Articles

Access the latest knowledge in applied science, electrical engineering, computer science and information technology, education, and health.

Advanced Search
Filter Icon

Filters article

Years

FAQ Arrow
Single Year
Range Year
0
0

Source Title

FAQ Arrow

Authors

FAQ Arrow

30,938 Article Results

Automated bacteria and fungi classification using convolutional neural network on embedded system

10.11591/ijai.v15.i2.pp1132-1142
Tarik Bouganssa , Maryem Ait Moulay , Samar Aarabi , Abedelali Lasfar , Abdelatif EL Afia
In this study, we created and applied novel concepts for hardware-based image identification and categorization. For artificial intelligence (AI) and image recognition applications, this includes putting algorithms for recognizing colors, textures, and shapes into practice. Our contribution uses an embedded device with a camera and a microcomputer (Raspberry-Pi4 type) to replace the optical assessment of Petri dishes. Our object recognition system processes images efficiently by using a state-of-the-art kernel function and a new neighborhood architecture. Using the well-known convolutional neural network (CNN) architecture, YOLOv8, as a pre-trained model, we evaluated the proposed CNN-based method for object recognition in a number of demanding scenarios. Several Petri plates, uncontrolled settings, and different backgrounds and illumination were used to evaluate the technology. Our dynamic mode integrates a CNN network with an attention mask to highlight the traits of bacteria and fungi, ensuring robust recognition. We implemented our algorithm on a Raspberry Pi 400, connected to a CMOS 3.0 camera sensor and a human-machine interface (HMI) for instant display of results.
Volume: 15
Issue: 2
Page: 1132-1142
Publish at: 2026-04-01

Deep learning for mental health analysis: long short-term memory approach to text-based condition classification

10.11591/ijai.v15.i2.pp1762-1770
Zaqqi Yamani , Dinda Lestarini , Sarifah Putri Raflesia , Purwita Sari , Ghita Athalina
The increasing prevalence of mental health disorders highlights the need for scalable and automated approaches to early detection. This study proposes a deep learning–based text classification framework using a long short-term memory (LSTM) network to identify mental health conditions from user generated textual data. A corpus of 103,488 labeled texts representing anxiety, stress, bipolar disorder, depression, personality disorder, suicidal ideation, and normal states was preprocessed through tokenization, padding, and word embedding. The proposed LSTM model achieved overall accuracy of 87% on test set, with strong class-wise performance reflected by precision, recall, and F1-scores, particularly for anxiety, personality disorder, and normal classes. Comparative error analysis using a confusion matrix revealed challenges in distinguishing depression from suicidal ideation, indicating semantic overlap between these conditions. The results demonstrate that LSTM-based models can effectively capture sequential linguistic patterns relevant to mental health classification. This framework shows potential as a decision-support tool for early screening and digital mental health applications, complementing clinical assessment rather than replacing it.
Volume: 15
Issue: 2
Page: 1762-1770
Publish at: 2026-04-01

Unified voting-based ensemble learning for rice leaf disease detection using improved pretrained models

10.11591/ijai.v15.i2.pp1646-1663
Govindarajan Subburaman , Mary Vennila Selvadurai
As a staple food for a large portion of the global population, rice is particularly susceptible to leaf diseases that adversely affect its yield and overall quality. This study utilizes four pretrained convolutional neural network (CNN) models to construct a unified voting-based ensemble approach for rice leaf disease classification. The models include VGG16, DenseNet121, InceptionV3, and Xception. The dataset used in this study was collected from Kaggle and further enriched with images obtained from Google sources. It comprises a total of 4,000 images categorized into six classes: bacterial leaf blight, brown spot, leaf blast, leaf scald, narrow brown spot, and healthy leaves. It was split into training (327 images/class), validation (140 images/class), and testing (200 images/class). Images were normalized to [0,1] and augmented through rotation, flipping, shifting, shear, zoom, brightness, and channel adjustments to improve generalization. Individually, the fine-tuned models achieved accuracies of 91.3% (VGG16), 95.6% (DenseNet121), 92.1% (InceptionV3), and 89.8% (Xception). The ensemble leveraged majority voting (93.6%), weighted voting (96.5%), and soft voting (97%), yielding an absolute gain of 1.4% over the best individual model and 4.8% over the average of all models. To our knowledge, this is the first ensemble combining these four architectures with unified voting for identifying diseases in rice leaves, delivering a scalable and computationally efficient solution suitable in advance diagnosis and timely execution in agricultural settings with limited resources.
Volume: 15
Issue: 2
Page: 1646-1663
Publish at: 2026-04-01

