandas, NumPy, and Scikit-Learn are utilized for data manipulation, feature extraction, and model training. The system employs state-of-the-art machine learning algorithms, such as deep neural networks, to extract high-level emotional features from audio data. Evaluation of the proposed system involves assessing its recommendation accuracy, user satisfaction, and the system's ability to adapt to dynamic changes in user preferences and emotional states. The results are obtained through user studies and objective metrics, demonstrating the effectiveness and efficiency of the implemented emotion-based music recommendation system.

Keywords: Machine Learning, Python, Emotion Recognition, Music Suggestions.

Abstract: Ensuring road safety through timely detection of traffic accidents is critical for reducing fatalities and improving emergency response times. This project presents a novel system that integrates YOLOv8, a state- of-the-art object detection algorithm, with OpenCV for real-time accident detection on roadways. The system processes live video streams, identifying critical events such as vehicle collisions or abnormal driving behavior. YOLOv8 enables precise and rapid detection of vehicles and pedestrians, while OpenCV enhances image preprocessing and motion analysis. These components are deployed within a Django web framework, providing an interactive interface for monitoring and alerting authorities. By automating the detection process, the solution minimizes human dependency, accelerates response coordination, and contributes to safer traffic environments. This AI-powered approach not only improves detection accuracy but also supports integration into existing traffic management infrastructures, offering a scalable solution for smart city applications.

Keywords: Accident Detection, Road Safety, YOLOv8, OpenCV, Real-Time Object Detection, Deep Learning, Traffic Monitoring, Django Framework, Computer Vision, Emergency Alert System

Abstract: Deep learning has been widely used in medical image processing, which has sparked the development of a wide range of applications and led to a notable increase in the number of therapeutic and diagnostic options available for a range of medical imaging problems. In the era of the Internet of Things (IoT), safeguarding the security and privacy of medical data is crucial to the advancement of sophisticated diagnostic applications for medical imaging. Deep learning-based brain tumor detection in smart health care systems with privacy preservation is proposed in this paper. The system under consideration is organized into three discrete stages that are then combined to provide an all- encompassing blueprint. During the first phase, patients with brain tumors are the primary target of an efficient healthcare system that is introduced. A Microsoft-based operating system-compatible application has been developed to accomplish this. Patient data is secure and only available to the hospital and the individual patient, which enables patients to engage with the system both locally and virtually. To obtain the anticipated outcomes, the user must first submit the patient’s MRI scan and then enter a special 10-digit code. In the second part, the authors develop a deep learning-based tumor identification platform which also incorporates the AES-128 algorithms and PBKDF2 for secure medical image storage on the server and data transmission via the internet from the client to the server and back to the client upon prediction..

Keywords: Brain tumor detection, classification, CNN, cryptography, deep learning algorithms, MRI, privacy preservation, smart healthcare systems.

Abstract: Accurate and timely diagnosis of gastrointestinal (GI) diseases is essential for effective treatment and improved patient outcomes. Endoscopy is a key diagnostic tool that provides direct visualization of the GI tract, but manual interpretation of endoscopic images is subject to human error, fatigue, and inter-observer variability. To address these challenges, this research explores the application of deep learning techniques for automated disease identification using endoscopic images. Leveraging convolutional neural networks (CNNs), the proposed approach aims to classify and detect abnormalities such as ulcers, polyps, and early-stage cancers with high accuracy. The model is trained and validated on a diverse dataset of annotated endoscopic images to ensure robustness and generalization. Experimental results demonstrate the effectiveness of the deep learning framework in enhancing diagnostic precision, reducing workload for clinicians, and supporting real-time decision-making in clinical settings. This study highlights the potential of AI-driven tools in transforming endoscopic diagnostics and improving the quality of healthcare delivery.

Abstract: Early and accurate identification of gastrointestinal (GI) diseases is critical for effective treatment and improved patient outcomes. Endoscopy provides high-resolution images of the GI tract, but manual interpretation is time-consuming and prone to human error. This study presents an automated approach for disease identification from endoscopy images using deep learning techniques. A convolutional neural network (CNN) model is trained on a labeled dataset of endoscopic images to classify various gastrointestinal conditions such as ulcers, polyps, esophagitis, and bleeding. The system incorporates image preprocessing, data augmentation, and model optimization to enhance detection accuracy. Experimental results demonstrate the model’s ability to achieve high classification accuracy, offering a reliable tool to assist clinicians in diagnostic decision-making. This approach has the potential to improve diagnostic efficiency, reduce workload on medical professionals, and enable scalable screening in resource-limited settings.

Abstract: Drones are increasingly being utilized for recreational purposes and across various fields such as engineering, disaster response, logistics, and airport security. However, their potential misuse has raised serious concerns regarding the safety and surveillance of critical infrastructures, particularly in airport environments. Incidents involving unauthorized drone activity have frequently disrupted airline operations in recent years. To mitigate this issue, this study proposes a novel deep learning-based approach for drone detection and recognition. The method demonstrates superior performance compared to existing systems by accurately identifying the presence of drones, distinguishing between two drone types, and differentiating them from birds, despite the visual and behavioral similarities that often confuse. This advancement significantly enhances aerial object classification and reinforces airspace security. Key Words: drone; UAV; deep learning; convolutional neural network CNN; drone image dataset; drone detection; drone recognition.

Abstract: By implementing a comprehensive school management system, educational institutions can greatly enhance their operational efficiency, resulting in improved educational outcomes and increased stakeholder satisfaction. We will also discuss the significance of establishing strong partnerships with local businesses and organizations and aligning academic programs with industry requirements. The session will showcase effective collaboration models, discuss successful outreach strategies, and underscore the importance of networking events in facilitating connections between students and potential employers. Our objective is to strengthen the connection between education and ultimately enhancing student preparedness for the job market and improving employment rates.

Keywords: services, platform, system, management, education, students, work.

Abstract: This paper presents an intelligent system for plant identification through the analysis of leaf images, utilizing image processing and machine learning techniques. The model extracts key leaf features such as shape, color, texture, and vein patterns to classify plant species accurately. This approach provides an accessible, efficient, and scalable method for botanical studies, conservation efforts, and educational purposes. The system emphasizes ease of use, requiring only a smartphone or basic imaging device. By leveraging convolutional neural networks (CNNs) for classification, the proposed model achieves high accuracy in species recognition. This paper explores the system's design, methodology, experimental evaluation, and future applications.

Keywords: Plant Identification, Image Processing, Leaf Recognition, Machine Learning, Convolutional Neural Networks (CNN).

Abstract: "Smart invoicing : From Transactions to Trends" streamlines invoice management by starting off with a safe sign up and login, followed by the ability to upload invoices in PDF and PNG. Using the Gemini API or google vision API, it inputs unstructured invoice data into structured formats by extracting such vital information as invoice number, date, and amount. This structured information is then rendered through interactive graphs, providing 6 users with actionable information for expense management, trend observation, and effective optimization of financial processes. Processing invoices manually is both time-consuming, prone to errors, and inefficient. In order to solve this issue, our project, "Smart Invoicing: From Transactions to Trends" applies Artificial Intelligence (AI) and data visualization to transform unstructured invoice data into actionable information. This project is designed to address the challenges of traditional invoice processing by automating and streamlining the process. It features functionalities such as secure user login, where users can upload invoices in PDF and PNG file formats. The core functionality involves the utilization of the Gemini API for the processing of unstructured invoice data into structured formats and retrieving crucial details such as invoice number, date, and amounts. Organized information is displayed through easy-to-understand graphs, providing end-users with actionable information that can be used to track spending, identify patterns, and optimize financial decisions. The system adapts to any size of business and can be applied in industries such as finance, e-business, and supply chain management. With a focus on reducing human error, increasing efficiency, and enabling sustainability by electronic management, the project remains easy to use and expandable in the future.

Keywords: Invoice processing, Optical Character Recognition(OCR), Machine learning Data Extraction, Financial Automation, gemini API or google vision API, data visualization, Automated accounting.

Abstract: As the online communication grows exponentially, the issue of toxic comments, varying from hate speech and cyberbullying to offensive and abusive content, has emerged as a pressing issue for social media sites, online forums, and news portals. Although a lot of headway has been achieved in the detection and moderation of toxic content in English, the task becomes more challenging in multilingual environments because of the varying linguistic frameworks, cultural environments, and the lack of sufficiently annotated datasets in most languages. This survey article delves into the recent research in multilingual toxic comment classification, emphasizing datasets, methods, and challenges used in this process. We detail a comprehensive critique of different strategies, such as rule-based and lexicon-based methods, old-school machine learning models, and state-of-the-art deep models. Particular emphasis is on the performance of transformer-based architectures like multilingual BERT (mBERT) and XLM- RoBERTa, based on large-scale pretraining that facilitates cross-lingual competency. In addition, we address the use of cross-lingual transfer learning in overcoming low-resource language issues and the effect of code-switching and transliteration on toxicity detection. There are still some challenges that exist despite progress, such as model and dataset biases, the absence of contextual awareness in some languages, and the dynamic nature of toxic language on the internet.

Keywords: Multilingual Toxic Comment Classification, Hate Speech Detection, Offensive Language Identification, Cross-Lingual Transfer Learning, Transformer Models, Natural Language Processing (NLP), Content Moderation.

