Journal of Recent Innovations in Computer Science and Technology
Journal of Recent Innovations in Computer Science and TechnologyNot a member yet
25 research outputs found
Sort by
Deep Learning Based Text Extraction from Video Using CNN, LSTM, and Transformer Models
Get PDFThis study offers a deep learning-based method for text extraction from video frames, addressing issues like motion blur, variable text orientations, and background noise. Traditional optical character recognition (OCR) methods like Tesseract suffer from these problems, while contemporary deep learning models offer notable advancements. The suggested model uses Convolutional Neural Networks (CNNs) to identify text regions, Transformer-based models to increase recognition accuracy, and Long Short-Term Memory (LSTM) networks to maintain sequences. Several tests demonstrate that by striking a balance between accuracy and real-time functionality, the CNN + LSTM architecture performs better than conventional OCR algorithms. The results show that transformer-based methods have the highest accuracy but the highest computational cost. deep learning models like CNN, LSTM, and Transformers can handle contextual recognition, temporal sequencing, and spatial detection, they are particularly well suited for video text extraction. This hybrid approach, in contrast to traditional OCR, guarantees high accuracy even in video frames that are noisy, blurry, or multilingual
Potato Leaf Disease Detection Method is based on a CNN model with a Genetic Algorithm
Get PDFThis paper puts forward an approach that combines Convolutional Neural Networks (CNNs) and Genetic Algorithm (GA) to detect accurately and quickly the diseases that affect potato leaves swiftly and accurately. The model of the CNN is employed for automated feature extraction from the images of leaves which are pivotal in differentiating between the healthy and the infected leaves. Optimization of CNN hyper parameters, like learning rate, the number of layers, and dimensions of filters, is not only time-consuming but also challenging to find optimal values, and genetic algorithms for the optimal values of the parameters of the Convolutional Neural Network. A genetic algorithm uses an iterative search process that aims to optimize a population of models of Convolutional Neural Networks over several generations. It achieves this through the use of optimal models to perform crossover and mutation hence efficiently searching for the optimal configuration of hyperparameters. The above combine will improve the performance of the model which leads to better performance in detecting the variety of diseases attacking the potato leaves, such as early blight and late blight. In this paper CNN model utilizing a genetic algorithm attains an accuracy of 98.3% in 50 epochs.
Internet of Underwater things (IoUT): A Systematic Review Research
Get PDFThe development of intelligent systems for the monitoring, exploration, and administration of underwater environments is made possible by the Internet of Underwater Things (IoUT), which is a revolutionary development in marine and environmental research. This thorough research examines the advancements, challenges, and promise of IoUT with a focus on its applications in domains such as resource extraction, the science of oceanography underwater tracking, and tracking the environment. IoUT systems require customised approaches in processing information, conservation of energy, connectivity, and sensor that is being tested design because to the particular difficulties of underwater settings. The new protocols and methods designed for underwater applications—such as acoustic, optical, and electromagnetic communications—as well as the incorporation of artificial intelligence (AI) and machine learning (ML) technologies for improved data processing and taking decisions are covered in this study. Furthermore, we draw attention to the urgent issues surrounding data security, environmentally friendly interaction, and installation expenses while providing suggestions for future lines of inquiry and technical advancements
ARTIFICIAL INTELLIGENCE AND CLOUD-BASED COLLABORATIVE PLATFORMS FOR MANAGING EMERGENCY OPERATIONS
Get PDFEmergency management operations increasingly depend on cutting-edge technological solutions to support better disaster response, resource coordination, and recovery. This research uses artificial intelligence (AI) and cloud-based collaborative platforms to enhance emergency management in pre-disaster, disaster, and post-disaster phases. AI predictive abilities allow for early risk estimation, enhancing disaster forecast accuracy by 47% for wildfires and 42% for earthquakes. In emergencies, real-time data analysis and AI automated response cut response times from 12–24 hours to 2–6 hours, boosting situational awareness and resource allocation by 55%. Cloud platforms enable real-time sharing of data between emergency responders, which promotes the number of individuals contacted within the initial 48 hours by 200% and cuts down on incident costs by 60%. The research highlights a gap in AI-based decision-making systems and the scalability of the cloud, especially in developing countries. It suggests more interdisciplinary research to create standardized AI models for emergency management. The results underscore that AI and cloud platforms improve disaster response effectiveness, resource optimization, and cost savings and overcome data security, privacy, and system integration issues
