Leaf disease detection using machine learning and deep learning: Review and challenges

Applied Soft Computing, Год журнала: 2023, Номер 145, С. 110534 - 110534

Опубликована: Июнь 22, 2023

Язык: Английский

---

Interpretable Machine Learning Techniques in ECG-Based Heart Disease Classification: A Systematic Review

Diagnostics, Год журнала: 2022, Номер 13(1), С. 111 - 111

Опубликована: Дек. 29, 2022

Heart disease is one of the leading causes mortality throughout world. Among different heart diagnosis techniques, an electrocardiogram (ECG) least expensive non-invasive procedure. However, following are challenges: scarcity medical experts, complexity ECG interpretations, manifestation similarities in signals, and comorbidity. Machine learning algorithms viable alternatives to traditional diagnoses from signals. black box nature complex machine difficulty explaining a model's outcomes obstacles for practitioners having confidence models. This observation paves way interpretable (IML) models as diagnostic tools that can build physician's trust provide evidence-based diagnoses. Therefore, this systematic literature review, we studied analyzed research landscape techniques by focusing on signal. In regard, contribution our work manifold; first, present elaborate discussion techniques. addition, identify characterize signal recording datasets readily available learning-based tasks. Furthermore, progress has been achieved interpretation using IML Finally, discuss limitations challenges interpreting

Язык: Английский

---

Enhancing coffee bean classification: a comparative analysis of pre-trained deep learning models

Neural Computing and Applications, Год журнала: 2024, Номер 36(16), С. 9023 - 9052

Опубликована: Апрель 1, 2024

Abstract Coffee bean production can encounter challenges due to fluctuations in global coffee prices, impacting the economic stability of some countries that heavily depend on production. The primary objective is evaluate how effectively various pre-trained models predict types using advanced deep learning techniques. selection an optimal model crucial, given growing popularity specialty and necessity for precise classification. We conducted a comprehensive comparison several models, including AlexNet, LeNet, HRNet, Google Net, Mobile V2 ResNet (50), VGG, Efficient, Darknet, DenseNet, utilizing coffee-type dataset. By leveraging transfer fine-tuning, we assess generalization capabilities classification task. Our findings emphasize substantial impact choice model's performance, with certain demonstrating higher accuracy faster convergence than conventional alternatives. This study offers thorough evaluation architectural regarding their effectiveness Through result metrics, sensitivity (1.0000), specificity (0.9917), precision (0.9924), negative predictive value F1 score (0.9962), our analysis provides nuanced insights into intricate landscape models.

Язык: Английский

---

Detection of plant leaf diseases using deep convolutional neural network models

Multimedia Tools and Applications, Год журнала: 2024, Номер 83(24), С. 64533 - 64549

Опубликована: Янв. 16, 2024

Язык: Английский

---

Deterioration identification of stone cultural heritage based on hyperspectral image texture features

Journal of Cultural Heritage, Год журнала: 2024, Номер 69, С. 57 - 66

Опубликована: Авг. 12, 2024

Язык: Английский

---

Dual view deep learning for enhanced breast cancer screening using mammography

Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Фев. 15, 2024

Abstract Breast cancer has the highest incidence rate among women in Ethiopia compared to other types of cancer. Unfortunately, many cases are detected at a stage where cure is delayed or not possible. To address this issue, mammography-based screening widely accepted as an effective technique for early detection. However, interpretation mammography images requires experienced radiologists breast imaging, resource that limited Ethiopia. In research, we have developed model assist mass abnormalities and prioritizing patients. Our approach combines ensemble EfficientNet-based classifiers with YOLOv5, suspicious detection method, identify abnormalities. The inclusion YOLOv5 crucial providing explanations classifier predictions improving sensitivity, particularly when fails detect further enhance process, also incorporated abnormality model. achieves F1-score 0.87 sensitivity 0.82. With addition detection, increases 0.89, albeit expense slightly lower 0.79.