Abstract: Higher education institutions depend on Enterprise Resource Planning (ERP) systems to sequence academic operations alongside administrative functions as well as execute data enhancement and guide institutional decision outputs. The primary intention of this study assesses Enterprise Resource Planning (ERP) adoption patterns along with technical hurdles during implementation as well as protective measures together with modern technology adoption within higher education institutions. The study evaluates multiple success factors which include user acceptance together with leadership involvement followed by specific system customization and full training programs. Enterprise Resource Planning(ERP) systems require improved security measures together with system efficiency gains and innovative technology implementation of cloud computing along with artificial intelligence and blockchain. Educational institutions can use cloud-based Enterprise Resource Planning (ERP) technologies to improve their operation efficiency at a lower cost than achieve superior student learning outcomes with Al-driven automation systems. In this paper we present an implementation of a dedicated feedback module within an existing ERP system at a higher education institution.

Keywords: ERP, Educational Framework, Artificial Intelligence, Cloud computing, LMS

Abstract: This project explores the application of blockchain technology in transforming food supply chain management. Blockchain, with its decentralized, immutable, and transparent ledger system, offers a promising solution to address the current challenges faced by the industry. By recording each transaction on the blockchain, stakeholders can trace the journey of food products from farm to fork with greater accuracy and accountability. The study presents an overview of how blockchain can enhance traceability, improve efficiency, reduce fraud, and ensure regulatory compliance in the food supply chain. Real-world case studies, system architecture, and potential implementation frameworks are discussed to demonstrate its practicality. The results indicate that blockchain has the potential to revolutionize food supply chains by promoting transparency, trust, and sustainability.

Abstract: International Visa programs (i.e. U.S. H-1B) have extremely high application volumes with limited quotas and rigorous variable outcomes. Complexity and uncertainty propel computer-aided decision systems. We launch an end to-end MLOps platform to provide real-time visa approval predictions. Our pipeline integrates data pre-processing (Pandas), training of the model (Scikit-learn), containerized deployment (Docker), and ongoing delivery (GitHub Actions) on AWS. The models and data reside in AWS S3 and EC2, while being monitored by Cloud Watch. This combined approach offers scalable, reproducible deployment of predictive models. Experiments illustrate system has good accuracy (similar to previous work) and can be retrained periodically with minimal human intervention. In brief, we present an end-to-end ML pipeline that bridges the gap between application and operational utilization, to the benefit of immigration authorities, employers, and candidates alike.Automated accounting.

Abstract: Volunteering plays a vital role in building socially responsible communities, yet individuals and NGOs often struggle to find meaningful connections and manage contributions efficiently. In today’s digital era, nonprofit organizations are increasingly turning to online platforms to improve transparency and expand their outreach through accountable, web-based practices [4]. The integration of Artificial Intelligence (AI) into such platforms offers smarter, more personalized volunteer matching, making engagement more effective and user-centric [1]. At the same time, social networks and peer recognition have been shown to significantly boost volunteer motivation and retention [3]. On the financial side, digital fundraising tools especially crowdfunding enable NGOs to reach a wider audience and secure support more flexibly and quickly [2], [6]. This paper introduces a survey-informed digital platform designed to address these needs by combining features like AI-driven opportunity matching, volunteer hour tracking, verified certification, university partnerships, and secure donation processing. By leveraging technology to simplify and strengthen collaboration among volunteers, NGOs, and academic institutions, the platform enhances civic participation and operational efficiency [5].

Keywords: Volunteering Platform, NGO Connections, Service Hours, Verified Certificates, Donations, Community Engagement, Student Activity Points, Volunteer Opportunities, Social Impact, Digital Certification, University Integration, Secure Donations.

Abstract: With growing reliance on technology for safe and efficient transit, real-time transport tracking systems in academic institutions have become essential. This paper surveys existing solutions for user expectations, system capabilities, design strategies, and implementation challenges towards developing such systems. Insights are drawn from a variety of contemporary studies to recommend optimal features and architectural considerations. It emphasizes the importance of user-centric design, mobile accessibility, and integration with GPS-enabled tracking servers to provide accurate bus location updates. Key findings underline the necessity of secure, scalable, and user-friendly systems tailored to the specific needs of students, parents, drivers, and college administrators. The paper also discusses various technologies and methodologies adopted in existing systems and proposes an optimized system architecture for a mobile-based college bus tracking solution. This study serves as a resource for developers and academic planners aiming to improve student and staff mobility leveraging digital solutions.

Keywords: Transport, Real-time tracking, GPS, mobile application, user experience, system design, student safety, student mobility.

Abstract: Public distribution system is a government-sponsored chain of shops entrusted with the work of distributing basic food and non-food commodities to the needy sections of the society at very cheap prices. Wheat, rice, kerosene, sugar, etc. are a few major commodities distributed by the public distribution system. Fair Price Shop does not open every day, nor do they keep regular hours. Even on the days that the Fair Price Shop is open, ration card holders have to stand in long queues. As social distancing was not followed at several fair price shops during the first phase of public distribution, the Civil Supplies Department has issued paper token to the beneficiaries, mentioning the date for them to avail food grains and relief fund. The proposed project aims to modernize the Public Distribution System (PDS) in India, specifically addressing challenges faced by Fair Price Shops. This not only saves time for beneficiaries but also aligns with social distancing measures crucial for public health .Additionally, the system allows for two re-slot allocations, providing flexibility for those who miss their initial collection slot. This not only streamlines the process but also allows individuals to view product details online, saving time and enhancing accessibility. By implementing a virtual queuing system through the Q Learning algorithm, the approach seeks to replace traditional physical queues with organized slot allocations. Ration cardholders would receive SMS notifications specifying the date and time for product collection, reducing the need for individuals to stand in long queues or frequent the Fair Price Shop every day. This not only saves time for beneficiaries but also aligns with social distancing measures crucial for public health .Additionally, the system allows for two re-slot allocations, providing flexibility for those who miss their initial collection slot. This not only streamlines the process but also allows individuals to view product details online, saving time and enhancing accessibility.

Keywords: Virtual Sign-up, Online Channels, Mobile Apps , QR Code , Traditional Channels , Virtual Waiting/Queues, Multi-channel Remote , Queuing Information Updates , Traditional Queue Management Components , Counter Plates, Digital Signage Screens.

Abstract: Markets for carbon credits are essential to international efforts to mitigate climate change. However, regional disparities, double-counting, and transparency problems plague traditional trading systems. In order to implement hard regional purchase caps and real-time fraud detection, this paper suggests a blockchain-based carbon credit trading platform that is integrated with AI analytics. To enforce geographic credit limits and prevent transactions that surpass regional thresholds, our system uses Ethereum smart contracts. AI models forecast market behaviour, identify anomalies, and track pricing trends. While a 3D interactive dashboard plots time, credit usage, and regions, IPFS and Zero-Knowledge Proofs (ZKP) guarantee secure data privacy. By improving trust, regulatory oversight, and fair trading, the platform benefits governments, traders, and environmental organisations. Automated blocking of users who try to cross area boundaries is validated by simulated use cases. IoT integration for real-time emission tracking is part of future work.

Keywords: Carbon Credit, Blockchain, Smart Contracts, AI Analytics, Zero-Knowledge Proofs, IPFS, Regional Caps, 3D Dashboard, Carbon Trading, Market Fraud Detection, Sustainability, ESG, Hyperledger, Ethereum, IoT Sensors, Environmental Policy.

Abstract: Brain stroke is a complicated disease that is one of the foremost reasons of long-term debility and mortality. Because of breakthroughs in Deep Learning (DL) and Artificial Intelligence (AI) which enable the automated detection and diagnosis of brain stroke as well as intelligently assisting post-brain stroke patients for rehabilitation, is more favorable than a manual diagnosis. Many publications on automated brain stroke detection, diagnosis, and robotic management using DL and AI approaches are now being published. This review provides a study of the detection, diagnosis of brain stroke and robotic management techniques of post-brain stroke rehabilitation from six different perspectives, namely, brain stroke datasets and modalities of brain stroke data collection, pre-processing approaches, DL-based detection and diagnosis of brain stroke, Al-based intelligent post brain stroke rehabilitation assistant, and performance measures. It also examines the conclusions and the consequences of the findings. There are also three ongoing research challenges in the fields of brain stroke detection and diagnosis, as well as post-brain stroke robotic treatment. For this investigation, 130 key papers from the Scopus, PubMed and Web of Science archives were chosen after a comprehensive screening method. This study gives a comprehensive overview of brain stroke detection and post-brain stroke robotic management strategies that may be useful to the scientist’s community working in the field of automatic brain stroke detection and robotic rehabilitation management.

Keywords: Brain stroke detection, internet of medical things, logistic regression, random forest, decision trees, support vector machine, shapley additive explanations

Abstract: This paper presents the design and implementation of LEGAL AI, an artificial intelligence-powered legal research and case prediction system customized for the Indian judicial context. It leverages a fine-tuned LLaMA-2 model and InLegalBERT using transfer learning and domain adaptation to provide functionalities such as case outcome prediction, legal explanation generation, and legal question answering (Legal QA). The system employs a Streamlit interface and FAISS-based vector search to retrieve relevant legal documents and provide contextual legal insights. With domain-specific fine-tuning and quantized models for CPU inference, LEGAL AI enhances accessibility, interpretability, and efficiency in legal research and decision-making.

