Revolutionizing lemon grading: an automated CNN-based approach for enhanced quality assessment

International Journal of Information Technology, Год журнала: 2024, Номер 16(7), С. 4155 - 4166

Опубликована: Май 2, 2024

Язык: Английский

---

BiCAE – A Bimodal Convolutional Autoencoder for Seed Purity Testing

Lecture notes in computer science, Год журнала: 2024, Номер unknown, С. 447 - 462

Опубликована: Янв. 1, 2024

Язык: Английский

---

SMOTE vs. KNNOR: An evaluation of oversampling techniques in machine learning

Gümüşhane Üniversitesi Fen Bilimleri Enstitüsü Dergisi, Год журнала: 2023, Номер unknown

Опубликована: Июнь 23, 2023

The increasing availability of big data has led to the development applications that make human life easier. In order process this correctly, it is necessary extract useful and valid information from large warehouses through a knowledge discovery in databases (KDD). Data mining an important part involves discovering developing models unknown patterns. quality used supervised machine learning algorithms plays significant role determining success predictions. One factor improves balanced dataset, where input values are distributed close each other. However, practice, many datasets unbalanced. To overcome problem, oversampling techniques generate synthetic as real possible. study, we compared performance two techniques, SMOTE KNNOR, on variety using different algorithms. Our results showed use KNNOR did not always improve accuracy model. fact, datasets, these resulted decrease accuracy. certain both were able increase indicate effectiveness varies depending specific dataset algorithm being used. Therefore, crucial assess methods case-by-case basis determine best approach for given algorithm.

Язык: Английский

---

Detection of Defects in Rolled Stainless Steel Plates by Machine Learning Models

International Journal of Applied Mathematics Electronics and Computers, Год журнала: 2023, Номер 11(1), С. 37 - 43

Опубликована: Март 31, 2023

Iron metal is the most widely used type. This metal, which in countless sectors, processed different ways and turned into steel. Since steel has a brittle structure compared to iron, defects may occur plates during rolling process. Detection of these at production stage great importance terms commercial safety. Machine learning methods can be such problems for fast high accuracy detection. For this purpose, using dataset obtained from stainless surface study, classification processes were carried out detect with four machine methods. Logistic Regression (LR), Decision Tree (DT), Support Vector (SVM) Random Forest (RF) algorithms processes. The highest was 79.44% RF model. Correlation analysis performed order analyze effects features on results. It thought that proposed models satisfactory challenging problem, but needs upgraded.

Язык: Английский

---

Enhancing Explainability in Plant Disease Classification using Score-CAM: Improving Early Diagnosis for Agricultural Productivity

Опубликована: Сен. 7, 2023

This study deals with the issue of explainability in classification plant diseases by deep learning methods. In particular, models is shown using Score-CAM method. a method used to identify important regions that affect decision model. this study, was applied analyzing images leaves, and it provided explain decisions model for diagnosis disease. As result, techniques help achieve more effective accurate results early diseases. turn helps reduce negative effects on economic food security increasing agricultural productivity.

Язык: Английский

---

Durum Wheat Classification Using Feature Selection, Bayesian Optimization and Support Vector

Опубликована: Янв. 11, 2024

Wheat is the primary component of majority everyday food items, and acquiring high-quality wheat grains a crucial concern for production products. Recognizing types durum vital during processing in food-processing facilities. A dataset that included two varieties extraneous substances was gathered. The objective this study to identify minimum number features from pool 236 morphological, color, wavelet, gaborlet features, which can yield highest accuracy with minimal difference between validation test kinds wheat: starchy vitreous foreign elements. This proposes machine learning approach optimal set distinguishing starchy, wheat, comprises feature selection, optimization, classification. First, five selection techniques, MRMR, ChiSquare, Relief, ANOVA, Kruskal-Wallis SVM, were evaluated identification wheat. After conducting analysis, it found out 50 yielded significant performance. However, also suffers decreasing 2-3% decrease accuracy. To compensate this, Bayesian optimization technique introduced achieved 99.8% 99.6%. methodology helps chain.

