Category Archives: Uncategorized

Authors: Om Sawant, Gunjan Shahade, Atul Sanap, Shailesh Pawar, Madhuri Shinde

Abstract: The widespread adoption of Internet of Things (IoT) systems has resulted in a large volume of time-sensitive data that requires fast and efficient processing. Although cloud platforms provide extensive computational capabilities, the physical separation between data-producing devices and remote cloud infrastructures frequently introduces noticeable delays, jitter, and bandwidth inefficiencies. Fog computing addresses these shortcomings by relocating processing tasks toward the network’s periphery; however, the decentralized and heterogeneous composition of fog resources complicates the design of effective scheduling strategies. Recent progress in Artificial Intelligence (AI), especially in the field of Deep Reinforcement Learning (DRL), have enabled adaptive and context-aware scheduling solutions capable of responding to dynamic changes in fog–cloud systems. This study presents an in-depth examination of AI-oriented scheduling mechanisms for fog computing, with emphasis on system design principles, algorithmic trends, and comparative performance outcomes. Conventional scheduling heuristics, machine-learning-based methods, and contemporary DRL approaches—including multi-agent and multi-objective frameworks—are critically analyzed. The review also identifies persistent challenges related to scalability, mobility, resource constraints, and security-aware decision-making. Overall, the findings demonstrate that AI-driven scheduling enhances responsiveness, load distribution, and resource utilization in emerging fog-supported IoT environments.

DOI: http://doi.org/10.5281/zenodo.20769734

Published by: vikaspatanker

Authors: Kiran P. Borase, Sachin K. Dahake

Abstract: This research focuses on the design and optimization of a motorcycle swing arm using a bio-inspired honeycomb structure aimed at achieving significant weight reduction while enhancing stiffness and durability. A conventional swing arm was modelled using Solid Works and compared with an optimized honeycomb-reinforced structure through Finite Element Analysis (FEA) in ANSYS. The inclusion of honeycomb geometry demonstrates improved structural efficiency, reduced stress concentration, lower deformation, and an expected weight reduction of 15–20%. The study establishes the feasibility of integrating nature-inspired geometrical patterns into mechanical components to achieve superior performance in lightweight engineering applications.

DOI: http://doi.org/10.5281/zenodo.20766895

Published by: vikaspatanker

Authors: Ashutosh Kale, Swapnil Yeole, Onkar Sonawane, Jayesh Wani, Vedant Rajput

Abstract: This analysis paper examines ResiPlan AI, an intelligent web-based system designed to automate residential floor plan generation using artificial intelligence and machine learning techniques. The system addresses significant barriers in traditional architectural design—such as high cost, complexity, and reliance on expert knowledge—by enabling non-expert users to generate optimized 2D and 3D layouts through simple inputs like plot size, room count, and architectural style. By integrating Stable Diffusion 1.5 with ControlNet, ResiPlan AI ensures structural adherence while maintaining creative flexibility. This paper critically evaluates the system’s architecture, technical approach, limitations, and future potential, positioning it within the broader context of generative AI in architectural design.

DOI: http://doi.org/10.5281/zenodo.20766823

Published by: vikaspatanker

Authors: Kishor Ugale, Tushar Pandhi

Abstract: Traditional Voting plays a very important role in modern republic systems. Electronic voting (e-Voting) refers to any method of casting or recording votes through electronic technologies. Voting machines consist of the whole combination of mechanical, electromechanical, or electronic components-along with the necessary software, firmware, and documentation-used for programming, controlling, and supporting the voting process. The e-Voting system discussed here uses biological validation, notably fingerprint identification, to verify voter identity. In this method, fingerprint matching is employed to confirm the user’s identity. This proposed work bears by differentiating sample fingerprint patterns to show whether the fingerprints real from the match individual. The primary objective of this system is to simplify and improve the regulation of the voting mechanism. The proposed solution is designed to encourage full participation by enabling every eligible voter to take part in elections. This is achieved through an Android application that permits human being to cast their balloting digitally. Implementing online voting across both Android and web-based platforms increases the reliability and effectiveness of the election process. The system aims to offer a convenient, user-friendly, and secure method for recording and counting votes. Online voting can reduce operational costs, boost voter turnout, and facilitate better communication in the middle of voters and candidates. The core target of the implemented system is to provide a voting mechanism that authorizes singles to submit secure and confidential ballots over a network, addressing the restrictions of traditional voting methods, which are often time-consuming and vulnerable to security issues.

