Jo Ce 1 INTERNATIONAL CONFERENCE On EMERGING TRENDS IN COMPUTER SCIENCE & ENGINEERING ICETCSE-2025 March 6th - 7th, 20252025 International Conference on Emerging Trends in Computer Science and Engineering MESSAGE Dr. S.$ Inamdar Chief Patron Vice- President Vishwaniketan It gives me immense pleasure to welcome all participants, researchers, academicians, and industry professionals to the Ist International Conference on Emerging Trends in Computer Science and Engineering (ICETCSE 2025). This conference serves as a platform to foster knowledge exchange, innovation, and collaboration in the dynamic field of Computer Science and Engineering. Technological advancements are transforming our world at an unprecedented pace. This conference aims to bring together brilliant minds to discuss cutting-edge research, emerging trends, and real-world applications that will shape the future of computing. I am confident that the deliberations and discussions in this event will contribute significantly to the advancement of technology and academia I congratulate the organizers of the Ist International Conference on Emerging Trends in Computer Science and Engineering (ICETCSE 2025) for bringing together researchers, academicians, and industry experts on a common platform. ICT has revolutionized the way we live, work, and communicate, Today, it is impossible to imagine a world without artificial intelligence, cloud computing, cybersecurity, and other emerging technologies Conferences like these are essential for fostering innovation, collaboration, and knowledge exchange, enabling India to contribute significantly to global technological advancements. I am confident that this conference will serve as a valuable forum for discussing cutting-edge research and exploring future possibilities in computing | wish the conference great success in achieving its objectives.2025 International Conference on Emerging Trends in Computer Science and Engineering MESSAGE Dr B. R. Pa Conference Chair Principal Vishwaniketan’s iMEET It is with great pleasure that I welcome all participants to the Ist Intemational Conference on Emerging Trends in Computer Science and Engineering (ICETCSE 2025). This conference serves as a vital platform for researchers, academicians, and industry professionals to exchange ideas, present innovations, and discuss the latest advancements in computing and engineering In today’s fast-paced digital era, emerging technologies such as artificial intelligence, data science, eybersecurity, blockchain, the Internet of Things, and cloud computing are reshaping industries and societies, As we move toward an increasingly technology-driven future, it becomes imperative to explore innovative solutions and interdiseiplinary approaches to address real-world challenges. This conference is designed to encourage such discussions and provide a collaborative environment for knowledge sharing and intellectual growth ‘Through this event, we aim to inspire groundbreaking research, foster academic and industry collaborations, and contribute to the ever-evolving field of Computer Science and Engineering. I am confident that ICETCSE 2025 will serve as a catalyst for innovation, enabling participants to gain valuable insights and contribute meaningfully to technological advancements. Textend my sincere gratitude to the organizing committee, distinguished speakers, research scholars, industry professionals, and all participants for their invaluable contributions, 1 appreciate the relentless efforts of the team in making this conference a reality. I am confident that the discussions and deliberations during ICETCSE 2025 will ‘open new avenues for research and development, ultimately benefiting the global academic and scientific community. Wishing you all a successful, insightful, and enriching conference experience!2025 International Conference on Emerging Trends in Computer Science and Engineering MESSAGE Dr S.B. Kadam Conference Co-Chair IQAC Director and Vice Principal Vishwaniketan’s iMEET It is my immense pleasure to welcome you all to the Ist International Conference ‘on Emerging Trends in Computer Science and Engineering (ICETCSE 2025). This conference is a significant step towards fostering collaboration among researchers, academicians, industry experts, and students in the rapidly evolving domain of Computer Science and Engineering ‘Technology is advancing at an unprecedented pace, influencing every aspect of our lives. Emerging fields such as artificial intelligence, machine learning, cybersecurity, cloud computing, and data science are redefining industries and creating new opportunities. ICETCSE 2025 provides a dynamic platform to discuss pioneering research, explore innovative solutions, and address the challenges of the digital era The objective of this conference is to bring together brilliant minds to share their expertise, engage in meaningful discussions, and contribute to technological progress. By facilitating knowledge exchange between academia and industry, we aim to inspire innovative ideas that will shape the future of computing and engineering I extend my heartfelt gratitude to our esteemed speakers, paper presenters, reviewers, organizing committee members, and all participants for their dedication and contributions, Your participation is what makes this event a success. I sincerely hope that this conference provides valuable insights, fosters collaborations, and encourages new perspectives in research and development. Wishing you all an enriching and successful conference.2025 International Conference on Emerging Trends in Computer Science and Engineering MESSAGE Prof. Shivani Agrawal Conference Co-Chair Vice Principal & FE Incharge Vishwaniketan’s iMEET Welcome Address — ICETCSE 2025 Dear Distinguished Guests, Researchers, and Participants, It is my privilege to welcome you to the Ist International Conference on Emerging Trends in Computer Science and Engineering (ICETCSE 2025). This conference is platform for exchanging ideas, showcasing innovation, and addressing the challenges shaping the future of technology. In an era driven by artificial intelligence, machine learning, cybersecurity, blockchain, and cloud computing, we are witnessing unprecedented transformation. These advancements offer immense opportunities but also bring complex challenges that require collaboration across disciplines. ICETCSE 2025 is designed to foster meaningful discussions, inspire innovation, and drive impactful research This conference is more than just paper presentations—it is a space where ideas take root, partnerships form, and solutions emerge. By uniting experts, scholars, and industry leaders, we aim to accelerate progress in Computer Science and Engineering I extend my sincere gratitude to the organizing committee, technical team, keynote speakers, authors, reviewers, and participants for making this event possible. I am confident that the insights shared here will leave a lasting impact. Wishing you all a thought-provoking, collaborative, and successful ICETCSE 2025!2025 International Conference on Emerging Trends in Computer Science and Engineering MESSAGE Dr. Ankush Pawar Conference Convener HOD CSE(AI&ML) ‘Vishwaniketan’s iMEET It is my great honor and privilege to welcome all esteemed participants, researchers, academicians, and industry professionals to the Ist International Conference on Emerging Trends in Computer Science and Engineering (ICETCSE 2025). As the Head of the Department of Computer Science & AIML and Convener of this prestigious conference, I am thrilled to see a gathering of intellectuals, innovators, and thought leaders coming together to explore the frontiers of computing technology. The rapid evolution of Artificial Intelligence, Machine Learning, Quantum Computing, Cybersecurity, Blockchain, IoT, and Edge Computing is reshaping industries and redefining the way we interact with technology. This conference serves as a erucial platform to share pioneering research, discuss emerging trends, and collaborate on solutions that will drive the future of computer science. I firmly believe that the knowledge exchange and discussions taking place at ICETCSE 2025 will play a pivotal role in advancing both academic research and industrial applications. I extend my heartfelt appreciation to all authors, speakers, panelists, and participants who have contributed their expertise and insights to make this conference a success. My sincere gratitude also goes to the organizing committee for their dedication in bringing together a diverse and distinguished set of minds on one platform. Conferences like ICETCSE 2025 are essential in fostering innovation, interdisciplinary collaboration, and technological progress, I am confident that this event will inspire groundbreaking ideas and open new avenues for research and development in computer science and engineering, I wish all participants a highly engaging, insightful, and productive conference. May this gathering mark the beginning of many fruitful collaborations and advancements in the field of computing2025 International Conference on Emerging Trends in Computer Science and Engineering MESSAGE Prof. Ashwini V. Garole Conference Co-Convener Assistant Professor CSE(AI&ML) Vishwaniketan’s MEET As the co-convener of the International Conference on Emerging Trends in ‘Computer Science & Engineering, I would like to express my heartfelt gratitude to all the dignitaries, speakers, participants, and organizing team whose unwavering support and dedicated efforts have contributed to the success of this event. I extend my sincere thanks to our esteemed keynote speakers and panelists for sharing their invaluable insights, which have undoubtedly enriched our understanding of the latest advancements and challenges in the field of Computer Science and Engineering A special note of appreciation to all the authors and presenters for their enthusiastic participation and for showcasing innovative research that reflects the spirit of exploration and technological advancement. I would also like to acknowledge the relentless efforts of the organizing committee, technical team, and volunteers who worked tirelessly to ensure the smooth execution of this conference Together, we have created a platform that fosters learning, collaboration, and innovation, and 1 look forward to continued contributions and partnerships in future endeavors. Thank you for your participation, and I look forward to the i that this conference will undoubtedly inspire. ightful exchanges2025 Intemational Conference on Emerging Trends in Computer Science And Engineering Advisory Committee Dr. Daniel Goldsmith, Professor, Coventry University, UK Dr. Adam Worrallo, Professor, Coventry University, UK Dr. Davide Fucci, Associate Professor, BTH Karlskrona, Sweden Dr. Rahul Mohanani, Assistant Professor, University of Jyvaskyla, Finland Dr. Naresh Marturi, Deputy Director, Robotics Dept, Birmingham University, UK Dr. Pranav Pawar, Assistant Professor, BITS Pilani, Dubai Dr. S. K. Shinde, Principal, LTCOE, Koparkhairane Dr. J. W. Bakal, Principal, HOC Pillai COE, Rasayani Dr. Deven Shah, Principal, SLRTCE, Mumbai Dr. S. M. Joshi, Principal, Pillai COE, Panvel Dr. Alam Shaikh, Principal, VPPCOE, Sion, Mumbai Dr. S.D. Nawale, Principal, Sinhagad COE, Solapur Dr. Vivek Sunnapwar, Principal, KISIT, Mumbai Dr S. M. Khot, Principal, FCRIT, Vashi Dr. V. J. Pillewan, Principal, KGCE, Karjat Dr. Sandhya D, Jadhav, Principal, BVCOE, Kharghar, Dr. Mashukar Lengare, Principal, MIT COE, Dr. Narendra Shekokar, Vice-Principal, DISCOE, Mumbai Dr. Shashikant Ghumbre, Professor, GOV. COE, Avasari, Pune Dr. A. W. Kiwaleker, Registrar, DBATU, Lonere Dr. Sunil Mane, Associate Professor, COEP, Pune Dr. Madhav Chandane, Associate Professor, VITI, Mumbai Dr. Shashikant Patil, IEEE Secretary, Mumbai Section Dr. Pinki Roy, HOD CSE, NIT, Silchar Dr. S. M, Patil, Professor, IGCOE, Koparkhairane Dr. Amol Adamuthe, HOD IT, RIT, Sangali Dr. Ashok Kante, Dean R & D, FCR COE, Mumbai Dr. Subhash Pingle, HOD CSE, SKN Sinhagad, Pandharpur Dr. Dinesh Puri, HOD, SSBT COE, Jalgaon Dr. Chetan Chaudhari, Associate Professor Raisoni COE, Jalgaon Dr. Dhananjay Kalbande, Professor & HOD CSE, SPIT, Mumbai Dr. Ankit Sanghvi, HOD CSE, ARMIET, Shahapur Dr. Bhushan Chaudhari, HOD CSE, SVKM'S, Dhule Dr. Vikas Shinde, Director PBLCOE, Vishwaniketan Prof. Sunita Jadhav, Director, T&P, Vishwaniketan Dr. Ambuj Kumar , Director, BMCOE, Vishwaniketan Prof. Pallavi Dongre,Director,UG Fellowship, Vishwaniketan2025 Intemational Conference on Emerging Trends in Computer Science And Engineering Organizing Committee Prof. Sneha Nimbare Prof. Aqsa Sarnaik Assistant Professor, CSE(AL&ML) Assistant Professor, CSE(AI&ML) Vishwaniketan’s MEET Vishwaniketan’s MEET os Prof Hemant Kamble Prof. Mahesh Sutar Assistant Professor, CSE(AL&ML) Assistant Professor, CSE(AL&ML) ‘Vishwaniketan’s MEET Vishwaniketan’s MEET Prof. Rutika Mohite Prof. Sanika Sutar Assistant Professor, CSE(AL&ML) Assistant Professor, CSE(AI&ML) Vishwaniketan’s MEET Vishwaniketan’s MEET Prof. Rohan Patil Assistant Professor, CSE(AL&ML) \Vishwaniketan’s MEET2025 Intemational Conference on Emerging Trends in Computer Science And Engineering INDEX [Link]. | Paper ID Title Page No. 1 | AIMLO1 | VaniConnect (Sign Language to voice transcriber) 1 2 | AIMLO2_ | CMS Detector Quality Assurance using Machine Learning 1 3. | AIML03 | Predicting Air Quality Index across Mumbai using Multiple Linear 2 Regression models 4 AIMLO7 | Advancements in Non-Invasive BCI Systems for Robotic Arm 3 Control: EEG Signals and Machine Learning Algorithms 5 | AIML08_ | Comprehensive Analysis of Machine Learning and Deep Learning 3 Models for Breast Cancer Detection 6 AIML12 | ChurSpark: Predicting Customer Churn in Telecommunications 4 through Data Analysis and Machine Learning 7 | AIML15 | Advanced Digital Asset Management through Object Detection and 5 Captioning Models 8 AIML18 | Color Blindness :A Comprehensive Review 5 9 AIML20 | Real-Time Social Media Analytic Dashboard 6 10 | AIML21 | CrossTalk - Seamless Language Translation a i AIML22 | Integrating Gamification and NLP: Analysis On An Adaptive 7 Learning Platform for Engineering Education 12. | AIML23 | DermaTech: A Machine Learning and Deep Learning-Based