IoT-enabled smart nutrition scale using fuzzy logic for dietary assessment and recommendation

10.11591/ijai.v15.i2.pp1194-1201
Wahyu Wijaya Widiyanto , Edy Susanto , Sri Suparti
Childhood malnutrition, particularly stunting, remains a major public health challenge that requires preventive and technology-supported nutritional interventions. This study presents an IoT-enabled smart nutrition scale integrated with fuzzy logic to support real-time dietary assessment and personalized recommendation. The system combines IoT-based sensing, mobile and web applications, and a fuzzy inference engine that evaluates child profiles and food composition data to generate nutritional adequacy scores and tailored dietary guidance. Experimental validation demonstrates high measurement accuracy of the sensing system, achieving a strong linear correlation (R² ≈0.9995). Comparison with expert nutritionist assessments shows strong agreement, supported by low error values (mean absolute error (MAE) =2.96; root mean square error (RMSE) =3.41), and Bland–Altman analysis. Usability evaluation involving community health workers and caregivers yields an excellent system usability scale (SUS) score, indicating strong acceptance for practical deployment. By integrating IoT sensing with fuzzy reasoning, the proposed system shifts nutritional monitoring from retrospective assessment toward proactive dietary intervention. This work highlights the potential of intelligent nutrition technologies to enhance decision-making in community-based stunting prevention programs and provides a scalable foundation for preventive digital health applications.
Volume: 15
Issue: 2
Page: 1194-1201
Publish at: 2026-04-01

A reinforcement-guided multi-phase hybrid architecture for threat profiling and defense towards IoT handheld device

10.11591/ijai.v15.i2.pp1497-1504
Pushpa Rajput Narayana Singh , Neelambike Siddalingaiah
The contribution of artificial intelligence (AI) towards offering proactive security in handheld devices of internet of things (IoT) is in evolving stage. Review of literature showcases noteworthy attempts of machine learning (ML) and deep learning (DL) models; however, they are a large scope of improvement towards bridging the trade-off between security and computational-communication efficiency. This problem is addressed in this manuscript by presenting a unique and innovative solution where reinforcement learning (RL) has been hybridized with standalone ML and DL models. The model reads the permission-based data in cloud, followed by vulnerability prediction carried out by hybridization of RL and logistic regression (LR). Further, RL is integrated with deep neural network (DNN) for exploring a secure path to facilitate data transmission. The proposed model witnessed 97.9% accuracy, 67.35% of higher accuracy, 55.14% of reduced latency, and 52.54% of faster response time in contrast to baselines.
Volume: 15
Issue: 2
Page: 1497-1504
Publish at: 2026-04-01

Recognition system based on artificial vision using OpenCV for discarding and detecting ceramics with defects

10.11591/ijai.v15.i2.pp1166-1173
Fernando Alvarado , Ricardo Yauri
Early detection of defects through preventive maintenance is important in industry to avoid economic losses, as in the case of ceramic tile manufacturing, where manual inspection allows defective parts to advance in production, causing delays. The research review shows that computer vision enables the automation of object detection, classification, and elimination tasks in industrial processes, using solutions based on Python, OpenCV, and MATLAB. For this reason, the design of a computer vision recognition system with OpenCV is proposed, which allows automatic discarding of ceramics with defects using an algorithm for detecting ceramics with a camera and Arduino-based hardware, comparing the captured images with a standard image on a conveyor belt. The machine vision system was integrated with a camera connected to a computer running OpenCV, achieving effective automatic detection with a threshold of 25% difference from the standard part. This percentage was calculated by comparing the grayscale pixel values with a reference image. The system calculates the proportion of pixels that exceed the similarity threshold. The conclusion is that the developed system contributes to production, highlighting the possibility of future industrial integration.
Volume: 15
Issue: 2
Page: 1166-1173
Publish at: 2026-04-01