Keywords: Legal AI, LLaMA-2, InLegalBERT, Legal Question Answering, Indian Judiciary, FAISS, Domain Adaptation , Retrieval-Augmented Generation.

Abstract: The "Smart Solar Water Management System" integrates renewable energy with intelligent monitoring to provide a sustainable and automated solution for water distribution. Powered by solar energy, the system enables real-time water usage tracking, automatic billing, and quality control through advanced sensors and data analytics. By leveraging IoT and wireless communication technologies, it ensures efficient resource management and minimizes human intervention while promoting transparency and sustainability.

Keywords: Solar-powered systems, Smart water management, Automatic billing, Water quality monitoring, IoT-based control.

Abstract: Underwater image quality is seriously degraded as a result of light scattering and absorption, which poses challenges of color distortion, reduced visibility, haze, and noise. Such visual degradation poses significant challenges to faithful object identification and hampers important applications such as marine exploration, underwater surveillance, and autonomous vehicle navigation. Traditional image-enhancement approaches are inefficient in restoring image fidelity. In order to tackle these problems, we introduce a sophisticated underwater image enhancement system that combines deep learning-based object detection with dedicated processing blocks for color correction, haze removal, and noise reduction. By using this combined approach, natural color tones are restored, scattering effects are minimized, and noise is reduced, hence improving visual quality and detection robustness. Our solution is aimed at facilitating real-time underwater operations like marine biodiversity analysis, autonomous navigation, and emergency response with enhanced accuracy and decision-making abilities.

Keywords: Underwater Image Processing, Object Detection, Color Correction, Dehazing, Denoising, Marine Research.

Abstract: In recent years, crowdfunding has emerged as a powerful tool for individuals and startups to raise capital for innovative ideas. However, traditional crowdfunding platforms are often centralized, lack transparency, and are prone to fraud or mismanagement of funds. This project presents a decentralized crowdfunding application built using blockchain technology, offering a secure, transparent, and trustless alternative. Leveraging Ethereum smart contracts written in Solidity and integrated through a React-based frontend, users can create and contribute to fundraising campaigns with MetaMask acting as the bridge for seamless blockchain interactions. The application ensures that funds are only released when campaign goals are met, and all transactions are immutably recorded on the blockchain. This solution aims to redefine the crowdfunding ecosystem by minimizing third-party interference and maximizing donor trust.

Keywords: Crowdfunding, Decentralized platform, Smart contract, Solidity, MetaMask wallet, Transparency, Efficiency, User Interface

Abstract: Tourism is fast growing, and travelers need improved digital tools. Many find it difficult to obtain reliable information and book authentic experiences. Tourists often find it difficult to move confidently in new destinations. This application helps travelers connect with credible travel agencies with easy bookings. Translucent pricing avails individuals of making mistakes of hidden charges or overcharging. Real-time updates for attraction, hotels, cafes, and cultural events. Solo travelers can join group tours to share experiences and save on costs. The app promotes social interaction and travel enjoyment. It also supports local businesses, thus boosting tourism in underrepresented areas. Built with React Native, it works smoothly on Android and iOS. Firebase ensures real-time updates and data synchronization. Cloud computing provides reliability, scalability, and high performance. Verified reviews help users make informed travel decisions. Users can explore cultural activities like dance performances and street shows. Safety features and correct navigation improve the overall experience. The app is designed for simplicity, accessibility, and reliability. Its flexible structure allows for future updates and new features. Combining technology with tourism, it creates a seamless travel experience. With this app, the world becomes easier to explore, more fun, and stress-free. It helps bridge the gap between travelers and authentic local experiences. With a simple and reliable design, travel becomes easy, fun, and connected. In addition to its robust feature set, the application integrates AI-driven personalization to recommend destinations, activities, and travel packages based on user preferences and past behavior. This enhances the user journey by curating relevant content, saving time and increasing satisfaction. The app also supports multilingual capabilities, breaking down language barriers and making it accessible to a global audience. By incorporating user feedback and analytics, continuous improvements ensure the app evolves to meet changing traveler needs. Furthermore, partnerships with local guides and artisans foster community involvement, giving users access to truly immersive and off-the-beaten-path experiences. This thoughtful blend of innovation, user-centric design, and cultural connectivity transforms how people discover and enjoy the world. To further enrich the travel experience, the app includes an integrated itinerary planner that allows users to customize their day-to-day schedules with ease, syncing accommodation, activities, and transport in one place. Push notifications keep travelers updated about weather changes, local alerts, or limited-time offers, ensuring they're always informed. For those concerned about sustainability, the app highlights eco-friendly stays and tours that support environmental conservation and responsible tourism. It also provides offline access to maps and essential travel information, making it reliable even in low-connectivity areas. A built-in expense tracker helps users monitor their spending and stick to their travel budget. With secure in-app payments and multilingual customer support, the platform offers convenience and peace of mind. By combining smart technology with a passion for exploration, this app redefines modern travel—making it more connected, culturally rich, and effortlessly enjoyable for everyone.

Keywords: destination ,modes, agencies, admin, Top places, Bookmarks , Details , Shareapp , Rating , Feedback

Abstract: The dynamic nature of unemployment rates presents a persistent challenge for policymakers and economists striving to maintain labor market stability. Fluctuations in employment levels are influenced by a multitude of factors, including economic shifts, policy changes, and global market conditions. This project introduces a predictive model designed to analyze unemployment trends using linear regression enhanced with recursive data analysis. By examining historical unemployment data, the model identifies critical patterns and key influencing factors, offering valuable insights into employment dynamics. The integration of recursive data handling allows the model to continuously update its predictions as new data becomes available, refining its accuracy over time. This adaptive approach ensures that the model remains responsive to evolving economic conditions, making it a reliable tool for labor market analysis. Through predictive insights, this system enables policymakers, economists, and other stakeholders to make informed, data- driven decisions aimed at mitigating unemployment.. Ultimately, this model serves as a robust analytical framework for understanding and managing employment trends in an ever changing economic landscape.

Keywords: Unemployment Prediction, Machine Learning, Random Forest, SVM, KNN, Data Analysis, Economic Stability, Workforce Management, Real-time Data, Predictive Modeling, Feature Engineering, Policy Formulation, NLP, Data Visualization.

Keywords: Automated Grading, NLP, OCR, Deep Learning, Machine Learning

Abstract: In today’s rapidly evolving digital landscape, podcast creators face significant challenges in content management, accessibility, and audience engagement. Traditional podcasting platforms often lack automation, requiring manual efforts for editing, transcription, and distribution. Additionally, discoverability issues and limited collaboration features hinder creators from reaching wider audiences efficiently. Our project introduces an advanced multimodal podcast orchestration framework, leveraging AI-driven tools and cloud-based automation to streamline podcast production. By integrating modern technologies such as Next.js, React.js, Convex, and OpenAI, our system enhances content management with automated `transcription, AI-powered content summarization, and intelligent recommendations. The platform offers real-time collaboration tools, allowing podcasters and editors to work seamlessly. Secure user authentication is implemented via Clerk, ensuring data integrity and controlled access. Unlike traditional solutions, our system provides a scalable and adaptive infrastructure, enabling smooth performance under high workloads. Through automated editing, voice processing, and intelligent tagging, our framework reduces the manual workload on creators, improving efficiency and content quality. The integration of machine learning models enhances content discoverability, making personalized recommendations based on listener behavior. By providing a unified solution for hosting, organization, and distribution, this system significantly simplifies podcast production. The proposed framework ensures accessibility, multi-device compatibility, and AI-enhanced automation, addressing current limitations in podcast management. Future advancements may include monetization tools, real-time translation, and deeper AI integration to further enhance the ecosystem. By leveraging cutting-edge technologies, this project aims to revolutionize podcasting, empowering creators with tools that optimize workflows, maximize audience reach, and redefine digital audio experiences.

Keywords: Podcast Management, AI Automation, Cloud-Based Framework, Content Discovery, Machine Learning, Audio Processing, Real-Time Collaboration, User Authentication.

Abstract: Side-hustles, or income-generating work performed alongside full-time jobs, have grown significantly in popularity as the gig economy provides opportunities for supplementary work. This increasing demand has created the need for platforms like AuraAssign, designed to connect individuals seeking temporary jobs with organizations requiring a flexible workforce. Events such as weddings, festivals, and corporate gatherings often require temporary staffing, but traditional systems relying on spreadsheets and static databases fail to meet these dynamic needs, leading to inefficiencies and missed opportunities. AuraAssign addresses these challenges by integrating advanced geolocation technologies and algorithms. The platform utilizes the Haversine formula to calculate precise distances between job seekers and employers and employs the OSRM algorithm (Open Source Routing Machine) with a multi-level A* algorithm to optimize routing, ensuring efficient navigation to job locations. Additionally, AuraAssign incorporates Location-Based Services (LBS) to provide real-time recommendations tailored to user preferences, including location, skills, and availability. It overcomes the limitations of traditional recommendation systems, such as collaborative filtering, which are unsuitable for dynamic and event-based social networks. By streamlining job matching and addressing challenges in short-term workforce management, AuraAssign not only improves user experience but also supports societal needs, offering students supplementary income opportunities to bridge financial gaps.

Keywords: Event-based Social Network, Haversine, Location-Based Service, Open Source Routing Machine.