Skincare | 8 Product Recommendation System 13 AIML24 | EvalMate: Handwritten Assessment Grading Tool Using LLM 9 Feedback 14 | AIML28_| Fake Social Media Bot Detection and Reporting 9 15 AIML29 | Forensic Age and Gender Estimation Using Deep Learning: A 10 Comparative Analysis of CNN, VGG-16, ResNet, and EfficientNet-B3 16 AIML30_ | MOODMATE: Bridging Technology and Mental Health W 17 | AIML31 | Anomaly Detection Using Machine Learning: Evaluating} 11 Autoencoder, ResNet50, and PatchCore on Industrial Defects2025 Intemational Conference on Emerging Trends in Computer Science And Engineering 18 AIML32_ | Forensic Voice Analysis Using MFCC and SVM Enhancing Speaker | 12 Identification and Verification 19 | AIML33_ | Deep Learning-Integrated loT Healthcare Monitoring System 13 20 | AIML36 | An Exhaustive Survey of Big Data Storage Reduction Techniques 13 21 AIML37 | "Deep Learning Approaches for ‘Text Summarization: A] 14 Comprehensive Survey" 22 | AIML43_ | Advancements in Deep Federated Learning for Autonomous Vehicles: | 15 Enhancing Real-Time Decision-Making Capabilities 23. | AIML4S_ | AyurGenixAl: An Ayurvedic Medication Advisor 16 24 AIML46_ | LearnSmart: Tool to Extract Key Concepts and Arguments from} 16 YouTube Lectures 25 | AIML47 |A GAN-Based Text-to-lmage Framework for Automated Comic | 17 Generation and Illustration 26 | AIML49 | Consensus of hourly visibility forecast for airports using NWP model | 18 and observations 27 AIMLS1 | Generative Al Approaches for Text Generation: Techniques, 19 Challenges, and Future Directions 28 | AIMLS3 | Enhancing Travel Itinerary Planning with Generative Al and} 19 Real-Time Weather Data 29 AIML54 | Enhancing Travel Itinerary Planning with Generative Al and| 20 Real-Time Weather Data 30 | AIMLSS | E-Learning Portal for GMVIT 21 31 | AIMLS8_ | Transforming Dyslexia Education: A Comparative Study of the 21 Personalized Learning Systems 32 | AIML60 | Cardiovascular disease prediction using ECG images and using ML 22 33 AIML62_ | From Rule-Based Systems to Transformers: The Evolution of NLP 23 Techniques, Applications, Limitations, and Challenges Across Industries 34 AIML65_ | Advancements In Deep Learning For Alzheimer’s Disease Detection 24 35 AIML67 | Image Captioning for blind People using Deep learning CNN-LSTM | 24 Technique2025 Intemational Conference on Emerging Trends in Computer Science And Engineering 36 | AIML68 | Real-Time Dance Performance Evaluation and Feedback System} 25 Using Motion Tracking and Pose Estimation 37 | AIML69 | AgroConnect: A Data-Driven Crop Recommendation System Using } 26 Machine Learning For Smart Farming 38 | AIML70 | Automatic Question Generation using Maximum Marginal Relevance | 27 Algorithm and TS transformer 39 | AIML73 |The Role of Deep Learning in Credit Card Fraud Detection: A} 28 State-of-the-Art Review 40 | AIML74 | ANKUR: A Crop Disease Prediction System using Machine Learning | 28 41 | AIML75 |Green Thumb Network: Urban gardening, home plant care and] 29 Community Platform. 42 | AIML76 | Optimized Fault Diagnosis in Three-Phase Voltage Source Inverters | 29 Using ANN and Machine Learning Techniques 43. | AIML78_ | AI Embedded QR Code Generator and Decoder 30 44 | AIML81_ | Sentiment analysis using pre-trained NLTK models BERT, TextBlob, | 31 Flair, VADER & RoBERTa with personalized results. 45 | AIML82 | User Recommendation System based on Project Metadata using | 32 Ensemble Learning 46 | AIML83_ | Ocular Biomarkers For Diseases Detection 32 47 | AIML84 | Deciphering Language using NLP Techniques 3 48 | AIML86_| Personalised Healthcare Recommendation System 34 49 | AIML87_ | Word Based Indian Sign Language Translator Using Seq-2-Seq Model | 35 50 | AIML88 | Machine Learning Approaches For Predicting Safe Landings Of The} 35 SpaceX Falcon 9 First Stage. $1 | AIML91 | Enhanced Structural Health Monitoring Through LSTM-Enhanced} 36 Gradient Boosting Regressor. 52 | AIML92 |The Real-Time Emotion and Mental State Monitoring System for| 36 ‘Train Drivers 53 | AIML96 | Deep Learning-Based Approach for Ambulance Detection & Traffic | 37 Management 34 | AIML98 | Integrated Learning Platform for Summarization, Quizzes and 38 Flashcards2025 Intemational Conference on Emerging Trends in Computer Science And Engineering bE] AIMLI101 | Predicting Social Media Addiction : Patterns and Consequences 39 56 | AIML102 | Augmented Reality Based Text Visualizer 39 57 | AIML103 | News Portal 40 58 | AIML106 | Real-Time Language Translation 40 59 | AIML107 | Advanced Text Extraction from Image and Voice Conversation 41 60 | AIML108 | EcoSort : Community-Powered E-Waste Detection & Recycling 2 61 | AIML109 | DealMate: A Conversia Chatbot Revolutionizing E-commerce 42 62 | AIML110 | Jarvis: Voice Assistant B 63 | AIMLII1 | Mock Interview Assistant (NLP-based Interviewer Bot) 4B 64 | AIMLI12 |Emo-Therapist Doctor Recommendation: Facial Recognition and | 45 ‘Chatbot 65 | AIML113 | Vikarshan: A Deep Leaming Approach for Plant Disease| 45 Identification 66 | AIML114 | Trinetra : A Deep Learning Approach for Eye Disease Recognition. 46 67 | AIMLII5 | [Link] (optimising household electricity through machine learning | 47 and data science) 68 | AIML116 | Detection And Classification of Copd Using Deep Learning 47 69 AIML117 | Implementation of healthcare GYANT using MERN Stack 48 70 | AIML118 | Crop Yield Prediction A Machine Learning Approach To Agriculture 49 71 | AIML119 | AyurSage: A Personalized Ayurvedic Healthcare Platform Leveraging | 49 Classical Wisdom and Modern Technology nR AIML120 | An Experimental Investigation on Data Mining Analysis of Academic 50 Performance and Record Technique Using Prediction Model Algorithms 73 | AIMLI21 | Agrosmart Ledger - Decentralized Marketplace for Demand and} $1 Supply. 74 | AIML122 | Quick Cash: A Membership-Based Emergency Financial Assistance 52 System 75 | AIML123 | Medical Image Segmentation with U-Net: Improving Diagnostic 33 Accuracy Through Deep Learning2025 Intemational Conference on Emerging Trends in Computer Science And Engineering 76 | AIML124 | Developing a Neural Network Model for Multi-Label Classification | $3 of Patient Readmission Rates 77 | AIML125 | A Comparative Study of Pre-Trained Convolutional Neural Networks | 54 for Feature Extraction in Computer Vision Tasks 78 | SSCO3_ |} A Survey of techniques for automatic malware Identification and| 35 Classification 79 | SSC04 | Vital Link Organ Donation Platform Using Blockchain Technology 55 80 SSCO7___|Enhancing Ransomware Protection through Moving Target 56 Defense Technique 81 SSCO8 _| Property Registration System using Blockehain 37 82 | SsCo9 | A Framework for Data Transmission to dedicated receiver in Visible | $7 Light Communication Network 83 | SSC10_| Cryptographic Technique To Secure Data In Rest 58 84 | SSCII__ | Security Framework for Cloud Servers in DevSecOps 38 85 IOTOI | Waste segregation and monitoring 59 86 | I0T03 | Fruit Detection System using Ultrasonic and Color Sensor 60 87 | 10704 | Pothole Detection System 60 88 IOTOS _ | Cattle Health Monitoring System 62 89 | 10707 _| Systematic Review of intrusion Detection in loT Botnet 62 90 1OTOS FusionFlux - Local Copilots for Linux systems 63 91 10109 _| Smart Highway Road Side Symbol Detection 63 92 | 10710 | Smart Intravenous with Raspberry Pi Monitoring System 64 93 1oTIL Smart Farm Assist: An loT-Based AI Solution for Precision 65 Agriculture 94 | 1OTI3 | Review of 6G Wireless Communication, Applications, and Use| 65 Cases 95 1OT14 | Performance Analysis of Fiber Bragg Grating Sensor with Varying 66 Grating Lengths for Medical Application 96 | IOTIS | Discrimination of Internal Faults and Other Transients in Power 67 Transformers using Tree based Classifiers2025 Intemational Conference on Emerging Trends in Computer Science And Engineering 97 RTIO1 | Integrating Multi-Criteria Decision-Making and the Taguchi Method 67 for High-Performance Material Selection in Rigid Flange Coupling Systems 98 RTI03__| Training & Placement Management System 68 99 RT105 | Carbon Footprint Calculator 68 100 | RT106 | Optimizing Auction Systems using Blockchain Technology 69 101 RTIO7 UdyogAl — Al-Driven Marketing Automation Platform 70 102 RTIO9 Academic Suite 70 103 RTH2 aoe: An Intelligent Warehouse Management S: 7 104, RTH3 | WatchHub: Unified Platform For Synchronized Video And Social | 72 Interaction 105 | RTHS | Sign Language Recognition and Translation System Using Machine | 72 Learning 106 | RTH6 _ | Simple Online Grocery Ordering System B 107 | RTH9 | Data-Verse Insight; Web-Application for Algorithm Visualization B 108 | R120 | Complexity to Efficiency: A Review of SCM ‘Transition Using] 74 Emerging Technologies 109 | RTI21_ | Multi- Factor Optimization of Burr Formation in Aluminium Sheet | 75 Blanking via Hybrid Taguchi Design and Predictive Regression Modelling 110 | RTI22 | Collaborative Coding Platform: Real-Time Code Editor 8 11 RTI25__ | AbilityHire : Implementation on employment for differently-abled 76 112 | RTI27__ | Blockchain Freight Network: A Decentralized Solution for Supply | 77 Chain Optimization 113. | RTI28 | Blockchain and Biometric based Student Identity Verification System | 78 with DigiLocker Integration 114 | OTHERO! | Model-Based Optimization of Hydrodynamic Characteristics in B Multiphase Reactors Using Reinforcement Learning with Markov Decision Processes 115 | OTHERO3 | Design & Develop Assessment Method For Learning Outcomes 792025 Intemational Conference on Emerging Trends in Computer Science And Engineering 116 | OTHERO4 | Swastika: A Comprehensive Mental Health App for Mood Tracking, | 80 Stress Relief, and Community Support 117 | OTHEROS | Comparative Study of Radiotracer’s Density Count Measurements | 81 Using FFT, Wavelet and Harrlike ‘Transforms Using Renal Scintigraphy Images 118 | OTHEROT | Swastika: Organ Donation System using mern stack 82 119 | OTHEROS | Improving the Efficiency of Hydraulic Ram Pump 822025 Intemational Conference on Emerging Trends in Computer Science And Engineering Paper id_AIMLO1 Title of the Article VaniConnect (Sign Language to voice transcriber) Authors Anu Sophiya Lingadurai, Ayush Sukhadan Nair, Ayush Sukhadan Nair, Poorva Surendra Raut, Prof Supriya Joshi, Melissa Rozario Sequeira [Link] Rodrigues Institute of Technology, Vashi Abstract The "Sign Language to Voice Transcriber" project aims to bridge communication gaps between individuals who use sign language and those who do not understand it proficiently Sis individuals worldwide. However, effective communication with non-signers, such as family members, colleagues, or healthcare providers, can be challenging without an interpreter present. This project proposes the development of a technological solution that translates real-time sign language gestures into spoken [Link] system will utilize computer vision techniques to capture and interpret sign language gestures from video input. Machine learning algorithms, particularly deep learning models, will be employed to recognize and classify these gestures accurately. a machine learning model trained on a diverse dataset of sign language gestures recognizes these gestures and translates them into text. By providing a reliable means of translating sign language into spoken language, this technology has the potential to promote inclusivity and reduce communication barriers in diverse social contexts, The proposed SLAT system aims to empower deaf individuals by providing them with an accessible means to interact with the spoken language environment independently. By leveraging advances in computer vision, machine learning, and natural language processing, the SLAT system represents a significant step towards fostering inclusivity and accessibility in communication for the deaf community language is a primary mode of communication for many deaf and hard-of-hearing Keywords Sign language, Machine Learning, computer vision, inclusivity, hard-of-hearing individuals. Paper id_AIMLO2 Title of the Article CMS Detector Quality Assurance using Machine Learning Authors Ayush Rakesh Gerra!, Richard Joseph', Shashikant Dugad’, Anisha Rohra', Sneha Tanna’, Preethika Shetty', Gagan Mohanty’, Irfan Mirza’, Kameswara Rao?2025 Intemational Conference on Emerging Trends in Computer Science And Engineering 1 Vivekanand Education Society's Institute of Technology, Mumbai, India 2 Indian Institute of Science Education and Research, Mohali, Punjab, India 3 Tata Institute of Fundamental Research, Colaba, Mumbai, India Abstract The Large Hadron Collider (LHC) at CERN accelerates proton beams to 7 GeV, with collisions recorded by the Compact Muon Solenoid (CMS) detector. For the CMS upgrade, 27,000 hexaboards are being manufactured for the High-Granularity Calorimeter (HGCAL), requiring efficient inspection of 700 checkpoints per module. Traditional Coordinate Measuring Machine (CMM) methods were slow, prompting the evaluation of flatbed scanners. While the HP MFP M277SDN was unsuitable, the EPSON v850 proved effective An inspection framework combining YOLOv8 and Hough transformations reduced inspection time to 2m 6s + 692ms per board at 3200 DPI, with individual hole extraction taking 21s 4 330ms. This novel system enhances Quality Assurance, improving scalability and accuracy in large-scale manufacturing for the CMS Detector upgrade. Keywords Automated Visual Inspection, CMS, Object Detection, Hough Transform, CHT, YOLOv8 Paper id_AIML03 le of the Article Predicting Air Quality Index across Mumbai using Multiple Linear Regression models Authors Prathmesh Kadam, Kartikeya Dangat, Jainam Varia, Dr. Shreya Patankar KJSIT Abstract Background: According to a TOI article, Breathing ailments have claimed an enormous amount of lives in the recent history of Mumbai, To help citizens reduce their breathing problems and take better care of their respiratory system, we decided to undertake this project. Methods: We aim to develop a predictive model for the Air Quality Index (AQI) in Mumbai by leveraging linear regression techniques. This study utilizes historical air quality data for the year of 2023 of 4 different locations spread out across Mumbai, including various pollutants such as PM2.5, PM10, NO2, and CO, to train the linear regression model Results: After all processing and predictions, the MSE