Enhancing imbalanced dataset diagnosis using class-based input image composition

10.11591/ijai.v15.i2.pp1613-1622
Azzeddine Hlali , Majid Ben Yakhlef , Soulaiman El Hazzat
Deep learning models often falter when faced with small, imbalanced datasets or degraded image quality, leading to unacceptably high false prediction rates. To bridge this gap, we introduce class-based image composition. This technique reformulates training inputs by fusing multiple intra-class images into unique composite input images (CoImg). By concentrating information density and amplifying intra-class variance, CoImg forces the model to discern subtle, nuanced disease patterns that might otherwise be lost. We validated this approach using the optical coherence tomography dataset for image-based deep learning methods (OCTDL), a collection of seven imbalanced retinal disease scan categories. From this, we engineered Co-OCTDL: a perfectly balanced variant where each training sample exists as a 3×1 layout composite. To measure the impact of this new representation, we benchmarked the original dataset against its composite counterpart using a VGG16 architecture. Precision was paramount. We maintained identical hyperparameters and model structures across all experiments to ensure a rigorous, fair comparison. The results were transformative. While baseline datasets struggled, the enhanced Co OCTDL achieved a near-perfect F1-score of 0.995 and an AUC of 0.9996. The method effectively neutralized the risks of class imbalance. It didn't just improve the numbers; it refined the diagnostic reliability of the model.
Volume: 15
Issue: 2
Page: 1613-1622
Publish at: 2026-04-01

Technical analysis model for stock prediction using a grammatical evolution algorithm

10.11591/ijai.v15.i2.pp1236-1246
Aditya Kusuma Setyanegara , Imas Sukaesih Sitanggang , Mushthofa Mushthofa
Stocks are a popular investment instrument but carry high risks, where investors may incur losses when stocks are bought at high prices and sold at lower prices. Technical analysis is used to study past stock price behavior to predict future prices. In this study, grammatical evolution (GE) is applied as an evolutionary computing technique to discover optimal functions or programs that represent historical stock price data. This study develops GE based prediction models by utilizing objective functions and search spaces defined through grammar. The model integrates technical indicators based on complex statistical models such as autoregressive integrated moving average (ARIMA), prophet, exponential smoothing, and Fibonacci retracements. Furthermore, this study employs GE to generate ensemble weights randomly, ensuring each model contributes equitably to the final prediction formula. Experiments were conducted using multiple stock datasets, including SMAR, S&P 500, the Johannesburg Stock Exchange (JSE), the New York Stock Exchange (NYSE), and Adani Enterprises (ADANIENT), to evaluate the model’s adaptability and generalization capability. The results demonstrate that the proposed GE model effectively captures complex market patterns and produces more reliable stock price predictions compared to deep learning-based approaches. Although GE requires greater computational time, the findings suggest that GE provides a flexible and effective framework for constructing hybrid stock price forecasting models in dynamic market environments.
Volume: 15
Issue: 2
Page: 1236-1246
Publish at: 2026-04-01

2D-CNN-GACL-ECGNet graph attention: a robust framework for electrocardiogram-based stress detection

10.11591/ijai.v15.i2.pp1529-1538
P. Kavitha , L. Shakkeera
Early detection of cardiovascular diseases (CVDs) via electrocardiogram (ECG) classification during physiological stress is critical and remains challenging due to stress-induced morphological variability, noise from ambulatory settings, and inter-class ambiguities. Existing models, such as 1D signal-based models with convolutional neural networks (CNNs) and graph convolutional networks (GCNs), struggle to adapt to dynamic stress conditions and generate interpretable insights. In response, we propose 2D CNN and graph attention network (GAT) for optimizer. The model 2D-CNN and GACL-ECG-Net, an innovative framework integrating GATs with adaptive contrastive learning (ACL) and morpho-temporal graph construction. Key innovations include 2D-CNN denoising, 2D transformation, dynamic morpho-temporal graphs modeling ECG beats as nodes with hybrid edges (70% morphological similarity, 30% temporal proximity), and stress-adaptive contrastive loss with learnable margins on stress-conditioned labels, reducing class ambiguity by 18%. Multi-head attention mechanisms provide interpretable heatmaps aligned with cardiologist annotations (κ =0.82) and are evaluated using Massachusetts Institute of Technology-Beth Israel Hospital (MIT-BIH) arrhythmia database, wearable stress and affect detection (WESAD) dataset for emotional stress, and stress at work, knowledge work (SWELL-KW) dataset for cognitive stress. 2D-CNN-GACL-ECG-Net achieves state-of-the-art performance with 98.7% F1-score (MIT-BIH), 94.2% (WESAD), and 92.8% (SWELL-KW), outperforming CNN-bidirectional long short-term memory (BiLSTM) and GCN baselines by 95%. The framework is computationally efficient and clinically validated for wearable health monitoring.
Volume: 15
Issue: 2
Page: 1529-1538
Publish at: 2026-04-01