AI-Driven Online Exam Proctoring: An Enhanced Machine Learning Approach
Get PDFThe proliferation of an online learning environment has opened up tremendous potential in the sphere of education, but has also posed significant problems to examination integrity. Conventional services of online proctoring, i.e., manual webcam supervision and lockdown browsers, failed to provide fairness since they were either inefficient or rather effortless to manipulate. This paper introduces an AI-based online exam proctoring (OEP) model that combines both visual and audio channels to identify cheating behavior, such as reading notes or using cell phones, muttering, or turning away the view on the examination. The research builds on a previously established framework that uses Support Vector Machines (SVMs), and tries different alternatives by using Random Forests (RF), Convolutional Neural Network (CNN), and Long Short-Term Memory (LSTM) models. A comparative analysis shows that the baseline system displayed an average True Detection Rate (TDR) of 87% under 2% False Alarm Rate (FAR) whereas the enhanced model with CNN-based visual processing and LSTM educated speech detection produced an improved system performance to 94% TDR at the same FAR limitation. The results indicate the potential of advanced ML to circumvent the drawbacks of earlier solutions and point to a way forward that results in scalable, fair, and privacy-sensitive proctoring solutions
Enhancing Organizational Performance and Strategic Forecasting Through Business Intelligence Technique
Get PDFIn this paper, show how BI applications can lead to sales forecasting and organizational performance improvements based on the case of a retail store. To detect patterns, performance issues and actionable results, a relatively simple business intelligence model based on descriptive, diagnostic and predictive analytics was applied. Although descriptive analytics revealed regional sales differences, diagnostic analytics found that too much discounting reduces profit. The comparison analysis confirmed that the proposals enable more profit, better discount management and improved performance in general when BI-based decision making is applied versus traditional methods. The present work evaluates five leading BI tools, namely Power BI, Tableau, Qlik Sense, Looker and SAP Business Objects based on their scalability, data integration and visualization features. In addition, future sales were predicted with the help of predictive models such as ARIMA, Prophet, etc., which contributed to inventory control and strategic alignment. The results highlight that BI solutions are core enablers of data-driven decision making, operational efficiency, and even continuous growth. With cloud and AI integration, BI provides real-time analytics to enable organizations stay competitive, agile, and responsive in today’s business landscape.
Facial Recognition and Object Detection using Machine learning
Get PDFFacial recognition and object detection are critical computer vision problems used in security, surveillance, autonomous systems, and human-computer interaction. This study investigates the of machine learning techniques. Use of facial recognition and object detection in deep learning has been develop to high level using machine learning. In enhancing the performance and enhancing the generalization of the model, this studies implements a recognition system on VGG16 model with data augmentation. To feed into the AI model, 1800 images were extracted from 17 classes and pre-processed, normalized and augmented also through rotation and flipping. The VGG16 model architecture was used and then trained by using categorical cross entropy loss function and Adam optimizer. The model achieved a validation accuracy of 76.39% within five epochs, signifying the model’s potential on various facial variations. Some of the issues that were pointed out include shortcoming such as misidentification and low quality images. Further researches propose to expand the dataset used for the study, to employ more complex base architectures such as ResNet, and to augment data pre-processing to improve recognition accuracy
Enhancing AI Decision-Making: Sensitivity Analysis, Hyperparameter Optimization, Multi-Agent Collaboration, and AI-Human Comparisons