was found to be 158.55 and R-squared to be 0.8247 for the classic Linear Regression model. The MSE and R-squared was determined to be 158.48 and 0.8248,respectively for Ridge Regression and for Polynomial Regression, the MSE and R-squared was found out to be 114.41 and 0.8735, respectively Conclusions: We can conclude that linear regression offers a foundational approach to predicting AQI, its limitations such as unpredictability of non-linear relationships ask for exploration of more advanced methodologies2025 Intemational Conference on Emerging Trends in Computer Science And Engineering Keywords Air Quality Index, Mumbai, Linear Regression, Machine Learning Paper id_AIMLO7 le of the Article Advancements in Non-Invasive BCI Systems for Robotic Arm Control: EEG Signals and Machine Learning Algorithms Author Aditya Deshbhartar, Nitesh Ghodichor, Lokesh Manapure, Aditya Wankhede, Atul Kumar Choudhary, Shival Jagdish Priyadarshini College of Engineering Nagpur, india Abstract This design trials to develop an innovative prosthetic hand that's controlled by EEG signals, exercising the BioAmp EXG Pill for real- time neural signal accession. using sophisticated machine knowledge ways, the system aims to restate brain exertion into practicable motor commands, enhancing stoner control. The design will employ MNE- Python and Deep EEG libraries for effective signal processing and type, landing EEG signals that will be strictly preprocessed to count noise and leftovers, therefore icing high- quality data for analysis. The machine knowledge model will be trained to fete specific patterns of brain application associated with colorful galette movements, allowing the prosthetic hand to respond directly to the stoner’s intentions. A critical element of this design is the establishment of a flawless communication interface between the laptop and the Arduino, easing real- time control of the servo motors rested on the classified EEG signals. eventually, this design aspires to rest the gap between neural engineering and prosthetic technology, furnishing a promising result for individualities with branch loss, and enabling further natural and intuitive control of prosthetic bias, thereby perfecting the quality of life and independence for addicts, while paving the way for farther advancements in brain- computer interfaces and assistive technologies Paper id_Al 1L08 Title of the Article Comprehensive Analysis of Machine Learning and Deep Learning Models for Breast Cancer Detection Authors Nirzaree Devasthale, Yukta Saraf,Deepali Patil, Sudhir Dhekane SVKM’s Dwarkadas J. Sanghvi College of Engineering Mumbai, India Abstract ‘The research highlights the performance of various machine learning and deep leaning architectures, including XGBoost, Decision Tree, Stacking Ensemble, CNN, ADA Boost,2025 Intemational Conference on Emerging Trends in Computer Science And Engineering ete., for breast cancer detection. We have used the Breast Cancer Wisconsin dataset, evaluated the performance for each model, and created a classification report encompassing. various metrics such as accuracy, precision, recall and Fl-score. Our findings show that ensemble methods, particularly Stacking Ensemble and ADABoost classifiers, outperform individual models, while deep learning models such as CNNs offer superior precision in capturing complex data patterns. Along with that, we can also see the models that are not as efficient in capturing the dependencies in the data such as SimpleRNN, XGBoost and Decision Tree Keywords Breast Cancer, Machine Learning, Breast Cancer Wisconsin Dataset, Deep Learning, Ensemble Methods, Comparative Analysis, Model Performance, Data Pattern Recognition Paper id_AIMLI2 itle of the Article ChurnSpark: Predicting Customer Churn in Telecommunications through Data Analysis and Machine Learning Authors Ajinkya Bhosale, Saarim Kazi, Aahad Khan, Shambhavi Ingole Ramrao Adik Institute of Technology Abstract Background: Customer churn poses a significant challenge in the telecommunications industry, leading to revenue loss and increased costs associated with acquiring new clients This study introduces ChurnSpark, a predictive analytics platform designed to identify customers at risk of churn and develop effective retention strategies. By leveraging exploratory data analysis (EDA) and machine learning techniques, ChunSpark aims to uncover key factors contributing to customer attrition and enhance retention efforts within the telecommunications sector. Methods: ChurnSpark utilizes a comprehensive telecommunications dataset to conduct exploratory data analysis, focusing on various factors such as payment preferences, contract terms, and service consumption patterns, The platform employs several machine learning algorithms, including Random Forest Classifiers, Decision Trees, and Logistic Regression, to predict customer [Link] address the challenge of data imbalance inherent in churn datasets, ChurnSpark implements SMOTE-ENN, a hybrid technique combining Synthetic Minority Over-sampling Technique (SMOTE) and Edited Nearest Neighbors (ENN). This approach ensures that the models can effectively learn from both churned and non-churned customer classes, Additionally, Principal Component Analysis (PCA) is applied for dimensionality reduction, optimizing model performance and minimizing overfitting Results: The results show that applying SMOTE-ENN significantly boosted the Random Forest Classifier’s accuracy to 94%, The Decision Tree model also saw improvements in precision, recall, and F1 score with SMOTE-ENN, demonstrating its effectiveness in addressing imbalanced churn datasets.2025 Intemational Conference on Emerging Trends in Computer Science And Engineering Conclusions: ChurnSpark is proven to be an effective tool for predicting customer churn by leveraging machine learning and data preprocessing techniques like SMOTE-ENN. The study highlights the importance of data-driven approaches in retention strategies. Future improvements could involve advanced feature selection methods and testing the model on larger, more diverse datasets. These findings are crucial for telecom businesses looking to enhance churn prediction and retention efforts. Keywords Customer Churn, Telecommunications, Predictive Analytics, Machine Learning, Exploratory Data Analysis, SMOTEENN, Principal Component Analysis, RandomForest, Retention Strategy, Data Imbalance Paper id_AIMLIS, Title of the Article Advanced Digital Asset Management through Object Detection and Captioning Models Authors Vihit Khetle, Rupesh Raut, Atharva Desai Stevina Correia, Neha Katre DJSCE, Mumbai, India Abstract In this study, we present an innovative digital asset management system that leverages advanced object detection and captioning models to enhance the organization and retrieval of visual content. By comparing and evaluating various state-of-the-art object detection models, including Facebook's DETR-ResNet-101 and ViT-GPT2, we identified the most accurate and efficient combinations for object detection and caption generation, Our results demonstrate that the selected models, particularly Facebook's DETR-ResNet-101, significantly ‘outperform other contenders in object detection accuracy. The proposed system integrates these models with cloud storage services and a metadata management database, providing a comprehensive solution for managing digital assets. Our findings indicate that this integrated approach not only enhances user experience but also promotes scalability as the volume of digital assets grows. Open-soureing the source code and datasets enhances reproducibility and fosters academic exchange within the computer vision community by providing all relevant materials on GitHub, complete with detailed usage instructions and versioned DOI information: Github Link Keywords Computer Vision, Object Detection, Image Tagging, Model Comparison, Comparative Analysis, Neural Networks Paper id_AIMLI8 Title of the Article "Color Blindness :A Comprehensive Review"2025 Intemational Conference on Emerging Trends in Computer Science And Engineering Authors Akshat Gandhi, Abbas Bootwala,Krish Doshi, Dr. Madhura Phadke [Link] Institute of Technology Abstract Background: Color vision deficiency, commonly referred to as color blindness, is an impairment of visual perception that impairs the ability to accurately identify and distinguish certain colors. This condition has a significant impact on the quality of life of individuals with color vision deficiency, and the use of technology and computational approaches has enabled a surge in research to address this issue Methods: The review focuses on the exciting progress made in the field, from deep learning-based approaches to generative adversarial networks (GANS), real-time applications, medical image correction, and the integration of transformer networks and diffusion models The integration of these innovations into everyday life, including augmented reality applications, promises a more colorful, inclusive, and equitable world. Results: various challenges remain, such as data availability and unreliability, image quality, and real-time processing, We highlight these gaps and call for further research to refine and expand these approaches, ensuring they are accessible and effective. Conclusions: ‘The paper underscores the trans-formative potential of AI/ML in reshaping the way we perceive and experience the world, with a focus on creating an inclusive environment for individuals with color vision deficiencies Keywords Color blindness correction Astificial Intelligence (Al) Machine Learning (ML) Color vision deficiency Deep learning Paper id_AIML20 Title of the Article Real-Time Social Media Analytic Dashboard Authors Esha Pal, Soham Niungare, Prof. Nita Ingale, Sudarshan Dongre, Nishad Bhujbal Vasantdada Patil Pratishthan’s College of Engineering, University of Mumbai, Mumbai, India Abstract The X Real-Time Social Media Analytics Dashboard helps users analyze and improve their X presence using machine learning and data analytics. It provides real-time insights into user interactions, sentiment, engagement, trending topics, and tweet performance. The dashboard shows how content resonates with followers by tracking audience behavior and highlighting key patterns, Users can identify what’s working, and what needs improvement, and refine their social media strategies accordingly. With data-driven recommendations, it helps create2025 Intemational Conference on Emerging Trends in Computer Science And Engineering more engaging and effective content, This report covers the dashboards design, implementation, and impact in enhancing X engagement. Keywords X Analytics, Real-Time Insights, Machine Learning, Data Analytics, Sentiment Analysis, ‘Trending Topics, Tweet Performance, Social Media Strategy, Pattem Identification Paper id_AIML21 le of the Article CrossTalk - Seamless Language Translation Authors Nidhi Poojary, Saniya Sonawane, Darshit Visariya, Vishnav Tuli, Chirag Ramesh Desai Somaiya Vidyavihar University Abstract ‘The CrossTalk project addresses the growing demand for seamless, real-time multilingual communication. It integrates advanced Natural Language Processing (NLP) techniques to provide accurate, real-time translations for both text and speech, thereby enhancing cross-cultural interactions in business, education, and social settings. By leveraging state-of-the-art machine learning models such as transformer-based Neural Machine Translation (NMT) and speech-to-text algorithms, CrossTalk ensures high-quality translations with minimal latency. The project emphasizes translation accuracy, real-time performance, and ease of use, demonstrating how NLP-driven translation services can improve global communication. Future expansions include additional language support, enhanced conversational AI, and an optimized user interface for broader accessibility Keywords CrossTalk, multilingual communication, real-time translation, natural language processing (NLP), speech-to-text, machine translation, neural machine translation (NMT), transformer models, scalability, conversational Al Paper id_AIML22 Title of the Article Integrating Gamification and NLP: Analysis On An Adaptive Learning Platform for Engineering Education Authors Siddhesh Thakarkar, Mohak Tamore, Prathmesh Tiwari, Vikrant Agaskar Vidyavardhini’s College of Engineering & Technology Abstract2025 Intemational Conference on Emerging Trends in Computer Science And Engineering + Background: The rapidly evolving field of engineering education necessitates tools that enhance student engagement and learning outcomes, Traditional learning methods often struggle to motivate students consistently, especially in technical disciplines. This study introduces a gamified learning platform tailored for engineering students, leveraging Natural Language Processing (NLP) to dynamically generate subject-specific problem sets, By integrating gamification principles, such as level-based challenges, streak tracking, and rewards, the platform seeks to make learning interactive and effective [2], [4] + Methods: The methodology for developing the gamified platform for engineering education using Natural Language Processing (NLP) involves a structured approach encompassing both frontend and backend development, NLP model integration, and gamification techniques. Keywords Gamified, NLP, Adaptive Learning, Engineering Education, React Native, Student Engagement Paper id_AIML23 Title of the Article DermaTech: A Machine Learning and Deep Learning-Based Skincare Product Recommendation System Authors Anushka Supe Rutuja Mestry, Kiran Rokade, Vikrant Agaskar VidyaVardhini’s College of Engineering & Technology Abstract Background: ‘The rising incidence of skincare concerns, such as acne and hyperpigmentation,has created a demand for intelligent and personalized solutions. This study introduces DermaTech, a skincare product recommendation system that utilizes machine learning and deep learning to provide tailored product suggestions based on detected skin conditions. Methods: DermaTech employs convolutional neural networks (CNNs), specifically Inception V3 and EfficientNet, to analyze facial images and identify skin concems such as acne and dark spots. Data augmentation techniques