IAES International Journal of Artificial Intelligence (IJ-AI)
Not a member yet
1769 research outputs found
Sort by
Catalysing precision in bone x-ray analysis for image detection and classification: the triple context attention model advancement
Get PDFAccurate detection and classification of fractures in bone x-ray images are crucial for effective medical diagnosis and treatment. In this study, we propose the triple context attention model (TCAN) as a novel approach to address the challenges in this domain. TCAN offers several key contributions that significantly enhance the accuracy and efficiency of bone x-ray image recognition and classification. Firstly, TCAN introduces the coordination attention mechanism, which considers both horizontal and vertical positional data during the recognition process. Secondly, TCAN mitigates the common issue of mislabelling fractures in bone x-ray images, particularly in the you only look once (YOLO) model, due to the absence of positional data during training. Thirdly, TCAN efficiently enhances positional data by focusing on weights, and increasing feature dimension while maintaining a manageable model size. This allows for effective utilization of positional data without computational overhead. Lastly, TCAN combines the visual attention network (VAN) with its capabilities, resulting in a comprehensive system that can handle diverse image dimensions and accurately classify various types of fractures across different body regions. Overall, TCAN presents a promising advancement in medical image analysis, improving fracture detection accuracy and classification efficiency in bone x-ray images, thus aiding in more effective clinical decision-making
Greywater treatment system based on fuzzy logic control
Get PDFGreywater from households and public facilities represents a major source of untreated wastewater, carrying high microbial loads and variable chemical composition that threaten environmental and public health. Conventional treatment systems often lack adaptive control mechanisms capable of handling the dynamic fluctuations of greywater quality. This study presents the design and validation of an intelligent greywater treatment system that integrates real–time sensing with a Sugeno fuzzy logic controller to regulate pump and solenoid valve operation. The system continuously monitors pH, total dissolved solids (TDS), dissolved oxygen (DO), and ammonia (NH3), and dynamically adjusts treatment cycles based on sensor feedback. Experimental deployment demonstrated significant improvements in effluent quality, with pH reduced from 9.04 to 8.08, TDS from 611.04 ppm to 393.96 ppm, and NH₃ from 0.52 ppm to 0.19 ppm, while DO increase from 2.52 mg/L to 6.07 mg/L. These results confirm that fuzzy logic–based control enhances system responsiveness and ensures effluent compliance under variable influent conditions. The proposed framework provides a scalable, cost-effective solution for decentralized wastewater management, advancing the development of intelligent treatment technologies for sustainable urban water systems
Designing a squeeze-and-excitation-capsule BiLSTM transformer for plant leaf disease recognition
Get PDFDeep learning (DL) is critical in plant disease recognition and classification with precision like those of expert human evaluators. However, development of effective systems is often disrupted due to the complexity and variability of disease pathogenesis. To address these challenges, this research applies to a hybrid DL architecture that integrates spatial encoding, sequential modelling, and attention for visual recognition. This proposed model can incorporate squeeze-and-excitation (SE) with residual blocks, capsule network (CapsNet), bidirectional long short-term memory (BiLSTM), and transformer network (TransNet)-based attention to realize spatial relationships and long-range dependencies for improving recognition accuracy. The proposed model is assessed on the corn leaf disease dataset (CLDD) and rice leaf diseases dataset (RLDD), and its performance is compared to leading-edge models. CLDD and RLDD achieved 99.88 and 99.10% training accuracy respectively. The area under the curve (AUC) reached almost ceiling recognition on CLDD, with 99.73, 99.96, 99.96, and 99.98% for blight (BL), common rust (CR), gray leaf spot (GL), and healthy (HE) result. RLDD results were also high, with 94.98, 93.70, 97.66, 84.57, 99.58, and 98.85% for bacterial leaf blight (BLB), brown spot (BS), HE, leaf blast (LB), leaf scald (LS), and narrow brown spot (NBS), respectively. The results of these tests show the remarkable promise and performance of the proposed model in plant disease recognition applications
Chest X-ray image classification using deep belief network with Al-Biruni earth radius and particle swarm optimization