DOI: http://doi.org/10.5281/zenodo.20766695

Published by: vikaspatanker

Authors: Khushboo Lokhande, Samruddhi Mahajan, Sayali Pawar, Janhavi Wankhede, Vijay More

Abstract: Vaani2Mudra is an online assistive communication platform that converts spoken or written language into gestures representing Indian Sign Language (ISL). The platform utilizes a compact speech recognition model to process voice input and employs natural language processing methods to restructure spoken content into a format compatible with ISL. Through a rule-based linguistic framework, the system eliminates redundant grammatical elements and standardizes text for gesture mapping. For multilingual functionality, Marathi language input is first converted to English before further processing. The output is presented through a series of pre-established ISL gesture visuals shown on a web-based interface. This system prioritizes ease of use, instantaneous processing, and user accessibility, positioning it as an effective tool for learning environments and assistive communication applications.

DOI: http://doi.org/10.5281/zenodo.20766405

Published by: vikaspatanker

Authors: Nupur Jadhav, Atharva Bhusnale, Pritesh Gupta, Sakshi Jadhav, Vaishali Hiray

Abstract: Magnetic Resonance Imaging (MRI) is very impor-tant in the detection of neurological defects because it possesses high resolution that enables good visualization of soft-tissue structure. However, diagnostic clarity is often hindered by low-resolution scans due to the short time of acquisition, motion artifacts and hardware constraints. Recent advances in deep learning, such as Enhanced Super-Resolution Generative Ad-versarial Networks (Real-ESRGAN), have demonstrated strong capabilities of perceptual-driven image enhancement.This paper discusses Real-ESRGAN-based MRI super-resolution strategies, their architectural advantages and clinical potential benefits, in preserving fine anatomical and pathological details much better than CNN-based and conventional interpolation methods. We also present a conceptual AI-enabled deployment framework, where Real-ESRGAN is handled by a clinician support chatbot for application in web-based interaction, tele-radiology accessibility and diagnostic help. Clinical validation including metrics such as PSNR, SSIM,LPIPS and sFRC is investigated. The study emphasizes the need for interpretable, regulation-ready models to bridge AI-driven MRI enhancement with real-world diagnostic workflows.

DOI: http://doi.org/10.5281/zenodo.20766158

Published by: vikaspatanker

Authors: Samruddhi Maheshkumar Aher, Harshali Rajendra Bagul, Diksha Ravindra Nirbhavane, Ashwini Nandu Pawar, Puneet Eknath Patel

Abstract: E-commerce platforms generate millions of product listings, often causing information overload and generic, non-personalized suggestions. Traditional recommendation systems operate as black boxes, resulting in limited user trust due to the lack of transparency. This paper proposes an AI-driven Explainable Product Recommendation System integrating Large. Language Models (LLaMA-2), FAISS semantic search, and SHAP-based interpretability. The system processes natural language queries, interprets intent, retrieves relevant products across multiple platforms, and generates human-readable explanations. Experimental evaluation demonstrates improved accuracy, transparency, and user satisfaction compared to traditional recommendation approaches.