improve model generalization, while a curated product dataset ensures relevant recommendations. Results: The system achieves high accuracy in skin condition detection using state-of-the-art CNN models. The integration of machine learning and a structured recommendation approach enhances the precision of skincare product suggestions. Conclusions: This research highlights the potential of Al-driven skincare solutions in addressing dermatological challenges. DermaTech serves as a foundation for future advancements in personalized skincare technology, improving accessibility to targeted skincare recommendations, Keywords2025 Intemational Conference on Emerging Trends in Computer Science And Engineering Skincare, Deep Learning, Convolutional Neural Networks, Personalized Recommendations, Dermatology. Paper id_AIML24 le of the Article EvalMate: Handwritten Assessment Grading Tool Using LLM Feedback Authors Yukta Saraf, Rujuta Jariwala,Jay Kansara,Nikita Raichada SVKM’s Dwarkadas J. Sanghvi College of Engineering Mumbai, India Abstract EvalMate is an innovative tool designed to simplify and enhance the grading process for handwritten assignments. By combining advanced text recognition techniques with automated evaluation, it offers a seamless solution for educators, EvalMate extracts handwritten text with precision and evaluates it against a model answer provided by the teacher, offering consistent grading that aligns with predefined criteria, Through an intuitive Streamlit-based interface, the tool integrates grading automation with a standout feature personalized feedback powered by language models This feedback not only highlights areas for improvement but also provides constructive suggestions to help students refine their understanding. For educators, EvalMate significantly reduces the time and effort required for manual grading, allowing them to focus on meaningful interactions with their students, With its unique ability to deliver detailed, actionable insights and simplify the grading workflow, EvalMate bridges the gap between technology and education. Designed for scalability, it empowers teachers and students alike by making the evaluation process efficient, insightful, and consistent. Keywords Automated Grading, Handwritten Text Recognition, Language Models, Personalized Feedback, StreamL it Interface, Grading Automation, Education Technology. Paper id_AIML28 itle of the Article Fake Social Media Bot Detection and Reporting Authors Jayesh Arjun, Pooja Chafekar, Sachin Jadhav VidyaVardhini’s College of Engineering & Technology Abstract Background'The rise of social bots on platforms like Twitter and Instagram has created significant challenges in maintaining the authenticity of online interactions. These bots,2025 Intemational Conference on Emerging Trends in Computer Science And Engineering powered by advances in natural language processing (NLP) and machine learning (ML), spread misinformation, manipulate public opinion, and disrupt organic conversations, This project aims to develop a reliable system for detecting and reporting fake social media bots using ML and NLP techniques. Provide the rationale for the study and its objectives Methods:The system collects data from social media platforms via APIs, focusing on tweets, user profiles, and interaction patterns. A custom algorithm, combined with FastText word embeddings, is used to classify posts as human-generated or bot-generated, Behavioral features such as posting frequency and interaction patterns are analyzed to enhance detection accuracy. The system is benchmarked against baseline models like Random Forest, SVM, and LSTM networks. Results'The proposed system achieved an accuracy exceeding 95% in detecting bots Real-time detection and reporting capabilities were successfully implemented through API integration. The system also provided insights into bot behavior and strategies Conclusions:The project successfully developed a scalable and accurate system for detecting and reporting social bots. The system's integration with social media platforms ensures the preservation of authentic online interactions and combats the spread of misinformation. Keywords Social bots, ML, NLP, Fake account detection. Paper id_AIML29 Title of the Article Forensic Age and Gender Estimation Using Deep Learning: A Comparative Analysis of CNN, VGG-16, ResNet, and EfficientNet-B3 Authors Yash Palkar, Nirzaree Devasthale, Dashrath [Link] Patil SVKM’s Dwarkadas J. Sanghvi College of Engineering Mumbai, India Abstract Accurately estimating age and gender is important for law enforcement, identifying people and guaranteeing social security. Specific facial feature patterns enable more precise age and gender classification due to several improvements in deep learning. This study offers a thorough comparative analysis of four meaningful deep learning models for forensic age and gender estimation. These models include Convolutional Neural Networks (CNN), VGG-16, ResNet and EfficientNet-B3, The performance of these models, which are trained and evaluated using a benchmark dataset, is determined by a high degree of accuracy, large computational efficiency, as well as large robustness. This thorough study precisely examines the meaningful strengths and weaknesses of each model, paying particularly close attention to their performance across diverse age groups, genders, lighting conditions and facial expressions. Deeper architectures, like ResNet and EfficientNet-B3, provide better classification accuracy than customary CNNs, but remain computationally feasible. This research points out the meaningful potential of deep learning in forensic applications and 102025 Intemational Conference on Emerging Trends in Computer Science And Engineering offers large guidance on selecting the optimal model for practical implementation. Future work will investigate several multimodal aspects. Keywords Forensic Age Estimation, Gender Classification, Deep Leaning, CNN, VGG-16, ResNet, EfficientNet-B3. Paper id_AIML30 itle of the Article MOODMATE: Bridging Technology and Mental Health Authors Ms, Rima Nathuram Nivale,Prof. Kadambari Metha [Link] Institute of Technology, Raigad, India Abstract Mental health issues are prevalent among young adults aged 20-35, who face stresses related to studies, work, finances, and personal relationships. We present Mental Health, a Web application to identify, analyze, and alleviate mental health problems at an early stage. The app has three key features: 1) An emotion detection module that divides the face into lines to extract and predict emotions, then recommends uplifting songs and podcasts based on the user’s current state, 2) An autonomous chatbot for users to privately discuss their feelings and receive suggested coping strategies without revealing their identity. 3) Anonymous group chat rooms where users can socialize and discuss common problems anonymously. Mental Health aims to improve mental well-being by allowing users to monitor their emotional state through facial analysis, vent feelings safely to an AI, and connect with peers undergoing similar issues while maintaining privacy. Evaluations demonstrate the emotion detection module can identify facial expressions indicative of mood with 82% accuracy. The chatbot is effectively designed to reduce negative sentiment through supportive conversations. Mental Health represents a novel mobile approach to enhancing mental health through emotionally intelligent systems and secure social features tailored to young adults’ needs. The app empowers users to privately manage mental health issues by increasing self-awareness privately, providing personalized support, and facilitating anonymous peer discussions. Paper id_AIML31 Title of the Article Anomaly Detection Using Machine Learning: Evaluating Autoencoder, ResNetS0, and PatchCore on Industrial Defects Authors Mr. Ali Karim Sayed',Dr. Ankush Pawar? 1 ARMIET College of Engg "2025 Intemational Conference on Emerging Trends in Computer Science And Engineering 2 Vishwaniketan (iMEET), University of Mumbai Abstract To sustain a quality manufacturing output while reducing operational inefficiencies, it is essential to incorporate Anomaly detection in a deep learning based approach, Anomaly detection techniques like PCA and Isolation Forrests face struggles with defect pattern collection and high miss classification rates. Recent years have shown the use of Deep learning to significantly improve defect detection and localization. This paper secks to compare and study novel approaches to Anomaly detection using Deep Learning which are (1) Autoencoder with 2 loss, (2) Feature extraction with KNN on ResNet50, (3) Multi-level feature extraction with autoencoder reconstruction on ResNetS0, (4) PatchCore anomaly detection. Performance was measured with AUC-ROC scores, visualization of reconstruction error, and heatmap based anomaly localization. Findings suggest PatchCore performs the best, achieving the highest AUC-ROC score of 0.97, while also providing superior computational efficacy and defect localization. Keywords: Anomaly Detection, Deep Learning, PatchCore, Feature Extraction, Autoencoders Paper id_AIML32 le of the Article Forensic Voice Analysis Using MFCC and SVM Enhancing Speaker Identification and Verification Authors Rahul Rathod, Rahul Pawar, Tanmay Pawar, Chandrakant Patil, Dr. Swarupa Wagh Atma Malik Institute of Technology and Research, Maharashtra, India Abstract Forensic voice analysis is a crucial tool in criminal investigations, particularly for speaker identification and verification from voice recordings. Traditional methods often fall short when dealing with noisy or low-quality audio, making automated techniques essential. This study explores the integration of Mel-Frequency Cepstral Coefficients (MFCC) and Support Vector Machines (SVM) to enhance the accuracy and reliability of forensic voice analysis The study employed a dataset of voice recordings collected from diverse sources, preprocessed using noise reduction and normalization techniques. Features were extracted using MECC, and classification was performed using SVM with hyperparameter tuning to optimize performance. The model's accuracy was evaluated using precision, recall, and Fl-score metrics, with cross-validation ensuring robustness. The MFCC-SVM approach achieved high accuracy in speaker identification and verification tasks, even under challenging conditions such as noisy or low-quality audio recordings. Including additional features, such as pitch and formants, further improved classification performance, demonstrating the method's potential for forensic applications. 122025 Intemational Conference on Emerging Trends in Computer Science And Engineering Integrating MFCC and SVM offers a reliable and effective forensic voice analysis solution, addressing the limitations of traditional techniques. Future research should focus on enhancing feature extraction and exploring deep learning techniques to improve performance further, Keywords: - Forensic Voice Analysis, MFCC, SVM, Speaker Identification, Speaker Verification, Machine Learning Paper id_AIML33 Title of the Article Deep Learning-Integrated loT Healthcare Monitoring System Authors Simran Kaur Inderdeep Singh Shembe', Satishkumar L. Varm: 1 Pillai College of Engineering 2 Dwarkadas J. Sanghvi College of Engineering, Abstract Background: Deep learning methods have greatly improved recent developments in medical imaging. By reducing the amount of manual input required for segmentation, classification, and feature extraction, these technologies have significantly improved diagnostic accuracy. Methods: In order to determine the severity of osteoarthritis (OA), this work offers an Internet of Things (loT)-powered healthcare monitoring system that makes use of deep learning models, specifically ResNet101 and DenseNet201. Sensor data gathered using a Raspberry Pi is processed and presented for examination via an interactive graphical user interface. Results: ResNet!01, DenseNet201, and the ensemble model all achieved training accuracies of 71.28%, 82.32%, and 84.2%, respectively, demonstrating the potential performance of the suggested approach. When it came to classification accuracy, the ensemble model fared better than solo models. Conclusions: Continuous vitals surveillance and early disease identification are made possible by combining deep learning with loT-based healthcare monitoring to improve patient care Keywords Deep Learning, Convolution Neural Network, Internet of Things, Healtheare, Raspberry Pi, and Ensemble Method Paper id_AIML36 Title of the Article An Exhaustive Survey of Big Data Storage Reduction Techniques 132025 Intemational Conference on Emerging Trends in Computer Science And Engineering Authors Samita Patil, Dr. Savita Sangam Shah and Anchor Kutchhi Engineering College, Chembur Mumbai Abstract The exponential growth of data in the era of big data has introduced significant challenges in storage, processing, and analysis, necessitating the development of effective data reduction methods. This survey provides a concise overview of key big data reduction techniques, systematically reviewing methodologies such as data compression, sampling, dimensionality reduction, and feature selection. By analyzing representative algorithms and frameworks, the study highlights the critical need for data reduction to manage storage demands, improve computational efficiency, and enable meaningful data analysis in large-scale systems. Key findings reveal that hybrid approaches combining multiple reduction strategies excel in maintaining data integrity while significantly reducing storage and computational overhead, with advanced techniques like deep learning-based compression and adaptive sampling, proving highly effective. The study concludes that innovative data reduction methods are essential for efficient big data management and emphasizes the importance of aligning these strategies with specific application requirements, paving the way for future advancements in scalable and optimized storage frameworks. Keywords Big data, Data Deduplication, Data compression, Dimensionality Reduction, Data Preprocessing Paper id_AIML37 Title of the Ai "Deep Learning Approaches for Text Summarization: A Comprehensive Survey" jicle Authors Nilofar Mulani, Dr Savita Sangam Shah and Anchor Kutchhi Engineering College, Chembur, Mumbai Abstract ‘Text summarization is essential for efficiently extracting key information from large textual data, The rapid growth of digital content necessitates effective summarization techniques. Traditional extractive methods select key sentences but often cause redundancy and incoherence. In contrast, abstractive methods generate new sentences while preserving meaning, improving fluency. Recent deep learning advancements include Seq2Seq models, transformers, and attention mechanisms, significantly enhancing summarization quality Hybrid models combining extractive and abstractive approaches show promising results Reinforcement learning optimizes summaries by improving coherence, informativeness, and factual accuracy. GANs enhance summary quality through adversarial training Transformer-based models like BERT and GPT enable context-aware summarization However, challenges such as factual inconsistency, redundancy, and high computational costs persist. Ensuring semantic accuracy and avoiding hallucinations remain critical concerns. 