Get PDFChest X-ray (CXR) is a widely employed radiological clinical assessment tool that provides a quick and effective means of classifying various diseases using CXR images. However, several researchers face challenges with CXR images due to imbalanced datasets and image quality issues. Pre-processing is performed using contrast limited adaptive histogram equalization (CLAHE) to enhance image quality and mitigate noise in the data. The synthetic minority oversampling technique (SMOTE) is applied to create synthetic samples for the minority class and handle class imbalance. The MobileNetV2 performs depth-wise separable convolution is used for feature extraction, while maintaining high efficiency for CXR images. This research proposes a deep belief network (DBN) to classify CXR, which helps capture hierarchical features and complex patterns in CXR images. The combination of particle swarm optimization (PSO) and Al-Biruni earth radius (BER) method is employed for hyperparameter tuning with enhanced DBN classification accuracy. Furthermore, BER is integrated with the PSO algorithm to balance exploration and exploitation while the fitness function is fine-tuned for optimal DBN classification performance. The proposed PSOBER-DBN achieves a high accuracy of 99.86% on the CXR14 dataset, in comparison to existing techniques such as the multi-level residual feature fusion network (MLRFNet)
Classification algorithm with artificial intelligence for the diagnostic process of obstructive sleep apnea
Get PDFObstructive sleep apnea (OSA) is a disease that affects millions of people worldwide, and a large proportion of them remain undiagnosed due to the high cost of polysomnography (PSG) tests. For this reason, it is crucial to develop affordable diagnostic tools to facilitate early detection of this condition. This study aims to analyze how an artificial intelligence (AI) based classification algorithm impacts the diagnostic process of OSA in Lima, Peru. The algorithm was developed following the Kanban methodology, which guaranteed an efficient and transparent follow-up during the development cycle, which is key in the medical context where software quality and traceability are fundamental. A decision tree (DT) was used for diagnosis and classification, employing a training dataset provided by the National Sleep Research Resource (NSRR), from which six relevant attributes were selected for analysis. The research results indicated that, although the improvement in clinical diagnostic accuracy was minimal at 10.81%, positive results were obtained in other aspects: diagnostic time was significantly reduced by 28.17%, and the number of tests required decreased by 24.07%
Trend analysis of machine learning techniques for traffic control based on bibliometrics
Get PDFMachine learning in traffic control for intelligent transportation systems (ML-ITSTC) aims to enhance user coordination and safety within transportation networks, ultimately improving overall traffic system performance. ML-ITSTC is achieved by leveraging data to execute machine learning algorithms in intelligent transportation management and optimizing traffic flow to prevent or reduce congestion. This paper conducts bibliometric analysis to explain the research status, development trajectory, and challenges of ML-ITSTC, drawing insights from literature in the Scopus database literature covering 2013 to November 2023. The bibliometric analysis of ML-ITSTC includes: performance analysis, science mapping analysis, and citation analysis. The evaluation of ML algorithm trends over the 10-year span indicates that traffic prediction (TP), neural networks, and deep learning are frequently used keywords. Further, an examination of keywords used over the entire period and in 2023 (up to November) shows that reinforcement learning (RL) is the latest popular approach for traffic control in transportation. The results provide a comprehensive view of the opportunities and challenges in ML-ITSTC, covering data, models, and applications, offering researchers insights into the current and future directions of ML-ITSTC research
An optimized transfer learning-based approach for Crocidolomia pavonana larvae classification
Get PDFThe increasing demand for mustard greens has driven farmers to continuously improve mustard greens cultivation. One of the challenges in mustard greens cultivation is the presence of insect pests. A significant pest in mustard greens is Crocidolomia pavonana (C. pavonana). C. pavonana damages plants by feeding on various parts, especially the leaves. The initial step in controlling them is insect pest monitoring. Monitoring aims to establish the control threshold. C. pavonana larvae have four instar stages: instar 1, 2, 3, and 4. Identification of the instar larval stages utilizes deep convolutional neural network (CNN) to classify C. Pavonana larvae on mustard greens using ResNet50V2 and DenseNet169 architectures optimized to enhance classification accuracy. The classification evaluation results show that both DenseNet169 and ResNet50V2 models achieve high accuracy, with DenseNet169 reaching the highest accuracy at 97.1%, while ResNet50V2 achieves an accuracy of 94.2%. The lower loss values on the test data compared to the validation data indicate that the deep learning models have successfully captured the patterns in C. pavonana images for classification. This classification process is expected to be one of the activities in monitoring the instar larvae to improve the accuracy of insecticide spraying and enhance mustard greens production