Abstract: Dream Path is comprehensive career guidance website designed to help individuals discover and pursue their ideal career paths by analyzing their skills, interests, and educational background. Featuring an intuitive and user-friendly front-end, the web site provides seamless navigation with essential functionalities such as registration, login, password reset, and profile management. Its robust back-end, developed using Python, efficiently manages user data, validates credentials, and generates personalized course recommendations based on individual profiles. By submitting their academic qualifications and skill sets, users receive tailored career course suggestions that align with their aspirations, enhancing their professional growth. Additionally, Dream Path allows users to modify their profiles, update preferences, and track their progress as they advance in their careers. By bridging the gap between education and employment, Dream Path promotes continuous learning and personalized career development, equipping users with the necessary resource to navigate the evolving job market with confidence and clarity. Index Terms—kills, career opportunities, Credential Validation, Course recommendations.

Abstract: The Automated Grocery Monitoring System for Elderly People is an IoT-based solution developed using the ESP32 microcontroller to intelligently manage household grocery inventory and enhance kitchen efficiency. The system continuously monitors the weight of two essential grocery containers using load cells, while also tracking environmental parameters such as temperature and humidity through the DHT11 sensor, and food spoilage gases using an MQ2 gas sensor. Real-time data is displayed on an LCD screen and communicated via Telegram, allowing users to receive instant updates on grocery availability, environmental conditions, and potential spoilage alerts. When the weight of an item falls below a predefined threshold or spoilage is detected, the system sends an automatic notification to prompt refilling or replacement. Furthermore, based on the available grocery items, the system provides intelligent recipe recommendations, helping users plan meals effectively. This smart Ordering approach reduces food waste, ensures timely grocery management, and introduces a personalized cooking assistant feature, all integrated through a user-friendly Telegram interface.

Keywords: Grocery Monitoring, ESP32, Load Cell, DHT11, MQ2 Sensor, Telegram Alerts, Recipe Suggestion.

Abstract: In recent years, vehicle numbers have surged, but road infrastructure and traffic systems have lagged, leading to inefficient management. The rise in vehicle types, poor traffic control, and technical failures in signal systems exacerbate congestion, emissions, and noise pollution in smart cities. Conventional traffic control systems do not handle the complex traffic flow at the junctions, whereas existing traffic control systems work on fixed time- based techniques. The number of new vehicles on the road is increasing rapidly, which in turn causes highly congested roads and serving as a reason to break traffic rules by violating them. This leads to a high number of road accidents. New technologies such as computer vision (CV) and artificial intelligence (AI) are being used to solve these challenges. The proposed system integrates automated traffic signal adjustments and violation detection to address the challenges of increasing vehicular density and non-compliance with traffic rules. With its ability to enhance traffic flow efficiency and promote disciplined driving behavior, this system represents a significant step toward smarter and safer cities. The use of algorithms such as YOLO has the potential to revolutionize traffic management in urban areas, leading to a more efficient and sustainable transportation system. As a result, these technologies have established a distinct identity in the surveillance industry, particularly for continuous traffic monitoring. Traffic violation detection systems using computer vision efficiently reduce violations by tracking and penalizing offenders while alerting compliant drivers, ultimately decreasing fatal motorcycle accidents. Effectiveness is measured through key metrics such as traffic density estimation, violation detection accuracy (for red-light and helmet violations), and processing speed, ensuring real-time decision- making and optimized traffic management.

Keywords: Smart Traffic, YOLOv8, Traffic Violation, Real-Time Detection, Signal Control, AI for Safety, Smart cities.

Abstract: This project is designed to create a comprehensive hospital-based website that includes a symptom checker chatbot, document storage, and an appointment management system to enhance healthcare services. The symptom checker utilizes AI and natural language processing (NLP) to allow users to enter their symptoms and receive tailored insights into possible health concerns, along with suggested next steps. It serves as an initial assessment tool while promoting professional consultations for accurate diagnoses. The platform also ensures the secure storage of patients' medical documents, making it easy to access and retrieve health records. Furthermore, users can book appointments with healthcare professionals, improving convenience and minimizing waiting times. With an emphasis on accessibility, privacy, and a user-friendly interface, this project aims to boost the efficiency of healthcare delivery, empower patients with self-care tools, and streamline hospital operations in a technology-driven healthcare landscape.

Abstract: The "Cotton Leaf Disease Detection and Automated Spraying System" offers an intelligent, image-based solution for identifying plant diseases and performing precision pesticide spraying with minimal human intervention. By utilizing real-time image acquisition, a CNN-based classification model, and embedded actuation via Raspberry Pi, the system ensures reliable, automated treatment of diseased cotton plants. A Flask-based interface, along with onboard sensors, supports responsive decision-making, while the mobile platform enables deployment across diverse field environments.

Keywords: Cotton Leaf Disease, Convolutional Neural Network (CNN), Image Processing, Raspberry Pi, Automated Spraying, Machine Learning, Precision Agriculture, Flask Web Interface, Pesticide Control, Smart Farming.

Abstract: The DR (Diabetic Retinopathy) is an eye variation which the human retina is influenced because of long haul diabetes. Diabetes is a chronic condition related to an expanding measure of glucose level. As the degree of glucose builds, a few adjustments happen in veins of the retina. As diabetes advances, the vision of patients may begin to cause Diabetic Retinopathy. It is exceptionally far reaching among moderately aged and older individuals. In this article, fundus images of eye (retina) are used and the features are extracted from these images using the image processing technique. Images are trained, tested and severity of the DR is classified using (CNN) algorithm.

Keywords: Diabetic Retinopathy Screening (DRS), Classification, Prediction, Image Processing, Machine Learning, Retinal Images, Data Analytics.

Abstract: Breast cancer continues to be a leading cause of mortality among women worldwide, necessitating early and precise diagnostic systems. While Convolutional Neural Networks (CNNs) have made significant strides in medical image analysis, their limitations in modeling long-range dependencies persist. This study proposes an advanced breast cancer detection model based on Vision Transformers (ViTs) integrated with transfer learning. Pre-trained ViT models were fine-tuned on histopathological breast cancer image datasets to address data scarcity and enhance classification accuracy. The model was evaluated using metrics such as accuracy, AUC, and F1-score, and showed superior performance compared to traditional CNNs. These results highlight the potential of ViTs in transforming breast cancer diagnosis into a more automated, robust, and accurate process.

Keywords: Breast Cancer Detection, Vision Transformers, Transfer Learning, Medical Image Analysis, Deep Learning, Histopathology

Abstract: This paper presents a flexible and an inexpensive chronic kidney disease prediction system by utilizing machine learning models including Deep Neural Networks (DNN), Support Vector Machines (SVM), and Extreme Gradient Boosting (XGBoost). The interface between the clinical data sets and advanced AI algorithms for accessing patient records and controlling disease progression remotely will be made by using comparative analysis of these three models. This study node connected to clinical attributes that can be controlled using smart data preprocessing and remotely controlled through an access point. The Smart CKD prediction system for healthcare development consists of two major parts that are smart diagnostic device and the access point. The main hardware for this system contain: Clinical Dataset, Machine Learning Models, Feature Selection, Data Preprocessing, Model Evaluation Metrics, Performance Analysis, Confusion Matrix, ROC Curves, and Statistical Analysis. Expected outcomes from this system: programming by using Python that comes built-in with Scikit-learn, TensorFlow module adapter to make connections between the clinical data and AI models for precise CKD prediction.

Keywords: Chronic Kidney Disease, Diagnosis, Deep Neural Networks, Support Vector Machines, XGBoost, Machine Learning, Artificial Intelligence, Clinical Decision Support Systems, Feature Selection, Early Detection, Health-care Analytic s, Accuracy, Sensitivity, Specificity.

Abstract: Credit card fraud remains a major challenge for financial institutions due to the growing number of online transactions and the sophistication of fraudulent techniques. Traditional machine learning methods often struggle with class imbalance and lack contextual understanding. In this study, we propose a credit card fraud detection framework leveraging Transformer-based architectures integrated with transfer learning. The model is fine-tuned on transaction data to detect fraudulent activities effectively. Experimental results demonstrate improved performance in comparison to conventional classifiers, suggesting that Transformer-based models are well-suited for time-series and sequential data in fraud detection scenarios.

Keywords: Credit Card Fraud Detection, Transformers, Transfer Learning, Deep Learning, Anomaly Detection, Financial Security

Abstract: Recent advances in recommendation systems have demonstrated significant potential through deep learning approaches yet challenges remain in computational efficiency and prediction accuracy. This research presents a novel framework that integrates deep learning compression techniques with statistical variance analysis to enhance movie recommendation performance while reducing computational overhead. The proposed system leverages multi-modal data from MovieLens and IMDB datasets, implementing vector quantization and embedding compression to achieve optimal memory utilization. Our methodology incorporates standard deviation analysis to evaluate recommendation consistency and employs quantization-aware training for model optimization. Experimental validation using MovieLens 25M and IMDB datasets demonstrates superior performance compared to baseline collaborative filtering methods. The system achieves a 15.3% reduction in Root Mean Squared Error (RMSE) while maintaining 79.2% compression ratio through INT4 quantization. Statistical analysis reveals improved recommendation consistency with standard deviation values of 0.89 for highly-rated content compared to 1.67 for polarizing content. The framework addresses critical gaps in computational efficiency and recommendation accuracy particularly in large-scale deployment scenarios. Results indicate significant improvements in both prediction quality and system efficiency with 68% reduction in memory requirements and 45% faster inference time. This research contributes to the advancement of efficient recommendation systems by demonstrating the effectiveness of combining compression techniques with statistical analysis for enhanced user experience and system scalability.