142025 Intemational Conference on Emerging Trends in Computer Science And Engineering ‘Computational efficiency is vital for large-scale applications. ROUGE, BLEU, and METEOR are common evaluation metrics for assessing model performance, Real-world applications include news summarization, academic research, and business intelligence. Future research should focus on multi-modal summarization and improved interpretability. Explainable Al (XAD) can enhance transpareney and trust in summarization models, Addressing these challenges will improve the effectiveness and reliability of summarization systems. Deep learning continues to revolutionize the field, providing innovative solutions, Integrating Al techniques will further improve efficiency and reliability. Optimizing models is crucial for balancing quality and computational efficiency. ‘The future of automatic summarization depends on achieving this balance Keywords ‘Text summarization, deep leaning, abstractive summarization, extractive summarization, reinforcement learning, RNN, seq2seq Paper id_AIMLA3 Title of the Article Advancements in Deep Federated Learning for Autonomous Vehicles: Enhancing Real-Time Decision-Making Capabilities Authors Vaibhav Dhage,Sivaram Ponnusamy School of Computer Sciences and Engineering, Sandip University, Nashik, Maharashtra, India Abstract: ‘The rapid advancements in autonomous vehicles require robust real-time decision-making systems that ensure safety, efficiency, and adaptability across diverse driving scenarios. While preserving data privacy, traditional federated learning approaches struggle to prioritize critical information, adapt to diverse environments, and maintain computational efficiency. These limitations make them less scalable and applicable to real-world autonomous driving systems, This study proposes an integrated framework that combines Hierarchical Multi-Layer Federated Attention Networks (HML-FANs), Adaptive Federated Curriculum Learning (AFCL), and Dynamic Federated Model Pruning (DFMP) to enhance federated learning for autonomous vehicles in the process of addressing these challenges. HML-FANs introduce a novel hierarchical attention mechanism operating at both edge and global levels to prioritize critical data from edge devices, improving object detection by 12% and reducing latency by 25%. AFCL uses a curriculum learning paradigm to order data progressively, ensuring incremental learning and achieving an 18% improvement in the generalization of the scenario while realizing a 15% increase in collision avoidance on high-density simulations. DFMP employs dynamic task-specific pruning, reducing the size of the global model by 40%, boosting inference speeds by 30%, and lowering energy consumption by 20%, These methods collectively optimize real-time decision-making: the model's efficiency, scalability, and adaptability improve in diverse road conditions, This paper shows that it is gaining much accuracy, training stability, and energy efficiency; this would be the most crucial step towards realizing real-time, privacy-preserved, 152025 Intemational Conference on Emerging Trends in Computer Science And Engineering and highly efficient autonomous driving systems for solving complex tasks in real-life scenarios. Keywords: Federated Learning, Hierarchical Attention, Curriculum Learning, Model Pruning, Autonomous Driving, Scenarios Paper id_AIML4S Title of the Article AyurGenixAl: An Ayurvedic Medication Advisor Authors Yadnesh Bhanushali, Kirti Rane, Nikita Yadav Datta Meghe College of Engineering, Navi Mumbai, India Abstract Traditional Ayurvedic healthcare lacks data-driven personalization, limiting its effectiveness in modern applications. This study introduces AyurGenixAl, an Al-powered Ayurvedic medication advisor that integrates machine learning, Generative Al, and a manually curated Ayurvedic database to provide personalized treatment recommendations. AyurGenixAl utilizes a Django-based web interface to process user-input symptoms using natural language processing (NLP), mapping them to a PostgreSQL-backed dataset. The system incorporates Ollama Generative AI to refine treatment plans by suggesting accurate remedies and precautions, Additionally, an interactive chatbot supports symptom analysis and treatment guidance. Evaluation results demonstrate high predictive accuracy in recommending Ayurvedic treatments. The integration of AI enhances accessibility and user engagement, improving the overall effectiveness of Ayurvedic healthcare. By combining traditional Ayurvedic knowledge with advanced AI technologies, AyurGenixAl offers an intelligent and accessible approach to holistic healthcare Keywords: Ayurveda, Machine Learning, Generative Al, Ayurvedic Medicine, PostgreSQL, Personalized Healthcare Paper id_AIML46 Title of the Article Learn$mart: Tool to Extract Key Concepts and Arguments from YouTube Lectures Authors Dr. Sujata Khedkar,Ayush Rakesh Gerra, Lintomon Chirakkara, Sania Khan,Prathamesh Joshi,Harsh Dhebe Vivekanand Education Society’s Institute of Technology, Mumbai, India 162025 Intemational Conference on Emerging Trends in Computer Science And Engineering Abstract Background: In today's fast-paced digital world, individuals face challenges in efficiently acquiring and processing vast amounts of online information. Traditional learning approaches struggle to keep up with the growing content volume, leading to cognitive overload. "Learn Smart is designed to simplify and accelerate knowledge acquisition by leveraging Al-driven content summarization and structuring. Methods: The system retrieves relevant YouTube videos based on user queries, extracts captions using Natural Language Processing (NLP) techniques, and processes them through clustering and Retrieval-Augmented Generation (RAG). The output consists of a well-structured article, a brief summary video, and an interactive flowchart. The implementation integrates Mistral AI for article generation, MiniSOM for clustering, and FAISS for vector search, ensuring optimized content extraction and summarization Results: The system significantly reduces the time required to process and comprehend digital content, It efficiently retrieves, processes, and presents summarized educational material in a multimodal format. Performance evaluations demonstrate that articles are generated in approximately 26.81 seconds, while image loading completes within 33.03 seconds. The clustering method, although yielding a low Silhouette score, effectively groups overlapping ideas from captions to provide meaningful summaries. Conclusions: "Learn Smart” revolutionizes digital education by addressing information overload and improving learning efficiency through Al-powered summarization. The system enhances accessibility and adaptability in learning by presenting content in multiple formats Future developments will focus on refining NLP models, improving summarization accuracy, and expanding content support for diverse learning needs, Keywords Al-powered learning, Natural Language Processing (NLP), Retrieval-Augmented Generation (RAG), Video summarization, Digital education, Information overload Paper id_AIML47 Title of the Article A GAN-Based Text-to-Image Framework for Automated Comic Generation and Illustration Authors Ranjana Singh, Jyoti Singh, Hitakshi Trivedi, Aditya Singh Watumull Institute of Engineering & Technology, Thane, India Abstract Comic generation is an innovative approach to storytelling that leverages the power of machine learning and deep learning techniques to automate the creation of comic panels. This paper presents Comix-Gen, a comprehensive system designed to transform user-provided text descriptions into fully realized comic illustrations featuring characters, environments, and dialogues. The system employs advanced methodologies, including Natural Language Processing (NLP) for text classification, Generative Adversarial Networks (GANs) for image generation, and Convolutional Neural Networks (CNNs) for feature extraction and style consistency. By integrating these components, Comix-Gen addresses the challenges of visual 72025 Intemational Conference on Emerging Trends in Computer Science And Engineering coherence, character representation, and dialogue authenticity, paving the way for a seamless comic creation experience, Through extensive experimentation, the system demonstrates high accuracy in character and emotion detection, producing visually compelling and narratively coherent comic panels, This research not only contributes to the field of automated comic generation but also inspires new avenues for interactive storytelling in the digital age. Keywords Comic generation, machine learning, GANs, CNNsNLP, text classification, ML, DL. Paper id_AIML49 Title of the Article Consensus of hourly visibility forecast for airports using NWP model and observations Authors Vishnukant Gupta, Rahul Barna, Shreyash Pingle, Priyanka Shinde, AbduRahim AbdulSattar Shaikh, Tushar H. Ghorpade. Ramrao Adik Institute of Technology Abstract Background: Visibility forecasting is one of the most important aspects of meteorology impacting aviation safety, urban transportation, and disaster management. Fog, pollution, or extreme weather conditions can cause adverse visibility, leading to operations being disrupted, flights delayed, road accidents, and increased risks. This paper introduces a machine learning-based visibility forecasting system that predicts hourly visibility for the next 12 to 24 hours. This approach will integrate NWP models with advanced ML techniques, such as XGBoost, to improve the accuracy and reliability of visibility predictions Methods: It involves the use of historical and real-time meteorological data, temperature, humidity, wind speed, wind direction, and air pressure, Data preprocessing includes unit conversion, handling missing values, outliers, and feature engineering. Given its efficiency for structured datasets, the XGBoost model captures complex relationships in meteorological variables, The hyperparameters are tuned with GridSearchCV for optimizing the performance of the model, and validation of the model is performed by using metrics such as Mean Absolute Error and Root Mean Squared Error. Results: The study evaluates multiple models, including LSTM-CNN, LSTM-XGBoost, and XGBoost Multivariate. The XGBoost Multivariate model achieves the lowest error rates (MAE: 424.81, RMSE: 651.09), outperforming other models in forecasting accuracy. The system effectively predicts visibility trends, with minor deviations observed during abrupt environmental changes, A correlation heatmap reveals that humidity and particulate matter concentration are the most significant factors affecting visibility. Conclusions: The developed machine learning-based system demonstrates high accuracy in visibility forecasting, with the XGBoost Multivariate model proving to be the most effective The research highlights the potential of data-driven approaches in improving decision-making, for aviation and urban transport safety. Future enhancements may involve integrating additional real-time parameters and refining feature selection techniques to further improve 182025 Intemational Conference on Emerging Trends in Computer Science And Engineering predictive performance. These findings contribute to the advancement of meteorological forecasting, offering a scalable and reliable solution for visibility-related challenges. Keywords Visibility Forecasting, Machine Learning, Numerical Weather Prediction, XGBoost, GridSearchCV, Aviation Safety, Fog Prediction. Paper id_AIMLS1 itle of the Article Generative Al Approaches for Text Generation: Techniques, Challenges, and Future Directions Authors Hrithik Singh, Vanita Mane Ramrao Adik Institute of Technology,Navi Mumbai,Maharashtra,India Abstract In this paper, an analysis of the current uses of generative Al particularly in text generation is presented under various categories including content generation, chatbots, and machine translation, Here, it also points to crucial transformer architectures, multimodal learnings, and privation preserve learnings that have advanced this field, Ethical dilemmas, computational complexity and data cration limitations are discussed to offer a clear understanding of the challenges that generative AI is bound to experience. In addition, from the future point of view, the paper defines key concepts of adaptable learning, the increase of efficiency, and the application of generative Al in low-resource and emerging domains. In this way, the study is seen as an educative platform that seeks to provide a prophetic outlook of what generative Al entails and how itis likely to shape the future of text generation to a broad audience Keywords: Generative Al, Text Generation Multimodal Learning Paper id_AIMLS3 Title of the Article Enhancing Travel Itinerary Planning with Generative Al and Real-Time Weather Data Authors Nishali zope, Prathamesh Waghmare, Jayshree Bankar, Onkar Gurav Saraswati College of Engineering. Navi Mumbai, 410210, India Abstract Planning trips demands long periods of