TMA-Net: a transformer-based multi-modal attention network for abnormal behavior detection

10.11591/ijai.v15.i2.pp1441-1450
Huong-Giang Doan , Ngoc-Trung Nguyen
Abnormal behavior detection in crowded environments remains challenging due to complex motion patterns, occlusions, and domain variability. This paper presents transformer-based multi-modal attention network (TMA-Net), a unified framework that integrates red, green, and blue (RGB), optical flow (OF), and heat map (HM) modalities through a dual-stage attention fusion mechanism. The system employs you only look once version 11 (YOLOv11) for human localization and vision transformer (ViT)-B/16 for feature encoding, followed by intra-modal self-attention and cross-modal fusion to capture fine-grained spatial–temporal and motion energy dependencies. Extensive experiments on six public benchmarks as UMN, Crowd-11, UBNormal, ShanghaiTech, CUHK Avenue, UCSD Ped2, and EPUAbN dataset, demonstrate that TMA-Net achieves up to 97.5% area under the curve (AUC) and 96–100% accuracy, outperforming previous other state-of-the-art approaches. These results highlight the framework’s strong generalization and robustness across both single- and cross-dataset evaluations, underscoring its potential for reliable deployment in real intelligent surveillance systems.
Volume: 15
Issue: 2
Page: 1441-1450
Publish at: 2026-04-01

Automated classification of apple bruises from hyperspectral images: an approach for fruit quality assessment

10.11591/ijai.v15.i2.pp1381-1389
Peddireddy Venkateswara Reddy , Alaguchamy Parivazhagan
Apple bruise detection plays a crucial role in post-harvest quality control; however, conventional manual inspection remains labor-intensive, subjective, and unsuitable for large-scale industrial deployment. This study proposes an automated classification framework for identifying bruised regions in apples using hyperspectral imaging combined with deep learning and adaptive optimization techniques. The proposed model integrates a long short-term memory (LSTM) network optimized using an adaptive sand cat swarm optimization (ASCSO) algorithm, along with a ResNet-50 feature extraction backbone. The adaptive behavior embedded within ASCSO dynamically adjusts the optimization parameters to enhance convergence and prevent premature stagnation during LSTM hyperparameter tuning. Hyperspectral images were processed to extract relevant spectral–spatial features, which were subsequently fed into the optimized classifier. Experimental evaluations demonstrate that the proposed hybrid model significantly outperforms conventional and baseline deep learning approaches, achieving a classification accuracy of 98.0% while maintaining robustness across varying bruise patterns and intensity levels. The results highlight the effectiveness of combining hyperspectral imaging with adaptive deep learning optimization for high-precision fruit quality assessment. This research contributes a reliable, scalable solution for automated bruise detection and quality grading in the fruit supply chain, offering strong potential to reduce post-harvest losses and improve operational efficiency in the agro-food industry.
Volume: 15
Issue: 2
Page: 1381-1389
Publish at: 2026-04-01

Breast cancer detection using residual DenseNets in deep learning

10.11591/ijai.v15.i2.pp1632-1645
Naganandini Gururajarao , Vishwanath R. Hulipalled
Breast cancer, the leading cause of cancer-related deaths among women globally, requires a prompt and precise diagnosis in order to increase survival rates via therapy. There is a possibility of bias and inconsistency in the results of traditional diagnostic procedures like mammography, ultrasound, and histological testing since they rely on the expertise of radiologists and pathologists. There are exciting new opportunities for breast cancer diagnostics to be enhanced by artificial intelligence (AI) and deep learning. The purpose of this research is to examine the feasibility of using convolutional neural networks (CNNs) and residual dense networks (ResDenseNets) used for breast cancer automated detection in medical images. Because of their superior capacity to learn hierarchical features from raw image data, CNNs are ideal for medical image interpretation. By including residual connections, which allow for the training of considerably deeper models, ResDenseNets—an extension of CNNs—mitigate the problem of vanishing gradient in deep networks. ResDenseNet and CNNs considerably enhance the accuracy of breast cancer diagnosis in comparison to conventional approaches, according to the findings. Notably, ResDenseNets outperform other types of networks because they are able to learn intricate and nuanced properties directly from the data.
Volume: 15
Issue: 2
Page: 1632-1645
Publish at: 2026-04-01

