JOIV : International Journal on Informatics Visualization
Not a member yet
786 research outputs found
Sort by
Object Detection with YOLOv8 and Enhanced Distance Estimation Using OpenCV for Visually Impaired Accessibility
Get PDFAccessibility challenges for the visually impaired are getting more serious yearly. To address this issue, this study presents an advanced object detection system that utilizes YOLOv8, enhanced with OpenCV for distance estimation. The methodology involves data preparation with diverse scenarios to test system accuracy, including environments like busy streets and indoor settings. Precision, recall, and F1-score metrics evaluate performance under varying lighting conditions. Results show a decrease in performance during low-light conditions, emphasizing the need for adequate lighting for effective detection. The system also includes a real-time implementation with a panic button feature, allowing immediate activation of object detection and distance estimation processes. The results are translated into Indonesian using a translation service and converted to speech, making the information accessible to users. By integrating YOLOv8 and OpenCV, the research achieves an average object detection accuracy of 91% with a low error rate of about 3.6%. Rigorous testing and evaluation under various conditions ensure reliability and effectiveness. The implications of this research extend to real-time applications like navigation assistance for the visually impaired, highlighting the potential for improved quality of life and independence. Future work will focus on optimizing detection in low-light conditions, incorporating additional sensors like infrared cameras, and enhancing real-time text translation services for accurate information delivery to visually impaired users. Additionally, continuous training with diverse datasets will be conducted further to improve the robustness and accuracy of the detection system
Processing of Brain Images Dataset: Introducing a Novel LBP Features Extraction Method to Enhance the Prediction System of Brain Hemorrhage
Get PDFTo detect brain bleeding in CT images, this study presents an improved Local Binary Pattern (NLBP) operator for texture analysis in medical imaging. The suggested NLBP utilizes an XOR operation with multi-radius feature extraction (r=1 and r=2) to capture fine-grained and larger texture patterns. Applied techniques compare pixel intensities over two radii and use the NLBP operator on image patches. To emphasize sudden changes in texture, the binary patterns produced by these two radii were processed using XOR to highlight variations in pixel intensities. To achieve the goal of this study, four machine learning models were applied to the CT brain images dataset to identify hemorrhage cases from non-hemorrhagic. According to the results, the NLBP approach considerably improved classification performance over conventional LBP. The random forest algorithm achieved a superior prediction accuracy of 94.05% while employing the NLBP strategy for feature extraction, in contrast to only 70.03% accuracy obtained using the LBP method for a similar algorithm. The NLBP approach improved edge recognition and classification accuracy by highlighting differences between surrounding pixel brightness and capturing multi-scale texture information. It concluded from these results that the NLBP operator provides a reliable method for medical image analysis by combining XOR-based refinement with multi-radius extraction. Additional investigation may examine the use of NLBP in different imaging modalities and refine the feature selection procedure for enhanced performance in various settings
Low-Resolution Face Image Reconstruction Using Multi-Stage FSRCNN to Improve Face Detection and Tracking Accuracy in CCTV Surveillance
Get PDFFace detection and tracking under real-world condition remain challenging under different illumination, crowded scenes, partial occlusions and small or low-resolution face images. In traditional face tracking schemes, these factors often cause the false positive rate to be high and the accuracy to be low. Specifically, little or no detailed information is presented for small or distant faces, here the reliability of detection is diminished and non-face-object can provoke false alarms thus degrading the performance of a system in general. Such problems are not unclear and need a sophisticated solution to improve the resolution and detection performance in various scenarios. In this paper, a new face tracking system based on a cascade classifier, a two-step model of Fast Super-Resolution Convolutional Neural Network (FSRCNN) and DLib face validator is presented. The low-resolution facial parts are first enhanced by the FSRCNN to optimize the detection by the cascade classifier. The DLib face validator improves the approach by validating the discovered faces, and reducing false positives. The system was tested over a CCTV scenario video corpus of several challenging conditions represented by crowded environments, dynamic object and human faces of different sizes and locations. The performance analysis focused on performance metrics such as precision, recall, and false positive rate, which provided a comprehensive overview of the system's robustness. The results demonstrate a significant improvement in face detection accuracy, as high as 98% precision and very few false positive detections. The synergy between the FSRCNN method and the DLib validation was especially effective on small and far-away faces, which are normally difficult to perceive. Whilst their improvements on memory consumption were small, they proved effective for face detection in challenging conditions. The ability of the system to maintain high measurement accuracy while avoiding errors makes it well suited for use in surveillance, security and monitoring systems. In conclusion, this research highlights the effectiveness of combining super-resolution techniques with traditional face detection methods to address the limitations of existing systems. The future work will focus on increasing recall rate and constantly maturing the extraction system to work well in various realistic conditions, thus making it effective and general for different applications