Keywords: Movie Recommendation, Deep Learning Compression, Vector Quantization, Standard Deviation Analysis, Movie Lens Dataset, IMDB Integration, Collaborative Filtering, Embedding Compression

Abstract: The increasing environmental consciousness among consumers necessitates the development of intelligent recommendation systems that balance user preferences with sustainability goals. This research presents a novel Context-Aware Fuzzy Recommender System for Sustainable Product Discovery (CAFRS-SPD) that integrates contextual information, fuzzy logic reasoning and statistical aggregation methods (mean and median) to recommend environmentally responsible products. The proposed system addresses the critical gap in existing recommender systems that primarily focus on user satisfaction while neglecting environmental impact. Our methodology combines fuzzy membership functions with contextual factors such as temporal preferences, location-based constraints and user sustainability awareness levels. The system employs mean and median statistical measures for aggregating multiple sustainability criteria including carbon footprint, recyclability index and energy efficiency ratings. Experimental validation using the Amazon Product Dataset and MovieLens-25M dataset demonstrates that CAFRS-SPD achieves a 23.7% improvement in sustainability score while maintaining recommendation accuracy within 5.2% of traditional systems. The fuzzy inference engine successfully handles uncertainty in sustainability assessments while contextual adaptation ensures personalized recommendations aligned with individual user contexts. Comparative analysis with five baseline methods reveals superior performance in terms of sustainability awareness (F1-score: 0.847), contextual relevance (precision: 0.823) and user satisfaction (recall: 0.791). The statistical aggregation approach using weighted mean and robust median estimators effectively combines heterogeneous sustainability metrics, resulting in more reliable sustainability assessments. This research contributes to the growing field of green recommender systems by providing a comprehensive framework that promotes sustainable consumption patterns while preserving user experience quality.

Keywords: sustainability recommendations, fuzzy logic, context-aware systems, statistical aggregation, green computing, machine learning, environmental impact, sustainable consumption

Abstract: Recognizing and identifying emotions is a key element of understanding mental health that can ideally lead to better emotional well-being. The project combines facial recognition of emotions with data analysis to assess for conditions such as anxiety and depression. This detects and provides real-time emotion profiles with individual perspectives for healthcare workers via deep learning. To detect facial expression, Convolutional Neural Networks (CNNs) are utilized along a series of facial characteristics to extract essential details. Artificial Neural Networks can be made to classify emotions and, thus, play a role in understanding patterns related to mental health conditions; e.g., feeling drowsy, anxious. This system essentially links technology and health care, effectively equipping mental health professionals with modern, data-driven tools for on-time and personalized interventions.

Keywords: Artificial Neural Network, Convolutional Neural Network.

Abstract: The construction industry faces significant challenges in resource allocation, workforce management and project scheduling, leading to cost overruns and delays. Traditional Building Information Modeling (BIM) approaches lack intelligent decision-making capabilities for dynamic resource optimization. This research presents a novel framework integrating machine learning (ML) and deep learning (DL) techniques with BIM for intelligent resource and workforce management in construction projects. The proposed system utilizes Support Vector Machines (SVM), Random Forests and Convolutional Neural Networks (CNN) to predict resource requirements, optimize workforce allocation and automate construction scheduling. The framework processes historical construction data through a multi-layered architecture that combines BIM model data with real-time project parameters. Experimental validation using the PSPLIB dataset and real-world construction projects demonstrates significant improvements in resource utilization efficiency (25% improvement), schedule accuracy (18% reduction in delays) and cost optimization (15% reduction in project costs) compared to traditional methods. The system achieved 89% accuracy in predicting resource requirements and 92% precision in workforce allocation decisions. Deep learning models showed superior performance in clash detection and conflict resolution, achieving 95% accuracy in identifying potential construction conflicts. The integration of predictive analytics with BIM data enables proactive decision-making, reducing manual intervention by 40% and improving overall project delivery timelines. This research contributes to the advancement of intelligent construction management systems and provides a foundation for future development of autonomous project management platforms.

Keywords: Building Information Modeling, Machine Learning, Deep Learning, Resource Optimization, Workforce Management, Construction Scheduling, Predictive Analytics, Intelligent Construction

Abstract: This research explores a framework for selective answer analysis using keyword-based filtering and semantic similarity techniques. With the increasing volume of textual data generated through surveys, feedback mechanisms, and question-answer systems, it is often impractical and unnecessary to process every response. Our approach filters and analyzes only those answers that align with a specified set of keywords or topics of interest, leveraging advanced natural language processing (NLP) algorithms to prioritize relevance and reduce computational overhead. By combining lexical filtering with semantic matching, we aim to improve the efficiency, scalability, and interpretability of text analytics. The framework is evaluated on a diverse set of survey responses and demonstrates improved focus, accuracy, and thematic coherence in analysis. Additionally, the methodology incorporates dynamic thresholding to adapt to varying data densities and context-specific requirements, ensuring robust performance across datasets. Practical applications span customer sentiment analysis, educational assessment automation, and large-scale social research, offering a versatile solution for targeted data exploration. Future enhancements will focus on integrating machine learning models for adaptive keyword refinement and automated thematic categorization, further bridging the gap between precision and scalability in text analytics.

Keywords: Selective answer analysis, keyword-based filtering, semantic similarity, natural language processing (NLP), lexical filtering, thematic analysis, computational efficiency, dynamic thresholding, automated categorization, text analytics, survey response evaluation, domain adaptability.

Keywords: Online food delivery, recommendation system, MERN stack, collaborative filtering, user experience .and MERN Stack, Food Delivery App, AI Recommendations, User Personalization, Diet Preferences, Health Based Suggestions.

Abstract: Deep learning has revolutionized various fields including computer vision, big data analytics, and automation. However, the same technologies that drive innovation have also enabled the rise of deepfakes—AI-generated media designed to mimic real human expressions and voices with alarming accuracy. This paper presents a comprehensive overview of the mechanisms behind deepfake creation and critically evaluates the current state of detection techniques. Through a review of literature and research methodologies, we examine the evolution of both generation and detection approaches, discuss emerging challenges, and propose future directions for enhancing the robustness of deepfake detection systems. This work aims to provide a solid foundation for researchers and developers striving to mitigate the misuse of deepfake technology and preserve digital integrity. sequences in videos, and inconsistencies in spatial features.

Keywords: Deepfake, Machine Learning, Convolutional Neural Network, Transfer Learning, FaceForensics++, Detection Algorithms.

Abstract: Stroke remains one of the leading causes of death and long-term disability worldwide, whose impact imposes a heavy healthcare burden on individuals and systems. Early and accurate detection of stroke is critical in order to prevent delays in seeking medical care, provide improved patient outcomes, and prevent complications. The application of CT and MRI scans, the traditional methods, is not only required to be interpreted by an expert but is also time-consuming and prone to variation among radiologists. Growth in artificial intelligence (AI), and increasingly in machine learning (ML) and deep learning (DL), offers promising avenues for enhancing the accuracy and speed of stroke detection. This paper provides a detailed overview of ML and DL techniques applied in brain stroke detection, detailing the methodologies, the prerequisites for application, and the challenges posed. It explains various image processing techniques and classifying algorithms intended for detecting and segmenting regions affected by strokes within brain scans. We also cover developing an AI-based system integrating image processing with ML algorithms for assisting medical professionals to diagnose strokes more effectively. Through an extensive review of the literature, the current work presents the most recent advances in AI-based stroke detection, considering both supervised and unsupervised learning approaches, feature extraction methods, model performance evaluation measures, and challenges regarding dataset access, model interpretability, computational intensity, and deployment in the real world. By synthesizing current research evidence, this paper aims to enlighten the emerging role of AI in stroke detection and diagnosis. It also offers future directions for research aimed at improving model generalization, developing explainable AI models, and integrating AI tools into clinical practice. The evidence provided contributes to the continuum of initiatives towards stroke diagnosis improvement through novel technological advancements, leading to improved patient care and outcomes.

Keywords: Stroke detection, Machine learning, Deep learning, Artificial intelligence, Medical imaging, Image processing, Neural networks, CT scan, MRI, Stroke classification, Healthcare technology.

Abstract: In the digital age, the integration of smart systems within academic institutions has become essential for improving communication, transparency, and student engagement. KSIT Nexus is a comprehensive, cross-platform mobile application designed to streamline various student-centric services within campus environments. This paper presents the development and implementation of KSIT Nexus, which features an Anonymous Complaint System, Reading Room Tracking, an AI-powered Query Chatbot, Study Group Finder, and a Digital Notice Board. The application prioritizes user privacy and accessibility by enabling anonymous submissions while maintaining identity traceability exclusively for administrators. It is built using React Native for the frontend and Django for the backend, ensuring cross-platform compatibility and efficient data handling. Unlike conventional cloud-based applications, KSIT Nexus leverages on-premise college servers for data storage, addressing institutional concerns around data security and cost. Designed to run on low-end devices, the system supports offline-first principles and manual data entry where automation is limited. The paper also discusses the unique priority-based rearrangement of complaints using natural language processing (NLP) techniques and outlines the architectural choices made to optimize performance across devices. KSIT Nexus demonstrates how technology can bridge administrative gaps, enhance student participation, and foster a responsive academic environment.