time and complicated work which requires people to 192025 Intemational Conference on Emerging Trends in Computer Science And Engineering conduct manual investigations of destinations alongside hotel research and checks of weather conditions and activities. Traditional itinerary planning currently provides no customization capabilities which results in similar suggestions for all travelers without personal consideration, The system implements Artificial Intelligence technology for planning customized itineraries through Google Gemini API services combined with real-time weather recommendations from Visual Crossing Weather API. The web application built with Flask technology delivers better user experiences because it personalizes travel schedules to match specific preferences. The execution of this functionality requires users to answer a collection of questions about their travel style with their interests and their budget preferences. The user responses feed into a generative Al model that generates optimized travel plans which combine budget, interest and weather considerations for seamless experiences, User requests, handled through Flask while SQLAIchemy manages authentication functions and REST APIs support both Al itinerary development and weather advisory services. The testing showed Al-made travel plans create structured results which suit individual requirements without the requirement of human intervention. Travel planning efficiency gains better feedback from users through individualized recommendation features. Research findings show how AI will continue expanding within tourism applications alongside integrated real-time data to deliver user-specific travel schedules that adapt based on weather conditions. Research in the future will combine real-time traffic information, accommodation features and advanced machine learning systems for advanced personalization capabilities, Keywords Travel Planning, Artificial Intelligence, Generative AI, Google Gemini API Paper id_AIMLS4 Title of the Article Fake Shield: Detection of Deepfake Detection Using RCNN Authors Varsha Nawale , Pramod Pawar, Prasanna Jagtap, Vivek Sharma Shivansh Tiwari Saraswati College of Engineering Navi-Mumbai, India Abstract Comprehensive safety and authenticity verification in online media are more essential than ever in the ever-changing digital ecological systems. A new internet tool called FakeShield uses Region-based Convolutional Neural Networks (RCNN) to accurately identify deepfake movies. This platform integrates additional features to improve media integrity while allowing real-time footage evaluation and verification. Through APIs, it enables smooth platform integration with several video sources, making it simple to retrieve and analyze material from websites like YouTube, Vimeo, and social media platforms. The architecture of FakeShield is built for great scalability, effectively managing several fake identification 202025 Intemational Conference on Emerging Trends in Computer Science And Engineering requests at once by combining cloud-based processing and distributed computing. Among its primary attributes are automatic content scanning, real-time deepfake proof of identity, user authentication, history monitoring, an interactive report system for doubtful media, as well as multi-platform interoperability for desktop and mobile apps. FakeShield revolutionizes digital content verification by combining these features, offering a complete and user-focused security solution to counteract disinformation and deepfake manipulation Keywords Deepfake detection, RCNN, video authentication, media integrity, real-time analysis, Al security, content verification, misinformation prevention, Paper id_AIMLSS Title of the Article E-Learning Portal for GMVIT Authors Shivam Vasant Ghodekar, Yash Rupesh Thakur, Sakshi Shankar Vinherkar, Prof. Sanjay E Gavwali Abstract E-learning refers to the process of learning through electronic devices like computers, tablets, or smartphones, instead of traditional classroom settings. An e-learning platform is a website or software where students can access courses, lessons, and other learning materials online These platforms often include videos, quizzes, and interactive activities to help students learn in a fun and engaging way. The main objective of e-learning is to make education more flexible, accessible, and personalized, so that students can learn at their own pace, from anywhere, at any time, It's a great way to lear without the need to be in a physical classroom, making education easier. Keywords E-learning portal, Online platform, Educational resources, Interactive tools, Flexible learning, Accessible education, Paper id_AIMLS8 Title of the Article Transforming Dyslexia Education: A Comparative Study of the Personalized Learning Systems Authors Trishita Panja, Dr. Vanita Mane 212025 Intemational Conference on Emerging Trends in Computer Science And Engineering Ramrao Adik Institute of Technology, D Y Patil Deemed to be University, Navi Mumbai, India Abstract Personalized AI learning systems that may be designed to fit each dyslexic person uniquely will even bring a drastic change in results of education systems throughout the world. This study provides comparison analysis of generative Al models like GPT, BERT, TS ‘Transformer-XL, CTRL, XLNet in assisting dyslexic readers to assist with providing personalized learning support. ‘The adaptability of the generated content, ease with which language complexity is reduced, and improvement in comprehension are compared across these models, The study compares the models on a separate dataset of 5 years’ worth of learning materials, concentrating particularly on addressing pervasive dyslexic issues like reading fluency, comprehension, and memory retention. Our study provides useful guidance ‘on what generative AI can do for personalized education and brings new hope to how this technology could revolutionize learning opportunities available to people with dyslexia. Keywords Personalized Learning, Dyslexia, Generative Ai, GPT, BERT, TS Paper id_AIML60 Title of the Article Cardiovascular disease prediction using ECG images and using ML Authors Shraddha Mithbavkar, Eshaan Sonavane,Manas Thakare, Ashwin Yadav, Dhammabhushan Waghmare Mumbai University, India Abstract One of the most prominent tools for detecting cardiovascular problems is the electrocardiogram (ECG). The electrocardiogram (ECG or EKG) is a diagnostic tool that is used to routinely assess the electrical and muscular functions of the heart, Even though it is a comparatively simple test to perform, the interpretation of the ECG charts requires considerable amounts of training. Till recently, the majority of ECG records were kept on paper. Thus, manually examining and re-examining the ECG paper records often can be a time-consuming and daunting process. If we digitize such paper ECG records, we can perform automated diagnosis and analysis. The main goal of this project is to use machine learning to convert ECG paper records into a 1-D signal. This can be achieved by extracting the P, QRS, and T waves that exist in ECG signals to demonstrate the electrical activity of the heart using various techniques, The techniques include splitting the original ECG report into 13 Leads, extracting and converting into the signal, smoothing, converting them to binary images using threshold and scaling. This final model will aid in the diagnosing of cardiac 222025 Intemational Conference on Emerging Trends in Computer Science And Engineering diseases, to detect whether a patient has/had Myocardial Infarction, Abnormal Heartbeat, or the patient is hale and healthy by inferring the ECG reports. Keywords: NFC, Android application, hospital admission, vpatient registration, healthcare technology Keywords ECG (Electrocardiogram), Cardiovascular, Myocardial Infarction, KNN, SVM, Random Forest,Deep Learning, Ensemble Learning, Dimensionality Reduction, Streamlit Paper id_AIML62 Title of the Article From Rule-Based Systems to Transformers: The Evolution of NLP Techniques, Applications, Limitations, and Challenges Across Industries Authors Darshan Sura, DrAmarsinh Vidhate Ramrao Adik Institute of Technology, D Y Patil Deemed to be University, Navi Mumbai, India Abstract NLP has seen its application extended far and wide from sentiment analysis to machine translation, due to the advancement of language model architectures. With the rapid changes in the current digital era, artificial intelligence is increasingly establishing the NLP in recent times by combining the latest machine learning, and deep learning methodologies. In this paper we provide insights of various language model architectures like N-gram and RNN types, and more advanced models, such as transformer based architectures, In the various tasks, such as text classification, sentiment analysis, language generation and more, the study highlights the advantages and limitations of each model. Key contributions of this review include: (a) a comparative study between some of the widely known models, such as N-grams, LSTMs, and Transformer models like BERT, GPT and more, to determine whether they are fit for the various tasks (b) a review of the appropriate language model architectures that can be utilized in NLP. Finally, this review highlights the role of AI driven models in bringing about the transformative change in understanding and generation of human language Keywords Language Models, Natural Language Processing(NLP), N-grams, Long Short-term Memory(LSTM), Recurrent Neural Networks(RNN), Transformers 232025 Intemational Conference on Emerging Trends in Computer Science And Engineering Paper id_AIML65 Title of the Article Advancements In Deep Learning For Alzheimer’s Disease Detection Authors Tejaswini Sanjay Todkar, Harsh Jambukumar Upadhye, Parshwa Motichand Bhandari, Prof. Babita S Gawate, Department Of Computer Engineering, A P Shah Institute Of Technology, Thane Abstract Alzheimer’s disease (AD) is a form of dementia, significantly impacting global healthcare systems. Early detection and managing the disease's progression. Recent advancements in artificial intelligence (Al), particularly deep learning (DL), have shown promising results in identifying intricate patterns in neuroimaging data, such as Magnetic Resonance Imaging (MRI) scans, enabling efficient and non-invasive AD diagnostics. This review paper provides a comprehensive analysis of nine state-of-the-art research studies utilizing DL techniques for AD prediction and classification. The studies employ diverse methodologies, including Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Transfer Learning, and hybrid architectures like VGG16, InceptionV3, and Autoencoder-based models. Key focus areas include multi-class classification, time-to-event prediction frameworks, innovative preprocessing methods, and strategies to address challenges such as class imbalance and overfitting. The findings highlight the strengths and limitations of these approaches, emphasizing the importance of advanced feature extraction, data augmentation, and performance optimization techniques in enhancing model accuracy and clinical relevance. Additionally, the review explores the potential of these models to bridge the gap between research and clinical application, offering insights into their scalability and interpretability, By consolidating current research trends and identifying future directions, this paper aims to advance the development of Al-driven, scalable, and clinically interpretable solutions for Alzheimer’s diagnosis and treatment, contributing to the broader fight against this debilitating disease Keywords Alzheimer's Disease (AD), Deep Learning, Brain MRI Scans, Neuroimaging, Convolutional Neural Networks (CNNs), Transfer Learning, Early Detection, Medical Image Analysis ntervention are essential for improving patient outcomes and Paper id_AIML67 Title of the Article Image Captioning for blind People using Deep learning CNN-LSTM Technique 242025 Intemational Conference on Emerging Trends in Computer Science And Engineering Authors Shaumya kumari', Satishkumar L Verma? 