DOI: http://doi.org/10.5281/zenodo.20766113

Published by: vikaspatanker

Authors: Sanket Dudhade, Sahil Gilbile, Aditya Gavali, Atul Chaudhari

Abstract: Food safety concerns, particularly the presence of undeclared allergies and artificial ingredients, have significantly increased worldwide as a result of the exponential growth in the consumption of packaged foods. Customers' manual label reading is inefficient, error-prone, and frequently hampered by multilingual packaging and complex ingredient nomenclature. An innovative technique for automating the detection of allergens and additives is provided by Artificial Intelligence (AI) through the use of Deep Learning (DL), Natural Language Processing (NLP), and Optical Character Recognition (OCR). A comprehensive analysis of AI-based smart systems for detecting chemicals and allergies in packaged foods is presented in this study. It looks at benchmark datasets, talks about different machine learning and transformer-based models, looks at key performance validation measures, and looks at the architectures that are already in place. The article also discusses difficulties such as data imbalance, interpretability problems, and computing constraints in real-time systems. Experimental trends show that hybrid OCR–NLP frameworks achieve detection accuracies of over 97% on benchmark datasets and demonstrate greater generalization across languages and package formats.The results of the study indicate that integrating state-of-the-art AI technology into food safety systems has the potential to revolutionize consumer protection, regulatory compliance, and public health. The findings emphasize that AI models must be globally scalable, interpretable, and privacy-preserving in order to guarantee transparency and confidence in automated food labeling.

DOI: http://doi.org/10.5281/zenodo.20765726

Published by: vikaspatanker

Authors: Sahil Thange, Karan Shinde, Rushikesh Pingal, Shailesh Mogal, Vishal Chaudhari

Abstract: The project titled “Development of a Smart Agro AI Drone” focuses on designing a cost-effective and intelligent aerial spraying system aimed at improving agricultural productivity through automation. Indian farmers often encounter labour shortages, uneven pesticide application and rising operational costs. To address these challenges, the proposed system integrates Artificial Intelligence (AI) and GPS-based autonomous navigation within a quadcopter platform equipped with a liquid tank, pump, and atomising nozzles for precise and uniform spraying. AI algorithms support crop recognition, optimised flight-path generation, and obstacle avoidance, ensuring safe and efficient field operations. An embedded microcontroller with a flight controller enables stable flight, real-time data transmission, and improved system reliability, while lightweight structural materials enhance endurance and payload capacity. This work also develops a cost-efficient agricultural drone platform by combining low-cost hardware components, open source flight control architecture, lightweight mechanical design, and optimised edge AI processing. The prototype is evaluated based on spray coverage, flight time, payload capacity, endurance, and detection accuracy under varying field conditions and cost-per-hectare performance is compared against existing commercial drone systems. Results demonstrate that strategic component selection, modular mechanical design, and computational model optimisation significantly reduce overall system cost while maintaining effective spraying and monitoring performance. Overall, the Smart Agro AI Drone provides an affordable, intelligent, and practical solution that supports sustainable precision farming, particularly for small and medium scale farmers

DOI: http://doi.org/10.5281/zenodo.20765542

Published by: vikaspatanker

Authors: Jenyfal Sampson, Vihash.A, B.R.V.Dharma Raju, Sounder.K, S.P.Velmurugan, Rishi Kumar

Abstract: Back pain from sitting too long and bad posture is now one of the most common health problems for students, office workers, and computer users. Posture-tracking cameras and high-tech ergonomic furniture are some of the more common solutions, but they can be expensive, hard to set up, or raise privacy concerns. This paper presents an IoT-based Posture Monitoring and Back Pain Alert System developed with an ESP32 microcontroller, a flex sensor, a force sensor, and an MPU6050 accelerometer–gyroscope module. It is a cost-effective and user-friendly alternative. A Flutter mobile app collects the user's height and weight, which lets the system automatically set sensor thresholds for different body types. The ESP32 checks your posture in real time and sends sensor data to the app over Wi-Fi. If the app notices that you are slouching, putting too much pressure on your back, or tilting your torso too much, it will immediately show a posture alert message, telling you to fix the problem. The system runs on a Li-ion battery with a TP4056 charging module, which makes it portable and allows for continuous use. Experimental observations demonstrate that the personalized threshold mechanism markedly diminishes false alerts while enhancing comfort and user acceptance. The suggested design is a good, cheap, and private way to fix your posture and keep your back from hurting.

DOI: http://doi.org/10.5281/zenodo.20761757

Published by: vikaspatanker