DeepForgery Images Detection Using Deep Learning Approaches and Error Level Analysis
Get PDFThe increasing of manipulated images, often shared on social media platforms, poses significant challenges for distinguishing authentic content from forgeries. This study aims to enhance the detection of tampered images by integrating Error Level Analysis (ELA) with Convolutional Neural Networks (CNNs). Specifically, the objectives are to evaluate the performance of two CNN architectures, VGG16 and MesoNet, combined with ELA preprocessing, and to identify potential avenues for future improvements in forgery detection. The dataset used comprises 7,492 authentic and 5,124 tampered images, sourced from the CASIA database, and is complemented with images from the Milborrow University of Cape Town (MUCT) dataset. Images were preprocessed using ELA to amplify discrepancies caused by tampering before being analyzed by the CNN models. The results indicate that the proposed ELA-VGG16 model achieved an accuracy of 86.786%, while the ELA-MesoNet model demonstrated superior performance, with an accuracy of 92.7%. These findings highlight the potential of combining ELA preprocessing with CNN architectures for robust image forgery detection. Despite fluctuations in training curves and instances of overfitting, the model effectively detects subtle manipulations in the majority of cases. However, challenges such as false positives and generalization to diverse datasets persist. Future research should explore enhancements such as expanded data augmentation, the integration of multi-model architectures,such as Xception or capsule networks, and advanced preprocessing techniques, which could further refine the model’s applicability and accuracy. These efforts would advance both the practical detection of forgeries and theoretical developments in informatics visualization, addressing critical challenges in digital forensics and media integrity
Improving YoloPX using YoloP and Yolov8 for Panoptic Driving Perception
Get PDFAutonomous driving technology (ADS) has seen significant advancements over the past decade, with car manufacturers investing heavily in its development to meet the growing demand for safer, more efficient, and eco-friendly transportation solutions. The panoptic driving perception system is central to ADS, essential for accurately interpreting the driving environment. This system requires high precision, lightweight design, and real-time responsiveness to detect surrounding vehicles, lane lines, and drivable areas effectively. This study introduces an enhanced YOLOPX model that combines YOLOP and YOLOv8 to create an adaptive multi-task learning network capable of traffic object detection, drivable area segmentation, and lane detection. The model integrates YOLOP's detection head with YOLOPX's anchor-free detection head to improve generalization, incorporates YOLOv8's advanced backbone structure to enhance feature extraction accuracy, and retains YOLOP's three-neck architecture to optimize multi-task processing. The improved model employs a mode loss function for segmentation tasks, enhancing generalization and improving lane detection accuracy. Experiments conducted using the BDD100k dataset demonstrated the model's effectiveness: achieving 98.8% accuracy and 27.6% IoU for lane line detection, 90.4% mIoU for drivable area segmentation, and 85.9% recall and 76.9% mAP50 for traffic object detection. This model represents a significant advancement in ADS, enhancing both the safety and reliability of autonomous vehicles
Comparative Analysis of Weighted-KNN, Random Forest, and Support Vector Machine Models for Beef and Pork Image Classification Using Machine Learning
Get PDFThe actual problem that occurs in the sale of meat by some conventional market traders is mixing beef with pork because of the high selling price. The difference between pork and beef lies in the color and texture of the meat. However, many people do not understand this difference. This study aims to provide a solution to distinguish the two types of beef through a classification process by obtaining the best accuracy using the W-KNN, RF, and SVM models based on machine learning. This study compares the model's performance based on the number of datasets, comprising 400 original images (200 beef and 200 pork images), using a 80:20 ratio for training and test data. The extraction process uses two algorithms: HSV (Hue, Saturation, Value) and RGB (Red, Green, Blue). The model evaluation uses a confusion matrix that includes accuracy, Precision, Recall, and F1-score. Based on the results of the model testing, it was found that the random forest algorithm gave the best overall results, with the highest accuracy of 98.75%, Precision of 97%, F1-score of 98%, and recall of 99% on the number of decision trees of 400. This shows the stability and generalization of the superior model. The random forest algorithm is the most effective for classifying beef and pork data with minimal errors. Implications for further research include using a deep learning approach, especially for image processing, to detect differences in each meat characteristic and increase accuracy
Exploring the Role of Machine Learning and Big Data Analytics in Enhancing Decision-Making Processes: A Systematic Literature Review
Get PDFThis Systematic Literature Review (SLR) analyzes the influence of Machine Learning (ML) and Big Data Analytics (BDA) on decision-making processes in several industries. The study aims to explore the potential of machine learning and big data analytics in enhancing decision-making, examining the tools and platforms used, and identifying the challenges encountered during deployment. Employing the PRISMA technique, 31 publications published from 2019 to 2024 were meticulously selected through a stringent screening process, using Scopus as the principal database. The results indicate that machine learning and big data analytics substantially enhance predictive accuracy, operational efficiency, and data privacy measures, while facilitating seamless integration with current systems. Furthermore, these technologies are becoming progressively accessible to Small and Medium Enterprises (SMEs). In the healthcare sector, machine learning models have exhibited a diagnosis accuracy of 99% in detecting breast cancer. Nonetheless, the report underscores other research deficiencies, particularly the necessity for more cost-effective solutions designed for SMEs. These limitations signify opportunities for future study to investigate ML and BDA applications in underexamined areas, such as logistics and manufacturing. This research highlights the necessity of creating economical, scalable, and industry-specific machine learning and big data analytics solutions to address existing difficulties. This systematic literature review (SLR) seeks to elucidate the function of machine learning (ML) and big data analytics (BDA) in decision-making, thereby assisting researchers and practitioners in enhancing the utilization of these technologies across many industrial applications