Keywords: Campus Management System, Cross-Platform Application, React Native, Django, Anonymous Complaint System, Student Engagement, Reading Room Tracking, AI Chatbot, Digital Notice Board, Study Group Finder, NLP Prioritization, On-Premise Storage, Educational Technology, Privacy-Preserving Systems.

Abstract: The rise in unsolicited emails, known as spam, has created an urgent need for more trustworthy and powerful antispam filters. Recent advances in machine learning techniques have enabled researchers and developers to effectively identify and filter spam emails. In this paper, we present a thorough analysis of several popular machine learning-based email spam filtering strategies. We provide an overview of key concepts, methods, effectiveness, and current research directions in spam filtering. We begin by examining how top internet service providers (ISPs), including Gmail, Yahoo, and Outlook, apply machine learning techniques in their email spam filtering processes. We also describe the general process of email spam filtering and highlight the various ways researchers have applied machine learning to combat spam. Our evaluation compares the strengths and limitations of existing machine learning techniques and identifies unresolved challenges in spam filtering research. Based on our analysis, we recommend adopting deep learning and deep adversarial learning approaches to more effectively address the problem of spam emails in the future.

Keywords: Analysis of Algorithms, Machine Learning, Spam Filtering, Deep Learning, Neural Networks, Support Vector Machines (SVM), Naïve Bayes.

Abstract: The COVID-19 pandemic posed unprecedented challenges to global health systems, economies, and societies, demanding rapid and innovative responses. In this context, Artificial Intelligence (AI), data analytics, and data engineering emerged as vital tools for understanding and managing the crisis. This research paper examines how these technologies were deployed to monitor virus transmission, predict future outbreaks, allocate resources, and support evidence-based decision-making. By integrating structured and unstructured data from authoritative bodies such as the World Health Organization (WHO), national health agencies, and non-traditional sources like mobility and social media data, researchers were able to derive meaningful insights through machine learning and analytical models. Furthermore, data engineering played a foundational role in enabling seamless data integration, processing, and access, supporting scalable analytical workflows. The application of AI-driven forecasting and visualization tools enabled real-time dashboards and predictive simulations, which significantly influenced global and local health policies. This study underscores how technological innovation—when grounded in ethical principles and robust infrastructure—can empower societies to navigate complex public health emergencies more effectively.

Abstract: Object detection systems, pivotal in computer vision, identify and localize objects in images or videos. This paper explores Convolutional Neural Networks (CNNs) and MobileNet SSD (Single Shot Multi Box Detector) for efficient object detection, particularly in resource-constrained environments. We review CNN- based detection, detail MobileNet SSD’s lightweight architecture, and assess its performance. Applications in autonomous driving, mobile devices, and surveillance are discussed, alongside challenges and future directions. This study underscores MobileNet SSD’s role in enabling real-time detection on edge devices.

Keywords: Data-Driven, Object Detection, MobileNet, Convolutional Neural Networks (CNNs), Bounding Boxes, Machine Learning.

Abstract: This project presents a comparative analysis of various Attendance Management Systems (AMS) to evaluate their efficiency, accuracy, security, and scalability. With the growing need for reliable attendance tracking in educational institutions and organizations, multiple technologies—such as manual registers, RFID cards, biometric systems (fingerprint, face recognition), and mobile-based GPS tracking—have been developed and deployed. The study critically examines each system's performance based on parameters like implementation cost, ease of use, susceptibility to fraud (e.g., proxy attendance), maintenance requirements, and integration with existing infrastructure. By analyzing data collected from real-world case studies and user feedback, the project identifies the strengths and limitations of each approach. The findings aim to guide institutions in selecting the most suitable attendance solution based on their specific needs, budget, and operational environment. The study concludes that while biometric and AI-based systems offer superior accuracy and automation, factors like privacy concerns and technical complexity must also be considered.

Keywords: Biometric Authentication, Face Recognition, RFID Technology, GPS-Based Attendance, Manual Attendance, Comparative Study, Proxy Attendance Prevention.

Abstract: This research paper presents a deep learning-based approach for the automated detection and classification of plant diseases through leaf image analysis. Early and accurate identification of crop diseases is crucial for sustainable agriculture and global food security. Our system leverages Convolutional Neural Networks (CNNs) to analyse images of plant leaves and identify diseases with high accuracy. The proposed model was trained on a comprehensive dataset comprising 38,000 images spanning 14 crop species and 26 diseases. Experimental results demonstrate that our CNN based system achieves an average classification accuracy of 96.7%, outperforming traditional image processing techniques and conventional machine learning approaches. The system can identify diseases at early stages, enabling timely intervention that reduces crop losses and minimizes pesticide usage. Furthermore, we have developed a mobile application interface that allows farmers to utilize this technology directly in the field, bridging the gap between advanced AI technologies and practical agricultural applications. The Convolutional Neural Network (CNN) resulted in a improved accuracy of recognition compared to the SVM approach. Keyword: Machine Learning, Image processing, Decision Tree, Random Forest, Crop disease detection, Extreme Learning Machine, K-means Clustering

Abstract: This research paper explores the application of data analytics to an Airbnb dataset containing 74,111 listings, focusing on variables such as room type, accommodates, bathrooms, cancellation policy, cleaning fee, instant bookability, review scores, bedrooms, beds, and log-transformed price. Using Python libraries including pandas, NumPy, and seaborn, we perform exploratory data analysis (EDA) to uncover trends and relationships within the data. The study highlights key statistical insights and identifies potential sources of human error that could impact data quality and analytical and implement KNN algorithm to treat outlier values . The findings provide a foundation for understanding pricing dynamics in the Airbnb market and underscore the importance of addressing human-induced inaccuracies in workflows.

Abstract: Face recognition technology has become an effective and automated solution for managing attendance in academic and organizational settings. With advancements in machine learning and computer vision, it is now possible to recognize individuals accurately in real time using facial features. This project presents a Face Recognition Attendance Management System that automates the attendance process by detecting and verifying faces through live camera input. The system uses techniques such as face detection, feature extraction, and classification, with the help of Convolutional Neural Networks (CNN). We analyze the system’s performance under varying conditions such as lighting and face angles, and evaluate its accuracy and reliability. Our study demonstrates that integrating face recognition with attendance systems enhances security, saves time, and minimizes the chances of proxy attendance.

Keywords: Face Recognition, Attendance Management, Machine Learning, Deep Learning, Computer Vision, Convolutional Neural Networks (CNN), Image Processing.

Abstract: This paper presents Movie recommendation and advanced with the exponential growth of digital content, recommender systems have become essential tools for improving user experience and driving engagement. This paper presents the design and implementation of a movie recommendation system using Python. We explore collaborative filtering, content-based filtering, and hybrid approaches. The system is built using publicly available datasets such as MovieLens and employs tools such as Pandas, Scikit-learn, and Surprise. Our results show that hybrid recommendation systems yield better performance and personalization compared to single-method models.

Keywords: Movie recommendation, collaborative filtering, content-based filtering, hybrid system, Python, machine learning.

Abstract: One of the leading causes of cancer-related deaths among women worldwide is breast cancer. Although improving survival rates requires early detection, traditional screening methods like mammography and biopsy have disadvantages like high cost, radiation exposure, and restricted accessibility. ThermoScan AI is an artificial intelligence-powered smartphone app that uses infrared thermography to detect breast cancer without invasive procedures. In order to scan thermal images and accurately identify anomalies, this technology uses deep learning in the form of the DALAResNet50 model. By eliminating the dangers of traditional screening methods, ThermoScan AI offers a highly accessible and reasonably priced alternative. The app offers regular screening reminders, telemedicine support, AI-powered thermal image analysis, and personalized health monitoring. ThermoScan AI's advanced technology makes screening quick and easy. People living in rural and impoverished areas, where there are few traditional medical facilities, will particularly benefit from this innovation. ThermoScan AI has the potential to improve global breast cancer treatment and lower mortality rates through better early detection and preventive treatment. This project marks a significant advancement in the use of AI to provide effective and reasonably priced healthcare solutions.

Keywords: AI, Infrared Thermography, Deep Learning, Early Detection, Non-Invasive Screening, Mobile Health, Medical Imaging, Telemedicine, Healthcare Innovation, Breast Cancer.

Abstract: In a globalized world, effective communication across linguistic and cultural boundaries is critical yet challenging due to diverse languages, accents, and cultural nuances. The Multilingual Communication Assistant (MCA) is an innovative system designed to overcome these barriers by integrating real-time speech-to-speech translation, accent adaptation, context-aware translation, and cultural nuance understanding. Leveraging advanced technologies such as neural machine translation, deep learning, and speech synthesis, the MCA ensures accurate, natural, and culturally sensitive communication. This paper presents the design, development, and evaluation of the MCA, highlighting its architecture, methodologies, and potential applications in education, healthcare, business, and diplomacy. Preliminary results demonstrate high translation accuracy and user satisfaction, with future enhancements aimed at offline functionality and broader language support. The MCA promises to foster inclusive and meaningful cross-cultural interactions, redefining multilingual communication.