Get PDFArtificial intelligence (AI) has significantly influenced decision-making processes across various domains, including law, healthcare, and autonomous systems. Despite its advancements, AI models face several critical challenges, including sensitivity to input variations, hyperparameter tuning complexities, coordination issues in multi-agent environments, and fundamental differences in decision-making compared to human cognition. This study investigates four key dimensions of AI decision-making: (1) the impact of input perturbations on AI-generated responses, (2) the role of hyperparameter tuning in optimizing AI performance, (3) the effectiveness of multiagent AI collaboration in ethical and strategic dilemmas, and (4) a comparative analysis of AI and human reasoning in realworld scenarios. The findings indicate that AI models exhibit response inconsistencies with minor input rewording, hyperparameter tuning significantly alters model accuracy and coherence, multi-agent AI systems struggle with consensus-building, and AI decision-making lacks ethical and emotional depth compared to human reasoning. This study highlights the need for robust AI training methodologies, structured decision-making protocols in multi-agent AI systems, and enhanced explainability frameworks to improve AI’s effectiveness and reliability in real-world applications
Development of a Smart Women Safety ID with Real-Time Gas Detection and Crime-Aware Emergency Alerts System
Get PDF In many parts of the world, women’s safety in public, educational, and professional settings is still a major concern. Conventional safety measures frequently depend on wearable technology or smartphone apps, which aren’t always reliable in an emergency or covert enough to keep potential criminals from spotting them. This study proposes a clever and affordable women’s safety ID card that offers both proactive and re active protection to overcome these drawbacks. This cutting-edge gadget incorporates essential technologies like a Bluetooth module (HC-05) for short-range device pairing, a SIM900A GSM module for emergency communication, a GPS module for location tracking, and an MQ2 gas sensor for hazardous gas detection. To ensure smooth operation at companies, educational institutions, and schools, the system is controlled by an Arduino Uno (ATmega328P) and housed in a standard ID card form factor. To detect the presence of dangerous compounds, including LPG, alcohol, and an aesthetic gases, labelled gas sensor data was used to train a supervised machine learning model, more precisely, a Random Forest Classifier. The model produced dependable real-time predictions with a high classification accuracy of 98.75% and strong precision, recall, and F1 scores. Furthermore, the system classifies areas as red zones by using historical crime data, such as location, crime type, and coordinates. The Haversine distance method is used to assess the user’s real-time GPS data and calculate how close they are to these high-risk areas. An automated alert is set off if the user goes inside a predetermined danger radius of four kilo metres. A manual SOS button is another feature of the safety system that allows users to send emergency alerts via SMS with their position and threat kind right away, even when there is no internet connectivity. Only authorized users, like guardians or institution leaders, can access critical data, track user movement, and receive alerts thanks to the secure dashboard and user authentication offered by the supporting mobile and online application. The interface, which was constructed with Gradio and Folium, provides real-time viewing of red zone boundaries, position, and gas detection status. This integrated system is an effective instrument to increase women’s safety because it not only guarantees situational awareness but also makes quick emergency reaction possible. Because of its scalable architecture, open-source. technologies, and low-cost hardware, it can be widely implemented in a variety of industries. The Women Safety ID Card is a step forward in utilizing artificial intelligence and embedded systems to create safer environments and give women the freedom, security, and self-assurance they need to go about their everyday lives
Human Skin Disease Detection and Classification Using Ensemble Learning
Get PDFSkin disorders are thought to be common in humans and carry several invisible risks, including the potential to cause psychological sadness, lower self-esteem, and, in more serious cases, skin cancer. Medical professionals must diagnose these skin conditions, but doing so requires highly sophisticated diagnostic tools because they have trouble seeing clearly while examining images of the conditions. This paper focuses on skin disease detection and classification using ensemble learning. This is done using Multiple Skin Disease Detection and Classification data sets from the ISIC Archive through employing the bagging, boosting, and stacking methodologies for better diagnosis. To compare the proposed ensemble of CNN models and individual CNN models, the observations were made. This strategy is useful for dermatologists to identify skin diseases at an early stage or at the first instance to prevent further deterioration of the skin health of their patients