1 Pillai College of Engineering, Navi Mumbai, India 2 SVKM's Dwarkadas J. Sanghvi College of Engineering, Mumbai, India Abstract Background: Image captioning, the automatic generation of textual descriptions for images, has garnered considerable attention in recent years due to its broad range of applications in domains such as accessibility, content understanding, and human-computer interaction. The advent of deep learning techniques has revolutionized this task, offering substantial improvements in both accuracy and efficiency Methods: The proposed work leverages a Convolutional Neural Network (CNN) to train on images, while a Long Short-Term Memory (LSTM) network is used to train on sentences, ‘These models are integrated using the Encoder-Decoder technique. When a test image is provided, the encoder pull out all the features from the image using CNN, and the decoder ‘generates the corresponding text based on the extracted image features, ultimately producing, an image caption. To train the CNN, we employ the Flickr image dataset and the VizWiz-VQA (Visual Question Answering) dataset. The caption sentences from both datasets are used to train the LSTM network Results: The fusion of CNN and LSTM in the Encoder-Decoder architecture enable our model to caption images with an impressive accuracy of 99%, Conclusions: This paper we begin by exploring the challenges inherent in image captioning, which include understanding image content, generating coherent and semantically meaningful descriptions, and handling diverse visual contexts with the help of deep learning techniques Keywords Image Captioning, Deep Learning, CNN-LSTM technology, Encoder, Decoder, Flickr8k and Vis-Wis datasets. Paper id_AIML68 Title of the Article Real-Time Dance Performance Evaluation and Feedback System Using Motion Tracking and Pose Estimation Authors Shraddha Mithbavkar, Achyut Auti, Jayant Chaudhari, Debashsish Halder Mumbai University, India Abstract Traditional dance learning methods often lack real-time feedback, making it challenging for 252025 Intemational Conference on Emerging Trends in Computer Science And Engineering learners to improve their skills effectively. To address this limitation, an interactive dance learning system is proposed, leveraging advanced technology to enhance the learning experience. The system utilizes pose estimation technology, specifically Mediapipe, to analyze user movements in real-time. It detects body Jandmarks, evaluates pose accuracy, and ‘generates performance scores. Gamification elements, such as challenges and competitions, are incorporated to increase user engagement and motivation. The solution is built using a scalable full-stack architecture to ensure seamless performance. The proposed system effectively provides instant feedback, allowing users to refine their dance movements Preliminary tests indicate its potential to enhance dance education by offering a structured and interactive learning experience. This interactive dance learning system demonstrates significant potential in transforming traditional dance education, Additionally, its applications extend to other domains, including fitness, sports training, and rehabilitation, where real-time motion analysis and feedback are valuable Keywords Pose estimation, real-time feedback, motion tracking, dance learning, gamification, deep learning, computer vision, interactive learning, human movement analysis, and performance evaluation Paper id_AIML69 Title of the Article AgroConnect: A Data-Driven Crop Recommendation System Using Machine Learning For Smart Farming Authors Aditi Honagekar, Arya Khot, Moksh Jain, Vanita Mane Ramrao Adik Institute of Technology,D Y Patil Deemed to Be University, Nerul, 400706, Maharashtra, India. Abstract Background: The Smart Crop Recommendation System aims to optimize crop selection by utilizing digital technologies in the agricultural sector. Traditional methods, which often depend on intuition or limited knowledge, can lead to inef- ficiencies and financial losses. To enhance agricultural productivity, AgroConnect uses machine learning techniques, such as Random Forest and Decision Trees, to deliver data-driven crop recommendations Methods: The system analyzes key agricultural parameters, including nitrogen, phosphorus, and potassium levels in the soil, along with temperature, humidity, rainfall, and pH levels Through a user-friendly interface, farmers can enter their soil and environmental data to receive personalized crop recommendations, Additionally, AgroConnect features multi-site technology to support multiple languages, increasing accessibility. A real-time weather 262025 Intemational Conference on Emerging Trends in Computer Science And Engineering forecasting component further refines recommendations by considering short-term climate variations, Results: The integration of machine leamming models enhances prediction accuracy, enabling farmers to make more informed decisions based on their environmental conditions, ‘The five-day weather forecast feature assists farmers in planning their cultivation activities and mitigating risks associated with adverse weather. Moreover, the multilingual support broadens accessibility, making the system advantageous for a diverse range of farmers. Conclusions: AgroConnect bridges the gap between traditional agricultural practices and modern technological advancements, promoting efficient resource utilization and higher crop yields, By aligning crop selection with environmental factors, it reduces risks, minimizes losses, and supports agricultural sustainability. The combination of machine learning and near real-time weather intelligence provides farmers with actionable insights, contributing to a more resilient and productive agricultural ecosystem Keywords: Crop Selection, Machine Learning, Random Forest, Decision ‘Trees, Soil Parameters, Climate Variations, Agricultural Technology, Weather Forecasting, Sustainability Paper id_AIML70 Title of the Article Automatic Question Generation using Maximum Marginal Relevance Algorithm and TS transformer Authors Prof, Pranita Jadhav, Dr. Shivajirao M. Jadhav Dr. Babasaheb Ambedkar Technological University, Lonere, Raigad, India Abstract An Automatic Question Generation (AQG) study addresses the growing demand for advanced educational systems and active learning. Automated MCQ generation creates ‘gap-fill questions (GFQs) with one correct and three incorrect options. Using the Python library seqval, sequence labeling was evaluated for question generation, particularly for chunking tasks like named-entity recognition and POS tagging. Roberta tokenizers combined with the Roberta-base model and the Conll2003 dataset identify meaningful phrases. AQG simplifies aptitude quiz creation for instructors, supporting online evaluations. The Maximum Marginal Relevance method ensures effective distractor selection, achieving a 98.90% accuracy in informative sentence selection. The study highlights deep learning methods, focusing on the TS transformer for generating context-based questions and answers. Keywords Gap-fill question (GFQ), Automatic Question Generation (AQG), Automatic Multiple Choices Question Gen- erator (AMCQG) and Named Entity Recognition (NER). 272025 Intemational Conference on Emerging Trends in Computer Science And Engineering Paper id_AIML73 Title of the Article ‘The Role of Deep Learning in Credit Card Fraud Detection: A State-of-the-Art Review Authors Pranita Sangit, DrSavita Sangam Shah and Anchor Kutchhi Engineering College, Chembur, Mumbai Abstract Credit card fraud detection is a critical challenge in the financial sector. With the increasing volume of transactions and the sophistication of fraudulent activities, traditional rule-based and machine learning approaches often struggle to provide accurate and timely fraud detection, Deep learning techniques have emerged as a powerful tool to enhance fraud detection by capturing complex transaction patterns and improving detection accuracy. This review explores various deep learning methods applied to credit card fraud detection, comparing their effectiveness, challenges, and future directions. Keywords Credit card frauds, deep learning, Machine learning Paper id_AIML74 le of the Article ANKUR: A Crop Disease Prediction System using Machine Learning Authors Abhishek Gadankush, Achal Patil, Siddhant Gawai, Vikrant A, Agaskar, Dr. ‘Umesh Kulkarni Vidyalankar Institute of Technology, Mumbai, India Abstract This study presents ANKUR: A Crop Disease Prediction System, a deep learning-based solution for the automated detection and recognition of crop diseases using image data. ANKUR leverages Convolutional Neural Networks (CNNs) to analyze crop images and accurately classify diseases in real-time, enabling farmers to take timely action, The model is trained on a diverse dataset comprising both healthy and diseased crops, ensuring scalability across multiple crop types and disease conditions. The system is integrated into a user-friendly web application, providing an efficient and practical tool for crop health management. By facilitating early disease detection, ANKUR aims to reduce agricultural losses, improve productivity, and contribute to sustainable farming practices. This automated approach offers a significant improvement over traditional manual inspection methods, 282025 Intemational Conference on Emerging Trends in Computer Science And Engineering which are often time-consuming, labor-intensive, and prone to human error. Keywords | Crop Disease Detection, Deep Leaming, Convolutional Neural Networks (CNNs), Image Classification, Agricultural Technology, Plant Disease Prediction ] Paper id_AIML75, Title of the Article Green Thumb Network: Urban gardening, home plant care and Community Platform Authors Prof. Uma Kishor Ade, Aman Yadav,Pritam Yadav, Amit Yadav, Roshan Shelar, Durvesh Pawar Watumull Institute of Engineering & Technology Abstract Background: Urban gardening and home plant care have become increasingly popular. However, many individuals lack the knowledge required to grow plants successfully. Green Thumb Network (GTN) is designed to assist users in identifying, categorizing, and learning, about plants, flowers, and vegetables. Methods: The platform provides an interactive home gardening tool where users can select plants and receive detailed requirements for temperature, soil type, sunlight, and watering, Additionally, GTN integrates an Al-powered plant disease detection system that allows users to upload images of affected leaves. The AI then diagnoses issues and suggests solutions, The platform also features a community page for users to share images and experiences, functioning similarly to Twitter Results: The Green Thumb Network provides an intuitive plant care guide and Al-based plant disease detection. The community page fosters knowledge exchange among plant enthusiasts Initial testing with users demonstrated high accuracy in plant disease detection and increased engagement in gardening activities, Conclusions: GTN is an effective solution for plant identification, disease diagnosis, and ‘community interaction, making plant care more accessible and engaging Keywords Plant Identification, AI Disease Detection, Home Gardening, Plant Community, Smart Agriculture, Green Thumb Network. Paper id_AIML76 Title of the Article Optimized Fault Diagnosis in Three-Phase Voltage Source Inverters Using ANN and Machine Learning Techniques. 292025 Intemational Conference on Emerging Trends in Computer Science And Engineering Authors Vaishali Sonawane', Seema Idhate’, Arya Deshpande', Jyoti Gangane* 1 SKNCOE, Vadgaon, Pune 2 School of Computer Engineering and Technology, MITWPU [Link] 3 VIMEET, Khalapur Abstract The reliability of three-phase Voltage Source Inverters (VSI) is critical for industrial and renewable energy applications, where open circuit faults in switches can lead to performance degradation, energy loss, and equipment damage. This study presents a machine Jearning-based fault diagnosis framework integrating Artificial Neural Networks (ANN) and ensemble classifiers to enhance fault detection accuracy under varying load conditions. Current signal waveforms were acquired using a Tektronix Digital Storage Oscilloscope and pre-processed using Discrete Wavelet ‘Transform (DWT) to extract meaningful time-frequency features, Various machine learning algorithms, including Support Vector Machines (SVM), k-nearest Neighbours (KNN), Decision Trees (DT), Random Forest (RF), and Adaboost, were compared for VSI fault classification, Experimental results indicate that Random Forest (RF) achieved the highest accuracy of 99.97%, followed by Decision Trees (99.49%), k-nearest Neighbours with k~3 (99.43%), and Artificial Neural Networks (99.6%) The ANN model with an 18-10-1 architecture demonstrated strong classification performance, outperforming Adaboost, which reached a peak accuracy of 96.95% at 500 estimators, Support Vector Machines (SVM) achieved an accuracy of 90.87%, making them less effective compared to ensemble-based classifiers, The study further validates that wavelet-based feature extraction combined with ANN reduces misclassification rates and improves real-time detection efficiency. Additionally, the proposed system reduced fault diagnosis time by 22.5% compared to conventional methods, making real-time implementation feasible, The findings confirm the effectiveness of the proposed ANN-based fault detection system in accurately classifying single and double open switch faults in VSI Future work will focus on integrating the system with IoT-based predictive maintenance and cloud-based fault detection for real-world applications, Keywords Artificial Neural Networks (ANN), Decision Trees(DT), Discrete Wavelet Transform (DWT), Fault diagnosis, Machine Learning, Paper id_AIML78 le of the Article Al Embedded QR Code Generator and Decoder Authors Mohammad Aasim Shaikh, Umer Azmi, Piyush Dubey, Niraj Rampal Shree L. R. Tiwari College of Engineering, Mumbai, India Abstract This research paper presents the design, implementation, and integration of an enhanced QR code system that not only generates and decodes QR codes from diverse data inputs but also 302025 Intemational Conference on Emerging Trends in Computer Science And Engineering embeds these QR codes within Al-generated artistic images. The project has two major features: first, a robust QR code application that handles multiple data types—including text, URLs, emails, WiFi credentials, and contact information—and provides accurate decoding from images; and second, an advanced Al module that leverages deep learning models (Stable Diffusion with ControINet) to generate artistic images that incorporate scannable QR codes. This integrated approach enhances both the functional and aesthetic aspects of QR code usage, broadening the application range from secure data sharing to creative marketing, solutions. Keywords QR Code, Stable Diffusion, ControiNet, Al-Generated Images, Deep Learning, Computer Vision, Natural Language Processing, Paper id_AIML81 Title of the Article Sentiment analysis using pre-trained NLTK models BERT, TextBlob, Flair, VADER & RoBERTa with personalized results. Authors Gaurav Ojha, Sahil Patil, Sonukumar Jha, Tanush Shetty, and Tushar Ghorpade DY Patil Deemed to be University), Navi Mumbai, 400706, Maharashtra, India Abstract Background: Sentiment analysis plays a crucial role in natural language processing (NLP), aiding in opinion mining, market research, and decision-making. The increasing availability of textual data from social media and online platforms has led to a growing demand for efficient sentiment classification techniques Methods: This study explores sentiment analysis using pre-trained models such as BERT, TextBlob, Flair, VADER, and RoBERTa, A comparative analysis is conducted to assess model accuracy, precision, recall, and Fl-score across multiple datasets, Additionally, ensemble learning techniques are employed to enhance performance by combining individual model predictions. ‘The study also incorporates personalized sentiment analysis, adjusting ‘outputs based on user preferences and context. Results: The results demonstrate that deep learning-based models outperform traditional lexicon-based approaches, with RoBERTa and BERT achieving the highest accuracy. The ensemble approach further improves