Keywords: Multilingual Communication, Real-Time Translation, Accent Adaptation, Cultural Nuances, Neural Machine Translation, Speech Synthesis

Abstract: With the adding reliance on video conferencing for communication, vacuity and intelligence- driven advancements have come vital for perfecting user experience. This literature check explores the development and impact of video conferencing operations integrating sign language paraphrase, AI- powered chatbots, and automated meeting recordings. subscribe language paraphrase leverages real- time computer vision and natural language processing( NLP) to make virtual meetings more inclusive for hail- crippled individualities. AI chatbots help in automating responses, scheduling, and enhancing user commerce, reducing the cognitive weight on actors. Meeting recording and recap features ensure indefectible documentation and recovery of pivotal exchanges. also, the check examines the bolstering technologies, including machine knowledge models, speech recognition, and NLP ways, to estimate their effectiveness in perfecting engagement and vacuity. The review also discusses being platforms, challenges, and future trends in intelligent video conferencing results. By addressing vacuity and automation, this study highlights the eventuality of AI- driven advancements in shaping the future of virtual communication

Abstract: A major turning point in the development of wireless communication has been reached with the introduction of 5G technologies, which promises higher speeds, ultra-low latency, and improved connectivity. The Internet of Things (IoT), a network of linked devices that depend on smooth communication to deliver intelligent services across multiple industries, is expected to be significantly impacted by this change. Many of the shortcomings of earlier generations, like network congestion, capacity constraints, and delayed communication, which have prevented IoT devices from reaching their full potential, are anticipated to be addressed by 5G. This study examines the effects of 5G on the Internet of Things, emphasizing how it improves scalability, dependability, and real-time data processing. Along with discussing the opportunities and challenges brought about by this technical revolution, it also looks at how 5G can open up new IoT applications, such as industrial automation and smart cities. By examining how 5G and IoT connect, The goal of this study is to present a thorough knowledge of how these technologies will revolutionize connected ecosystems in the future.

Keywords: 5G Technology, Internet of Things (IOT), Low Latency, Smart Devices, Network Security.

Abstract: MapNest is a web-based platform developed to automate the design of residential floor plans and optimize utility layouts, including electricity and water systems. The system leverages HTML, CSS, JavaScript, Python, and artificial intelligence to collect user requirements and generate customized house maps. The platform simplifies the architectural design process for users without technical knowledge, offering a low-cost and accessible alternative to traditional CAD software. This paper outlines the methodology, system architecture, and implementation strategy of MapNest, and discusses its effectiveness, scalability, and potential for future enhancement.

Keywords: House Mapping, Utility Design, AI, Web Application, Automation, Floor Plan Generator.

Abstract: Community detection in complex networks remains a fundamental problem in network science, with wide-ranging applications from sociology to biology and recommendation systems. Building on recent advances in label diffusion techniques, we propose a novel hybrid approach—GraphSAGE-LBLD—that combines the structural awareness of the GraphSAGE embedding model with the local balance and speed advantages of the Label Balanced Label Diffusion (LBLD) algorithm. Our method integrates representation learning into the label propagation process, allowing for more semantically meaningful diffusion and improved stability across diverse network topologies. We empirically evaluate GraphSAGE-LBLD on multiple real-world SNAP datasets and benchmark against Louvain, classic Label Propagation, and the original LBLD. Results demonstrate that our model consistently achieves higher modularity and Normalized Mutual Information (NMI) scores, while maintaining comparable runtime. The integration of GraphSAGE enhances the representation of local neighbourhoods, resulting in finer community boundaries and better detection of small or overlapping clusters. Our method offers a practical, scalable, and more accurate alternative for modern community detection tasks.

Keywords: Community Detection, GraphSAGE, Label Propagation, Graph Neural Networks, Node Embeddings, K-core, Graph Autoencoder, Modularity, Weighted Diffusion, Cosine Similarity.

Abstract: The increasing reliance on artificial intelligence (AI) for ground breaking medical advancements has led to innovative solutions in botanical medicine. "Botanic Cure: AI-Driven Medicinal Leaf Analysis" represents a pioneering approach to identifying and analysing medicinal properties in leaves using AI technologies. This study harnesses machine learning algorithms to extract, classify, and evaluate bioactive compounds within various plant leaves, aiming to streamline the discovery of new medicinal agents and enhance therapeutic efficacy. By integrating AI with traditional botanic practices, this research demonstrates significant potential in revolutionizing pharmaceutical development, providing an efficient pathway for discovering novel treatments for a wide array of diseases. Key findings reveal the accuracy and speed of AI in medicinal leaf analysis, promising substantial contributions to the field of herbal medicine and pharmacognosy.

Keywords: Leaf identification, Medicinal properties, Plant species, Herbal medicine, Image recognition, Plant database, Xception

Abstract: Respiratory illnesses are among the most prevalent child health conditions globally, with the potential to cause considerable morbidity and mortality if not promptly diagnosed and treated. Proper and timely classification of respiratory conditions like asthma, bronchiolitis, pneumonia, and upper respiratory infections is essential to provide proper treatment and avoid complications. This research investigates the creation of a pediatric patient-specific respiratory disease classification system based on clinical signs, auscultation results, and diagnostic imaging information. Taking advantage of machine learning methods, such as decision trees, support vector machines, and deep learning, we seek to enhance the accuracy of diagnosis and facilitate clinical decisions within pediatric healthcare environments. We have used annotated pediatric clinic medical records, considering pediatric patients aged between 6 and 14 years. Initial findings show exceptional classification accuracy, particularly in demarcation between viral and bacterial infections. This research highlights the ability of data-driven methods in promoting pediatric respiratory management and provides the groundwork for putting intelligent diagnostic tools to clinical use.

Keywords: Respiratory sound classification, Adventitious respiratory sounds, Respiratory diseases, Deep learning.

Abstract: PharmAssist is a cloud-based Pharmacy Management System with artificial intelligence for better stock management in pharmacies, telemedical services for doctors and patients. Unlike offerings available in the market today, it has an AI-driven alternative medicine suggestion module that recommends alternatives by comparing medicine compositions. The system also provides e-prescriptions, where doctors e-mail prescriptions to patients. It also provides real- time pharmacy stock management with low-stock and expiration date automatic reminders. Location search for pharmacies assists users in searching for pharmacies around them using Google Places API, displaying shop information. PharmAssist is built with cloud infrastructure (AWS) and Flask, MongoDB Atlas, and artificial intelligence models (TensorFlow), thus enhancing efficiency, accessibility, and reliability in the digital pharmacy platform.

Keywords: Pharmacy Management System, Artificial Intelligence, Machine Learning, Telemedicine, Alternative Medicine Recommendations, Cloud Computing, E-Prescriptions, Stock Management, Healthcare Technology, Pharmacist-Patient- Doctor Integration, AI-Based Drug Substitution, MongoDB Atlas, AWS Lambda, Flask API, Google Places API, Video Consultation, Twilio API, UPI-Based Payments.

Abstract: FLASHO is a cutting-edge platform that enables on-demand fuel delivery for both corporate and individual customers. Customers can easily order fuel via a mobile application, and it will be delivered promptly to the location of their choice. To enhance the user experience, the platform makes use of secure payment methods, real-time notifications, and GPS technology for tracking. FLASHO offers fleet management tools for businesses and drastically reduces waiting times and fuel waste by eliminating the need to visit gas stations. The platform is based on a cloud-based, scalable architecture that respects legal requirements, puts safety first, and encourages eco-friendly behavior. FLASHO's reliable, effective service is poised to revolutionize the fuel industry.

Keywords: Cloud-based platform, Safe payment methods, GPS technology, On-demand fuel delivery, Safety procedures, and Improved user experience.

Abstract: Women’s safety has become an urgent concern in today’s world, especially when they are traveling alone. To address this issue, we developed Safe Journey Navigator, a mobile application designed to help women travel more safely and confidently. The app suggests the safest possible routes using location data and keeps track of the user’s live location throughout the journey. If the user strays from the selected route or shakes the phone in an emergency, the app immediately sends alerts along with the live location to predefined emergency contacts. Additionally, it provides quick-call buttons for ambulance and police services. We built the app using Flutter and integrated it with various APIs to ensure real-time tracking and notification delivery. Our goal is to combine technology and safety in a simple, user-friendly interface that can help reduce risk and provide peace of mind for women during travel. Initial testing showed promising results, and we plan to improve the system further by adding features like voice commands and AI-based threat detection.

Keywords: Women’s Safety, Safe Route Navigation, Emergency Alerts, Location Tracking, Mobile Application

Abstract: Phishing attacks have emerged as one of the most prevalent cybersecurity threats, targeting unsuspecting users by imitating legitimate websites to steal sensitive information. This project aims to develop an advanced phishing detection and prevention system using machine learning and browser extension-based security awareness. The system is designed to simulate phishing websites, analyze user interactions, and extract key dataset features to improve phishing detection mechanisms. The proposed solution consists of three core components: 1. Phishing Website Simulation – A controlled environment where phishing websites are created to mimic real-world attack patterns. User interactions are analyzed, and dataset features such as URL structure, SSL certificate status, and JavaScript behavior are extracted to enhance detection accuracy. 2. Machine Learning-Based Detection – The system trains various machine learning models (Random Forest, Support Vector Machine (SVM), and Neural Networks) using datasets of phishing and legitimate websites. Key extracted features like URL length, domain age, presence of HTTPS, and script execution patterns help in real-time classification to differentiate between phishing and authentic sites. 3. Browser Extension for Prevention – A real-time browser extension that integrates with the machine learning model to scan webpages before loading. It warns users via pop-up notifications when a phishing attempt is detected and blocks access to malicious websites. The extension also logs phishing attempts, displaying IP addresses, geolocation, and additional metadata for further research and reporting. The system architecture follows a multi-layered approach, leveraging client-side security mechanisms, cloud-based threat intelligence, and AI-driven classification for effective phishing detection. The methodology ensures real-time protection for users while also generating datasets for continuous model training and enhancement. This project contributes to enhancing cybersecurity awareness by educating users about phishing tactics and equipping them with proactive security measures. Additionally, real-time logging of phishing attempts provides cybersecurity researchers and organizations with valuable data to refine detection strategies and mitigate threats. Through extensive testing and validation, the proposed phishing detection and prevention framework achieves high accuracy in identifying phishing attempts while maintaining low false positive rates. Future scope includes expanding detection capabilities to mobile browsers and integrating blockchain-based threat validation for added security. This project ultimately aims to empower users with real-time phishing protection, improve cybersecurity resilience, and enhance global efforts in combating phishing-related cyber threats.