Smart Control Upgrade: IoT-Enhanced Remote Management of Straightening Machine with NodeMCU ESP8266
No full textStraightening machines are essential in the oil and gas industry. These machines are intended to straighten bent drill pipes after the drilling process. The operator must be near the straightening machine for manual operation and control. This limitation prevents the operator from monitoring the machine's performance directly and efficiently. In response to this challenge, this study proposes to use the Internet of Things to enable remote control and monitoring of the straightening machine using the NodeMCU ESP8266 microcontroller. The proposed system integrates sensors and electronic components to collect and process critical data from the straightening machine. The INTER-METH method develops an IoT system to control the straightening machine remotely. The RemoteXY cloud server receives the collected data via a Wi-Fi network using the Message Queuing Telemetry Transport protocol. This setup enables real-time data retrieval by an Android-based application. The proposed system allows the user to monitor the machine's operational status and control it remotely, thereby increasing efficiency by up to 45% and reducing the total cost of operation by about 60%. With the implementation of IoT, the operator could remotely monitor the performance of the straightening machine, thereby increasing operator safety by more than 50%. These advances improve operators' safety conditions and simplify maintenance and operational processes. In addition, future research could focus on robust data security measures. These developing systems could be operated with other industrial IoT platforms and utilize data analytics for predictive maintenance, extending the machine's life
QSroute: Enhancing Software Defined Networking Routing Scheme through Advanced QoS Metrics Integration
Get PDFThe surge in global internet traffic, primarily fueled by the rise of streaming platforms, cloud computing, and data-intensive applications, has posed significant challenges in effectively managing network resources. Ensuring consistent Quality of Service (QoS) across various types of traffic, such as video streaming, online gaming, and real-time communications, is becoming an increasingly challenging task. Traditional routing techniques, such as best-effort and shortest-path methods, are increasingly falling short in meeting these demands due to their inability to account for different traffic requirements. To address these limitations, this paper introduces QSroute, a novel QoS-based routing (QBR) scheme specifically designed for Software Defined Networking (SDN) environments. QSroute leverages the global view provided by SDN to dynamically optimize routing decisions based on critical QoS metrics, including available bandwidth, packet delay, and packet loss ratio. Our approach computes the optimal path between source and destination nodes by minimizing a composite link cost derived from these metrics. Extensive simulation results show that QSroute significantly outperforms existing QBR schemes, particularly in terms of reducing end-to-end delays, minimizing jitter, and improving overall throughput. These performance gains underscore QSroute’s potential as a highly effective solution for addressing the complex demands of modern network environments, offering enhanced scalability and network efficiency. Future research will investigate the integration of additional QoS metrics, as well as the scheme's scalability in more complex, multi-domain SDN environments, to further enhance its applicability
Optimizing Artificial Neural Network for Customer Churn: Advanced Data Balancing and Feature Selection
Get PDFCustomers are valuable assets in the dynamic business world. However, service dissatisfaction often leads them to switch to competitors, a phenomenon known as customer churn. In the telecommunications industry, churn poses a significant challenge as it directly impacts revenue and influences other customers within their social networks to do the same. Consequently, predicting churn has become essential, with numerous researchers employing various methods to classify potential churners. This study builds upon prior research that utilized Artificial Neural Networks (ANN) or Deep Learning to predict churn, achieving an accuracy of 88.12%. To improve model performance, this research implements an Artificial Neural Network (ANN) as the primary algorithm, along with Random Over-Sampling (ROS) and Synthetic Minority Oversampling Technique (SMOTE) for data balancing, and three feature selection methods: Minimum Redundancy Maximum Relevance (mRMR), Lasso Regression, and XGBoost. The results demonstrate a 0.38% increase in accuracy compared to previous studies. The finding suggests opportunities for further exploration. Future studies can consider alternative feature selection techniques, such as Wrapper Methods or Heuristic/Metaheuristic approaches, which may produce more optimal feature combinations. Other data balancing methods, such as Undersampling techniques (e.g., Random Undersampling, Tomek Links) or Hybrid Methods (e.g., SMOTE combined with Tomek Links), could be explored to address imbalanced datasets effectively. These approaches are expected to provide better combinations and to improve overall prediction performance, enabling researchers to develop more robust and accurate models for customer churn prediction in subsequent studies