classification performance, offering a more reliable sentiment analysis framework Conclusion: The findings contribute to the ongoing advancement of sentiment analysis methodologies, providing insights into model selection for various applications. Future 312025 Intemational Conference on Emerging Trends in Computer Science And Engineering research can focus on further optimizing ensemble models and exploring domain-specific adaptations to improve classification accuracy Keywords: Sentiment analysis, BERT, TextBlob, Flair, VADER, RoBERTa, Ensemble leaming, Deep learning, Data augmentation Sentiment analysis, BERT, TextBlob, Flair, VADER, RoBERTa, Ensemble learning, Deep learning, Data augmentation Paper id_AIML82 le of the Article User Recommendation System based on Project Metadata using Ensemble Learning Authors Prabhakar E, Sebe S, Ajaykumar N, Gowtham R, Surya Nandha College of Technology, Tamil Nadu, India . Kiruthikadevi K Abstract For upcoming days, gather required users to be a challenging one, Each user has their own skills, preferences to contribute and work for multiple projects. Many projects have been facing this issue due to hire required candidates based on their project requirement and it causes delay based on project development, execution. The proposed system will help project managers, project team leaders to overcome these difficulties to enhance their screening operations to recommend users who they prefer to work on their project. The proposed system consists of user dataset which includes user metadata based on skills, experiences, projects, and preferences etc. This recommendation leverages multiple approaches to enhance user recommendations. The user profile recommendation can be done by ensemble learning to combine multiple approaches, so here system used content based, collaborative based, popularity based and pre-requisites-based approach to provide diverse recommendations This solution will enhance user recommendation based on projects and user matching platform, The proposed user recommendation system helps to provide users based on metadata matching and highlight users based on score with a user-friendly approach. To enable real time user data matching, this recommendation model integrated with a web application using emerging technologies, Keywords Ensemble Learning, Content based recommendation, Pre-requisites based recommendation, Popularity based recommendation, metadata matching Paper id_AIML83 le of the Article Ocular Biomarkers For Diseases Detection 322025 Intemational Conference on Emerging Trends in Computer Science And Engineering Authors Vedika Avhad , Bilal Shaikh , Aniket Patil , Tanvi Mayekar, Harshal Chimane VPPCOE&VA, University of Mumbai Abstract The development of non-invasive and early illness detection techniques is crucial in mode healthcare, This initiative has provided us with a unique technique to use biomarkers, This method aids in the knowledge of the human eye by utilizing iridology concepts and powerful machine learning algorithms, with an emphasis on certain biomarkers seen in the iris, sclera and retina, To categorize and forecast probable illness states using ocular pictures, the software uses a convolutional neural network (CNN). The system is trained on a large dataset of pre-processed eye pictures and uses augmentation approaches to increase model generalizability. Furthermore, it is important since it provides real-time data and analyzes submitted photos, making it available to medical experts and academics. The project's goal is to enhance diagnostic accuracy and provide a cost-effective, scalable tool for disease screening, Hence, the results from this system could significantly contribute to early disease detection, leading to timely medical interventions and improved patient outcomes. Keywords - ocular biomarkers, iridology, deep learning and analysis, EfficientNetBO Paper id_AIML84 Title of the Article DECIPHERING LANGUAGE USING NLP TECHNIQUES Authors Prof Gayatri Dharap, Atharva Bhave, Shriram Lokhande, Bhakti Mhatre, Almeera Patel Saraswati College of Engineering, Navi Mumbai, Maharashtra, India Abstract Human-computer interaction has in recent years undergone a rapid evolution, and such technologies as Natural Language Processing and hand gesture recognition have been brought firmly to the forefront to enable new levels of accessibility and experience, Natural Language Processing lets programs understand, interpret, and respond to human language, while hand gesture recognition aids in seamless nonverbal interaction and communication. In spite of years of research, the integration of these technologies in one unified package remains commercially underexploited space full of real-world potential. This project aims to create a unified system incorporating NLP functionalities such as automatic text summarization, stemming, lemmatization, sentiment analysis, text similarity, part-of-speech tagging, named entity recognition, next-word generation, and chatbot implementation into a hand gesture recognition module for intuitive interaction. The NLP modules will harness the 332025 Intemational Conference on Emerging Trends in Computer Science And Engineering power of state-of-the-art techniques like deep learning and transformer-based models while the gesture recognition module will use convolutional neural networks to classify and interpret dynamic hand gestures. In addition to these core functionalities, the project includes auxiliary features such as an interactive blogging platform for knowledge sharing and a custom code editor for live experimentation. Together, these components constitute a versatile platform designed to foster communication, enhance productivity, and engage users. Tying together textual and gestural modalities, this framework faces the core human-computer interaction challenges that releases space for future innovations in education, assistive technology, and smart interfaces. Bringing together NLP and gesture recognition enhances the speed of diagnosis with supplementary user convenience and lays the basis for further development of multimodal interaction systems. Keywords Human-computer interaction, Natural Language Processing (NLP), Hand gesture recognition, Text summarization, Sentiment analysis, Named entity recognition (NER), Next- word generation, Chatbot, Blogging platform, Code editor, Education Paper id_AIML86 tle of the Article PERSONALIZED HEALTHCARE RECOMMENDATION Authors Varad Pawar, Jayendra Nar, Sahil Nar, Sahil Bunkar, Ashwini Mulik Vasantdada Patil Pratishthan’s College of Engineering & Visual Arts Mumbai, India Abstract ‘A Content Personalization Engine is an adaptive system that curates digital content tailored to individual users by analyzing their preferences, behaviors, and contextual information Utilizing machine learning and data analytics, these engines assess user interactions and classify audiences to forecast and deliver relevant content in real time, The primary aim is to boost user engagement and enhance the overall experience by offering personalized recommendations, articles, or produets on platforms such as websites, mobile applications, or email campaigns. Widely used in sectors like e-commerce, media, and entertainment, these engines help achieve greater engagement, improved conversion rates, and increased customer satisfaction through the strategic delivery of customized content. Keywords Content Personalized Engine, Adaptive System,User Preferences, Behavioral Analysis, Machine Learning, Data Analytics Real-Time Content Delivery.2025 Intemational Conference on Emerging Trends in Computer Science And Engineering Paper id_AIML87 Title of the Article Word Based Indian Sign Language Translator Using Seq-2-Seq Model Authors Ansari Fatima Anees, Vivek Chouhan, Nishikant Raut, Rehan Sayyed, Rohit Deshmukh, Nishikant Raut M, H, Saboo Siddik College of Engineering Abstract Background: Communication barriers significantly impact the Deaf and hard-of-hearing community in India due to the unique structure of Indian Sign Language (ISL). This project aims to address this by developing an ISL translator to facilitate real-time communication Methods: The system employs OpenCV and MediaPipe for landmark recognition of ISL gestures. A sequence-to-sequence (seq2seq) model, built using TensorFlow and trained on a comprehensive ISL dataset, is used to predict the corresponding text or speech Results: The Al-driven system accurately translates ISL gestures into text or speech in real-time, Conclusions: This ISL translator demonstrates the potential to bridge communication gaps, promote social inclusion, and improve accessibility for ISL users by providing accurate and timely translations. Keywords ISL, translation, seq2seq, modelling, RNN, GRU Introduction Paper id_A\ 1L88 Title of the Article Machine Learning Approaches For Predicting Safe Landings Of The SpaceX Falcon 9 First Stage. Authors Ashutosh Lalji Mishra,Pradnya Bhagwat Bhakte, Aaditya jitendra Salvi [Link] Institute of Technology Abstract This project aims to predict whether the SpaceX Falcon 9 rocket landing will be successful or not using data science and machine learning techniques. The data comes from the SpaceX API, We use Pandas and NumPy for handling the data, create visualizations to explore it, and apply machine learning models to make predictions. The main steps include gathering data, cleaning and analyzing it, identifying important features for prediction, and building a 352025 Intemational Conference on Emerging Trends in Computer Science And Engineering machine learning model to forecast landing outcomes. The project helps identify key factors that influence the success or failure of these landings Paper id_AIML91 Title of the Article Enhanced Structural Health Monitoring Through LSTM-Enhanced Gradient Boosting Regressor. Authors Aryan Kesarkar, Chrisil Dabre, Prof, Ruhina Karani, Raghav Agarwal, Yash Chavan DJ Sanghvi College Of Engineering Abstract Background: Structural deterioration in critical infrastructure, such as bridges and buildings, occurs due to factors like material fatigue and environmental stress. Accurate failure prediction is essential for safety and maintenance planning. Traditional models struggle to capture complex interactions in structural health data, necessitating a more robust approach. Methods: This study proposes a hybrid model, the LSTM-Enhanced Gradient Boosting Regressor (LE-GBR), which integrates Gradient Boosting Regressor (GBR) for static pattern recognition and Long Short-Term Memory (LSTM) networks for capturing temporal dependencies. The model was trained and evaluated on a collected structural health dataset, Results: Experimental results demonstrate that LE-GBR outperforms existing models, achieving higher accuracy in predicting structural health outcomes. The model effectively captures both static and dynamic patterns, enhancing prediction reliability. Conclusions: LE-GBR presents a promising tool for improving structural failure predictions, even without explicit time-series data. Its application can aid in proactive maintenance and risk mitigation, ensuring safer infrastructure management, Keywords Structural Health Monitoring, Failure Prediction, Gradient Boosting Regressor, LSTM Network, Hybrid Model, Infrastructure Safety, Predictive Modeling, Structural Integrity. Paper id_AIML92 le of the Article ‘The Real-Time Emotion and Mental State Monitoring System for Train Drivers Authors Parth Patil, Vaibhav Nemani, Vaibhav Nemani, Padmashri Panda, Vrishali Patil 362025 Intemational Conference on Emerging Trends in Computer Science And Engineering Ramrao Adik Institute of Technology, Navi Mumbai, Maharashtra , India Abstract The system works by picking up real-time data on pilots’ physical and mental states—things like heart rate, skin conductivity, and breathing patterns, Why? Because these details can help us in if they're feeling stressed or worn out. Plus, it observes their behavior too, keeping tabs ‘on facial expressions, eye movements, and even speech patterns using advanced video analysis. All this info gets passed through machine learning to spot any patterns that might signal emotional strain or mental overload. If the system senses a potential problem, it sends out an automatic alert to the pilot or dispatchers, allowing them to act fast. This could mean, suggesting a quick break, some relaxation techniques, or even handing tasks over to another operator. By delivering these real-time insights and helpful suggestions, the system not only gives pilots a breather but also improves the safety and efficiency of train operations. Under the hood, we're using all sorts of tech to make it happen: Python for coding, OpenCV for processing video, and Tensor Flow/Keras for the deep learning models. The setup’s pretty flexible, so it ean be adjusted to monitor different emotions and store data for future study ‘That's a big plus—not just for safety but for long-term research on job-related stress in high stakes fields. In the end, this whole setup aims to build a safer railway network by making sure that loco pilots are in good shape mentally and emotionally while they’re working, And down the line? We're hoping to add more emotions to the list, sharpen the system’s accuracy, and see what other benefits it might bring to the table. Paper id_AIML96 le of the Article Deep Learning-Based Approach for Ambulance Detection & Traffic Management Authors Ganesh Mohane, Anuj Mishra, Aaditya Prajapati, Yashwardhan Pandey, Ms. Priya Deshpande, Dr. Nandini C. Nag Lokmanya Tilak College of Engineering, Navi Mumbai, Maharashtra, India Abstract Ambulances require minimal response times to ef- fectively manage critical situations and prevent loss of life. How- ever, they frequently encounter delays due to traffic congestion, primarily caused by conventional traffic signal systems that lack emergency vehicle prioritization, While modern smart traffic lights have been introduced to optimize traffic flow, they do not incorporate effective mechanisms to detect and prioritize ambulances. To address this limitation, this study proposes a deep learning- based traffic management system that employs the YOLO (You Only Look Once) model for real-time detection of ambulances 37