Язык: Английский

---

Classification of field wheat varieties based on a lightweight G-PPW-VGG11 model

Frontiers in Plant Science, Год журнала: 2024, Номер 15

Опубликована: Май 14, 2024

In agriculture, especially wheat cultivation, farmers often use multi-variety planting strategies to reduce monoculture-related harvest risks. However, the subtle morphological differences among varieties make accurate discrimination technically challenging. Traditional variety classification methods, reliant on expert knowledge, are inefficient for modern intelligent agricultural management. Numerous existing models computationally complex, memory-intensive, and difficult deploy mobile devices effectively. This study introduces G-PPW-VGG11, an innovative lightweight convolutional neural network model, address these issues. G-PPW-VGG11 ingeniously combines partial convolution (PConv) partially mixed depthwise separable (PMConv), reducing computational complexity feature redundancy. Simultaneously, incorporating ECANet, efficient channel attention mechanism, enables precise leaf information capture effective background noise suppression. Additionally, replaces traditional VGG11's fully connected layers with two pointwise a global average pooling layer, significantly memory footprint enhancing nonlinear expressiveness training efficiency. Rigorous testing showed G-PPW-VGG11's superior performance, impressive 93.52% accuracy only 1.79MB usage. Compared VGG11, 5.89% increase in accuracy, 35.44% faster inference, 99.64% reduction also surpasses networks inference speed. Notably, was successfully deployed Android its performance evaluated real-world settings. The results 84.67% time of 291.04ms per image. validates model's feasibility practical classification, establishing foundation For future research, trained model complete dataset made publicly available.

Язык: Английский

---

U2-NET SEGMENTATION AND MULTI-LABEL CNN CLASSIFICATION OF WHEAT VARIETIES

Konya Journal of Engineering Sciences, Год журнала: 2024, Номер unknown, С. 358 - 372

Опубликована: Фев. 28, 2024

There are many varieties of wheat grown around the world. In addition, they have different physiological states such as vitreous and yellow berry. These reasons make it difficult to classify by experts. this study, a workflow was carried out for both segmentation according its vitreous/yellow berry grain status classification variety. Unlike previous studies, automatic images with U2-NET architecture. Thus, roughness shadows on image minimized. This increased level success in classification. The newly proposed CNN architecture is run two stages. first stage, sorted vitreous-yellow second these separated wheats were grouped multi-label Experimental results showed that accuracy binary 98.71% average 89.5%. study has potential contribute making process more reliable, effective, objective helping

Язык: Английский

---

Nondestructive Identification of Wheat Species using Deep Convolutional Networks with Oversampling Strategies on Near-Infrared Hyperspectral Imagery

Journal of Nondestructive Evaluation, Год журнала: 2024, Номер 44(1)

Опубликована: Дек. 4, 2024

Язык: Английский

---

Classification of Emotion with Audio Analysis

Yüzüncü Yıl Üniversitesi Fen Bilimleri Enstitüsü Dergisi, Год журнала: 2023, Номер 28(2), С. 467 - 481

Опубликована: Март 21, 2023

Classification is an important technique used to distinguish data samples. The aim of this study classify according emotions by extracting audio features. Two male and two female individuals expressed four different as "fun", "angry", "neutral" "sleepy" in the voice data. We “MFCC” a Cepstral feature, “Centroid, Flatness, Skewness, Crest, Flux, Slope, Decrease, Kurtosis, Spread, Entropy, roll off point” Spectral Feature, “Pitch, Harmonic ratio” Periodicity Features sound After, we applied that all classification algorithms located learner toolbox Matlab tried emotion with algorithm provides highest accuracy. Each has twenty-six features inputs one labeled output value. According results, support vector machine provided accuracy performance. Considering performances obtained, reveals it possible sounds using sentimental feature parameters.

Язык: Английский

---