Novel convolution neural network model for dysgraphia affected handwriting classification

10.11591/ijai.v15.i2.pp1418-1427
Nisha Ameya Vanjari , Prasanna J. Shete
It is estimated that 10% of the population in the world suffers from learning disabilities like dyslexia, dysgraphia, and dyscalculia. Learning disabilities are neurological disorders in which children struggle with reading, writing and mathematical skills. Dysgraphia disorder impacts on writing abilities of students and thus may be a hurdle in their learning and evaluation of subject matter. Hence early detection/prediction of learning disability (LD) in school going children will greatly help in providing necessary accommodations so as to ease their future learning curve. In recent years researchers have used several deep learning algorithms that produce automated and trained models which can be useful in the handwriting classification. To properly capture the distinct handwriting inconsistencies linked to dysgraphia, this study contains experiments that determine how various convolution neural network (CNN) model layers contribute to performance. To address it, this research focused on the improved novel model based on CNN and targeted dysgraphia English handwriting classification with 98% accuracy with 102,691 trainable parameters. The model is trained on both normal and dysgraphia-affected handwriting, increasing its accuracy in identifying individual differences.
Volume: 15
Issue: 2
Page: 1418-1427
Publish at: 2026-04-01

Hybrid machine learning for imbalanced lettuce disease classification

10.11591/ijai.v15.i2.pp1783-1789
Fazlur Ihzanurahman , Wayan Firdaus Mahmudy
This study investigates a hybrid machine learning framework combining EfficientNet-B3 feature extraction with classical classifiers for lettuce disease classification under conditions of extreme class imbalance. The system utilizes EfficientNet-B3 to extract high-dimensional feature embeddings from 2,337 images, which are subsequently classified using support vector machine (SVM), random forest (RF), and k-nearest neighbors (KNN). Although the proposed SVM-based model achieves a high overall accuracy of 94.01%, experimental results reveal a substantial performance discrepancy compared to the macro F1-score of 37.94%. This critical gap indicates that while the model successfully identifies the majority classes, it fails to recognize rare disease categories with limited samples. Theoretical analysis suggests that while SVM handles high-dimensional feature spaces more effectively than RF and KNN, the deep features extracted are biased toward majority class characteristics. These findings highlight the severe limitations of accuracy-centric evaluation in agricultural diagnostics and demonstrate that deep feature extraction alone is insufficient to guarantee robust detection for minority pathologies. The study concludes that relying on aggregate accuracy can mask diagnostic failures, emphasizing the urgent need for per-class performance analysis and data-level mitigation strategies in future research.
Volume: 15
Issue: 2
Page: 1783-1789
Publish at: 2026-04-01

Image feature extraction for road surface damage classification

10.11591/ijai.v15.i2.pp1578-1592
Octaviani Hutapea , Sarifuddin Madenda , Hustinawaty Hustinawaty , Iffatul Mardhiyah
Road surface deterioration poses a critical risk to driving safety and comfort, necessitating timely and accurate detection to support effective maintenance. Manual inspection methods are often inefficient, underscoring the need for automated approaches based on computer vision. This study investigates the integration of feature extraction techniques histogram of oriented gradients (HOG) and local binary pattern (LBP) with convolutional neural network (CNN) architectures ResNet50 and InceptionV3 for the classification of road damage. A dataset of 1,580 images was categorized into five damage types: alligator crack, longitudinal crack, other crack, patching, and potholes. Experimental results indicate that HOG–ResNet50 achieved 79% accuracy, while LBP–InceptionV3 yielded the best performance at 97%. The contributions of this study are threefold: i) an automated framework is proposed that combines texture-based features with deep learning for road damage detection, ii) the LBP–InceptionV3 combination is shown to provide superior accuracy compared to conventional pairings, and iii) the approach offers a scalable and reliable alternative to manual inspection methods, supporting more efficient road maintenance planning.
Volume: 15
Issue: 2
Page: 1578-1592
Publish at: 2026-04-01
1 2 3
...
2062 2063
Show 73 of 2063

Discover Our Library

Embark on a journey through our expansive collection of articles and let curiosity lead your path to innovation.

Explore Now
Library 3D Ilustration