Camera-based advanced driver assistance with integrated YOLOv4for real-time detection
Get PDFTesting object detection in adverse weather conditions poses significant chal lenges. This paper presents a framework for a camera-based advanced driver assistance system (ADAS) using the YOLOv4 model, supported by an electronic control unit (ECU). The ADAS-based ECU identifies object classes from real-time video, with detection efficiency validated against the YOLOv4 model. Performance is analysed using three testing methods: projection, video injection, and real vehicle testing. Each method is evaluated for accuracy in object detection, synchronization rate, correlated outcomes, and computational complexity. Results show that the projection method achieves highest accuracy with minimal frame deviation (1-2 frames) and up to 90% correlated outcomes, at approximately 30% computational complexity. The video injection method shows moderate accuracy and complexity, with frame deviation of 3-4 frames and 75%correlated outcomes. The real vehicle testing method, though demand ing higher computational resources and showing a lower synchronization rate (> 5 frames deviation), provides critical insights under realistic weather condi tions despite higher misclassification rates. The study highlights the importance of choosing appropriate method based on testing conditions and objectives, bal ancing computational efficiency, synchronization accuracy, and robustness in various weather scenarios. This research significantly advances autonomous ve hicle technology, particularly in enhancing ADAS object detection capabilities in diverse environmental conditions
Comparing bidirectional encoder representations from transformers and sentence-BERT for automated resume screening
Get PDFIn today’s digital age, organizations face the daunting challenge of efficiently screening an overwhelming number of resumes for job openings. This study investigates the potential of two state-of-the-art natural language processing models, bidirectional encoder representations from transformers (BERT) and sentence-BERT (S-BERT), to automate and optimize the resume screening process. The research addresses the need for accurate, efficient, and unbiased candidate evaluation by leveraging the power of these transformer-based language models. A comprehensive comparison between BERT and S-BERT is performed, evaluating their performance across multiple metrics, including accuracy, screening time, correlation with job descriptions, and ranking quality. The findings reveal that S-BERT outperforms BERT, achieving higher accuracy (90% vs. 86%), faster screening time (0.061 seconds vs. 1 second per resume), and stronger correlation with job descriptions (0.383855 vs. 0.1249). S-BERT though has a smaller vector size of 384 enables capturing richer semantic information compared to BERT’s vector size of 768, contributing to its superior performance. The study provides insights into the strengths and limitations of each model, offering valuable guidance for organizations seeking to streamline their talent acquisition processes and enhance candidate selection through automated systems
Intent detection in AI chatbots: a comprehensive review of techniques and the role of external knowledge
Get PDFArtificial intelligence (AI) chatbots have become essential across various industries, including customer service, healthcare, education, and entertainment, enabling seamless, and intelligent user interactions. A key component of chatbot functionality is intent detection, which determines the underlying purpose of user queries to provide relevant responses. Traditional intent detection methods, such as rule-based and statistical approaches, often struggle with adaptability, especially in complex, dynamic conversations. This review examines the evolution of intent detection techniques, from early methods to modern deep learning and knowledge-enriched models. It introduces the domain type-conversation turns-adaptivity-external knowledge (DCAD) classification, highlighting its significance in improving chatbot accuracy and contextual awareness. The paper categorizes existing intent detection models, analyzes their applications across various sectors, and discusses key challenges, including data integration, language ambiguity, and ethical concerns. By exploring emerging trends and future directions, this review underscores the critical role of external knowledge in enhancing chatbot performance and user experience