Keywords: Phishing Attack, Cyber Security, Machine Learning, Artificial Intelligence, Browser Extension, Website Detection, Cybercrime Prevention

Abstract: Recent scenarios of companies are struggling with complex vulnerabilities. Hence, there is need of an Automated tool to overcome the longest time frame to cover the gap from forensics and cybercrime. This paper presents an innovative approach to automating Cyber Threat Intelligence processes designed to ingest, analyze, display and respond to emerging threats in real time. Formerly, CTI used to rely heavily on manual methods for collecting, interpreting and analyzing data which was not only time consuming but also prone to inefficiencies, especially when rapid information dissemination is critical. IntelShield integrates multiple open-source intelligence feeds, does real time active network monitoring, severity scoring, triggers automated alerts via SMS using Twilio and displays threats on a user friendly dashboard. It also employs Natural Language Processing(NLP) to extract indicators of compromise (IOCs) from unstructured sources such as security blogs, reports, and dark web forums. Machine learning techniques are used to classify, prioritize, and correlate IOCs based on their severity and contextual relevance. This research highlights the transformative potential of AI-driven technologies to enhance both the speed and accuracy of CTI.

Keywords: Cyber Threat Intelligence, Automated CTI, Real Time Security, Open Source Intelligence, OSINT, CVSS, SIEM, Dashboard, Natural Language Processing (NLP), Indicators of Compromise(IOC).

Abstract: Brain-Computer Interface (BCI) technology enables communication between the human brain and external devices without requiring physical interaction. This paper presents a novel, non- invasive EEG-based BCI system for smart home automation, designed to assist individuals with physical disabilities. The system uses Bio amp EXG Pill for brain signal acquisition and the Maker UNO microcontroller to interpret EEG signals and control various home appliances. By focusing on brainwave patterns such as alpha and beta waves, the prototype converts cognitive commands into control signals. Experimental results indicate promising responsiveness and accuracy, highlighting the potential for enhancing independent living through assistive BCI technologies.

Keywords: Brain- Computer Interface, Home Automation, EEG, Assistive Technology, Maker UNO, Bio amp EXG Pill

Abstract: In the present study, the wear analysis of an Indian railway wagon wheel (IRWW) was modelled using modelling software CATIA and applied various material combinations and tested for its performance and wear slippage at distinct load applications. In the real-world applications, wear produces wheel-surface slippage, resulting in deformation and movement of the wheel beneath the track surface. To address this issue, a thorough investigation of rolling contact on train wheels was undertaken to lessen the likelihood of failure. In the present investigation, the IRWW was initially designed and modelled in CATIA and uploaded to ANSYS to make the analysis. The stress generated by increasing contact load at rail-wheel assembly in terms of stress, strain, total deformation, and safety factor were determined for various load applications. Later, the acquired results were validated using the Artificial Neural Network (ANN) of Machine Learning (ML) Approach. The results showed that the overall deformation applied under various loads was within the limit.

Keywords: CATIA, ANSYS, stress, strain, total deformation, IRWW, ANN

Abstract: This project presents an AI-driven DDoS protection system that detects and mitigates HTTP-based attacks on cloud-hosted Apache web servers. Real-time network traffic is captured using Scapy, extracting features such as request count and time intervals. The Isolation Forest algorithm is used for unsupervised anomaly detection, enabling identification of malicious IPs without labeled attack data. Detected attackers are automatically blocked using iptables to maintain server performance. A Tkinter-based GUI dashboard provides live visualization of system health and traffic status for effective monitoring. Tested against simulated attacks like GoldenEye and Slowloris, the system achieves high accuracy with low false positives. Its lightweight and modular design makes it practical for cloud environments, with possibilities for future enhancements like encrypted traffic analysis and advanced AI integration.

Abstract: Mechanization can greatly improve farm productivity, but for many small and marginal farmers, buying expensive equipment is difficult. This survey paper introduces a practical solution—a cloud-based rental platform that connects farmers with equipment owners, making access to machinery more affordable and convenient. Instead of purchasing machines, farmers can rent what they need, when they need it. The platform is easy to use, with features like a simple registration process, clear equipment listings with all the details, real-time booking, and the option to hire trained operators. It also supports secure payments, offers help in both Kannada and English, and uses location-based search to show nearby equipment. Extra services like transportation support, dynamic pricing, data insights, customer help, and a shared knowledge hub make the platform even more useful. This paper explores how the system works, its potential benefits, and what it takes to implement it, ultimately aiming to empower farmers, increase efficiency, and bring digital change to rural India.

Keywords: Agricultural Mechanization, Cloud-Based Platform, Equipment Rental, Agri Tech, Operator Hiring System, Geolocation Services, Dynamic Pricing, Multilingual Support.

Abstract: Electric vehicles (EVs) are getting to be increasingly common, which has highlighted the pressing require for a reliable and helpful framework for charging them. Finding available and congruous charging stations is one of the greatest impediments EV clients experiences, which blocks the smooth move to eco-friendly shapes of transportation and includes to run uneasiness. In arrange to bolster the expanding request for EVs and empower their broad appropriation, this issue must be settled. In arrange to move forward the productivity and comfort of EV utilize, this consider portrays the plan, improvement, and appraisal of an application for finding EV charging stations. The application, which was made with the Vacillate system, gives clients up-to-date data on charging stations in their range by coordination the Google Maps API. In arrange to help clients in making well-informed choices, the app appears comprehensive station information, such as connector sorts, accessibility status, estimating, and operational conditions. A number of user-friendly highlights, counting filter-based station looks, course arranging with coordinates charging stops, and client announcing for out-of-service stations, are included within the application to advance move forward the client encounter. The interface is made to be simple to utilize, and route is made smooth by intuitively maps. To guarantee the precision and constancy of the data shown inside the app, real-time information is combined from a few sources. The application's usefulness and client interface were assessed through convenience testing. Concurring to the comes about, clients thought the app was exceptionally user-friendly and compelling. The app encourages a more consistent EV proprietorship encounter by radically bringing down stresses around the accessibility of charging stations. This think about fills a basic framework hole in modern transportation frameworks and progresses clean versatility arrangements.

Keywords: Electric Vehicles (EVs), Charging Station Finder, Flutter, Google Maps API, Slot Booking, Real-time Data, Firebase, Sustainable Transportation, Mobile App Development, EV Infrastructure.

Abstract: For the previous years, thousands of gym goers are looking for the solution to having an efficient and customized workout. Currently, the majority of users go through a poor posture while doing some exercises that leads to pain or decreased result. To fight this issue, the "AI-Based Workout Guide" is a technology solution which utilizes computer vision and AI to provide real-time posture correction and rep counting. The project utilizes AI and machine learning algorithms to determine users' body motion while performing a workout. Recorded by means of a camera or phone, the system will then compare this posture with the most common model ones for the optimal methods of performing the exercises. In case of a wrong posture, it will always provide immediate feedback regarding what one needs to change. Another aspect of the system is that the repetitions are automatically counted, and therefore no manual counting is needed and even greater focus on correct form by the user. The model has been trained on the database of different exercise poses, like squats, push-ups, and lunges. OpenPose or Mediapipe computer vision libraries are utilized to detect important landmarks, i.e., joint angles and alignment. In real-time, the system checks these milestones in order to give a correct posture analysis and rep count. Eventually, this would assist the users in becoming more efficient in their workouts, minimizing the possibility of injury, and achieving fitness objectives better. It's a readable, scalable AI-driven workout manual, from which it follows that it can be easily converted into any web or mobile application. Its usability reaches to beginners and intermediate fitness enthusiasts. This project demonstrates how technology can revolutionize personal training in fitness: the marriage of cutting-edge AI methods with a pragmatic solution for the implementation of fitness. Real-Time Feedback for Ongoing Improvement: The real-time feedback allows users to correct immediately, and therefore maintain ongoing improvement in workout performance. This instant advice discourages poor habits from developing, important to ongoing fitness gains. Increased Precision: The system takes advantage of leading machine learning methodologies, such as deep learning methodologies, to analyze and suggest precise posture adjustments against a huge dataset of correct postures during workouts. Precision is guaranteed while following complicated movements, and it's even able to modify the fit according to distinct users' variance in form.

Keywords: AI, Computer Vision, Workout Guide, Pose Estimation, MediaPipe, OpenCV, Exercise Form Correction.