Язык: Английский

---

Review of Botnet Attack Detection in SDN-Enabled IoT Using Machine Learning

Sensors, Год журнала: 2022, Номер 22(24), С. 9837 - 9837

Опубликована: Дек. 14, 2022

The orchestration of software-defined networks (SDN) and the internet things (IoT) has revolutionized computing fields. These include broad spectrum connectivity to sensors electronic appliances beyond standard devices. However, these are still vulnerable botnet attacks such as distributed denial service, network probing, backdoors, information stealing, phishing attacks. can disrupt sometimes cause irreversible damage several sectors economy. As a result, machine learning-based solutions have been proposed improve real-time detection in SDN-enabled IoT networks. aim this review is investigate research studies that applied learning techniques for deterring Initially first major SDN-IoT thoroughly discussed. Secondly commonly used detecting mitigating Finally, performance presented terms models' metrics. Both classical (ML) deep (DL) comparable attack detection. ML require extensive feature engineering achieve optimal features efficient Besides, they fall short unforeseen Furthermore, timely detection, monitoring, adaptability new types challenging tasks techniques. mainly because use signatures already known malware both training after deployment.

Язык: Английский

---

Lightweight Model for Botnet Attack Detection in Software Defined Network-Orchestrated IoT

Applied Sciences, Год журнала: 2023, Номер 13(8), С. 4699 - 4699

Опубликована: Апрель 7, 2023

The Internet of things (IoT) is being used in a variety industries, including agriculture, the military, smart cities and grids, personalized health care. It also to control critical infrastructure. Nevertheless, because IoT lacks security procedures lack processing power execute computationally costly antimalware apps, they are susceptible malware attacks. In addition, conventional method by which malware-detection mechanisms identify threat through known fingerprints stored their database. However, with ever-evolving drastic increase threats IoT, it not enough have traditional software place, solely defends against threats. Consequently, this paper, lightweight deep learning model for an SDN-enabled framework that leverages underlying resource-constrained devices provisioning computing resources deploy instant protection botnet attacks proposed. proposed can achieve 99% precision, recall, F1 score 99.4% accuracy. execution time 0.108 milliseconds 118 KB size 19,414 parameters. performance high accuracy while utilizing fewer computational addressing resource-limitation issues.

Язык: Английский

---

Advances and Challenges in Computer Vision for Image-Based Plant Disease Detection: A Comprehensive Survey of Machine and Deep Learning Approaches

IEEE Transactions on Automation Science and Engineering, Год журнала: 2024, Номер 22, С. 2639 - 2670

Опубликована: Апрель 29, 2024

As advancements in agricultural technology unfold, machine learning and deep approaches are gaining interest robust plant disease identification. Early detection, integral to productivity, has propelled innovations across all phases of detection. This survey paper provides a meticulous examination detection systems, elucidating data collection methodologies underscoring the pivotal role datasets model training. The narrative navigates through complex areas image processing techniques, segueing into an exploration various segmentation methods. emphasizes importance feature extraction selection illustrating their efficacy increasing classification accuracy. It examines process, embracing both traditional avant-garde methods, with particular spotlight on Convolutional Neural Networks (CNNs). study over one hundred seminal papers, anatomizing dataset utilizations, considerations, strategies. Overall, contemplates challenges permeating this vibrant field, addressing critical issues such as diversity, generalization, real-world applicability. Note Practitioners —To ensure crop health yield, timely precise is crucial. Our research, titled "Advances And Challenges Plant Disease Detection: A Comprehensive Survey Machine Deep Learning Approaches", datasets, advanced processing, techniques presents practitioners guide latest for enhanced by emphasizing significance highlighting capabilities convolutional neural networks By understanding highlighted challenges, diversity industry professionals can better equip themselves integrate these technological applications.

Язык: Английский

---

Deep learning-based image compression for enhanced hyperspectral processing in the protection of stone cultural relics

Expert Systems with Applications, Год журнала: 2025, Номер unknown, С. 126691 - 126691

Опубликована: Янв. 1, 2025

Язык: Английский

---