A Comprehensive Review of Advanced Deep Learning Approaches for Food Freshness Detection

Food Engineering Reviews, Год журнала: 2024, Номер unknown

Опубликована: Дек. 6, 2024

Язык: Английский

---

Machine Learning-Based Classification of Infected Date Palm Leaves Caused by Dubas Insects: A Comparative Analysis of Feature Extraction Methods and Classification Algorithms

2022 Innovations in Intelligent Systems and Applications Conference (ASYU), Год журнала: 2023, Номер unknown

Опубликована: Окт. 11, 2023

This study investigates the utilization of machine learning techniques for effectively classifying infected date palm leaves caused by Dubas insects. Three distinct feature extraction methods, namely Inceptionv3, SqueezeNet, and VGG16, are combined with five diverse algorithms: K-Nearest Neighbors (KNN), Neural Network (ANN), Random Forest (RF), Artificial Support Vector Machine (SVM), Logistic Regression (LR). The dataset comprises a collection images depicting leaves, performance evaluation metrics, including accuracy, recall, precision, F1 score, computed each algorithm. results unveil varied levels accuracy among methods algorithms. Specifically, Inceptionv3 achieved an 80.4% KNN, while SqueezeNet attained 75.3% VGG16 obtained 76.6% accuracy. For SVM, scores were 72.9%, 66%, 62.4%, respectively. ANN demonstrated promising 83.8%, 80%, 80.1% Lastly, LR yielded 83%, 76.2%, 80% These findings offer useful information about how various perform in thereby facilitating development effective pest management strategies plantations.

Язык: Английский

---

Optical evaluation of internal meat quality deterioration in a tuna fillet based on second-harmonic generation anisotropy measurement

Journal of Food Engineering, Год журнала: 2024, Номер unknown, С. 112422 - 112422

Опубликована: Ноя. 1, 2024

Язык: Английский

---

Detection of Fungal Infections from Microscopic Fungal Images Using Deep Learning Techniques

Proceedings of the International Conference on Advanced Technologies, Год журнала: 2023, Номер unknown

Опубликована: Авг. 19, 2023

Fungal infections, due to their diverse manifestations and varying characteristics, present significant challenges in medical diagnosis. This study delves into applying deep-learning techniques for detecting fungal infections from microscopic images. By harnessing the power of Convolutional Neural Networks (CNNs), we propose an approach that employs transfer learning accurately classify different species. The dataset comprises images various types, enhance model performance, utilize data augmentation techniques. Furthermore, aim boost performance by fine-tuning model's layers. Initially starting at 84.38% accuracy, our experimental results progressively reached high values 95.35% 97.19%. These underscore effectiveness deep precisely identifying classifying infections. success holds promising potential aid professionals timely accurate diagnoses. findings presented this contribute ongoing research image analysis drive advancements field automated disease detection.

Язык: Английский

---

Enhancing Quality Control: Defect State Classification of Taralli Biscuits with MobileNet-v2 and DenseNet-201

Опубликована: Сен. 7, 2023

Industrial production and packaging face significant challenges, such as product damage, color changes, the presence of foreign bodies. These issues greatly impact quality, profitability, marketability, leading to increased consumer complaints. To address these concerns, this study presents a novel method for classifying Taralli biscuits using image processing techniques. The research encompasses dataset 4,900 images, featuring four types defects: no defect, defect-shape, defect-object, defect-color. Leveraging advanced deep learning architectures, including MobileNet-v2 DenseNet-201, classification process achieves impressive accuracy rates 98.71% 99.39% respectively. By automating detection biscuit proposed enhances quality control inspection processes within food industry. combination state-of-the-art techniques in provides an effective solution automatically detecting categorizing defects.

Язык: Английский

---

Detection of Defects in Soybean Seeds by Extracting Deep Features with SqueezeNet

Опубликована: Сен. 7, 2023

A vital part of ensuring the quality soybean products is detecting defects. The current study presents a five-category classification seed defects: broken, immature, intact, skin-damaged, and spotted seeds. Our goal to improve overall through accurately identifying categorizing Computer vision techniques machine learning algorithms are combined comprehensively achieve this goal. To begin with, images analyzed using SqueezeNet model, deep-learning architecture known for its efficiency in image analysis. Features extracted from soybeans indicate types defects they present their key visual characteristics. Then, we applied three widely used algorithms, including Artificial Neural Network (ANN), Logistic Regression (LR), Random Forest (RF), classify images. labeled dataset with train fine-tune each algorithm. An appropriate evaluation metric assessing It provides valuable insights into application improving product by detection

Язык: Английский

---

Classification of Hazelnut Species with Pre-Trained Deep Learning Models

Intelligent Methods in Engineering Sciences, Год журнала: 2023, Номер unknown

Опубликована: Июнь 1, 2023

A form of shelled nut in the Betulaceae family is hazelnut. The majority it grown Türkiye internationally. It grows provinces Türkiye's Black Sea region, which a significant global production hub. Hazelnuts can be eaten variety ways and are great source protein, fat, fiber, vitamins, minerals. There numerous applications for hazelnuts food business. This study uses pre-trained networks to categorize eight most popular hazelnut kinds farmed Türkiye. In this study, locally named varieties were examined. An automated computer vision system was used capture images different kinds. Our dataset includes total 2722 images, consisting 155 palaz, 340 yagli, 399 deve disi, 236 tombul, damat, 354 cakildak, 437 kara findik, 402 sivri hazelnuts. Using transfer learning, DenseNet121 InceptionV3 models convolutional neural employed these images. split into training testing portions, respectively. With DenseNet121, respectively, research revealed classification accuracy 96.99% 96.18%.

Язык: Английский

---

Comparative Analysis of YOLOv8 Models in Skipjack Fish Quality Assessment System

Опубликована: Дек. 13, 2023

Skipjack Tuna (Katsuwonus pelamis) holds significant economic importance in Indonesia, and ensuring its freshness is paramount for consumer satisfaction. Traditional methods of assessing fish freshness, such as sensory evaluation, face limitations accuracy. Deep learning, particularly Convolutional Neural Networks (CNN), presents a non-contact solution visually identifying freshness. This study aims to assess compare the performance various YOLOv8 models detecting Tuna. Five are considered: YOLOv8n, YOLOv8s, YOLOv8m, YOLOv8l, YOLOv8x, each distinguished by size complexity. Evaluation metrics encompass accuracy, precision, recall, specificity, F1-score, providing comprehensive analysis identify most suitable model The employs structured methodology involving dataset preparation, architecture, image pre-processing, implementation, classification. relies on confusion matrix metrics. Results reveal exemplary across all Training accuracy surpasses 98%, validation exceeds testing above 99%. Precision, F1-score values both classes consistently exhibit high levels, indicating effective discrimination between fresh non-fresh fish. These findings emphasize robust detection.

Язык: Английский

---

Fish Freshness Detection Through Artificial Intelligence Approaches: A Comprehensive Study

Turkish Journal of Agriculture - Food Science and Technology, Год журнала: 2024, Номер 12(2), С. 290 - 295

Опубликована: Фев. 26, 2024

Fish is regarded as an important protein source in human nutrition due to its high concentration of omega-3 fatty acids In traditional global cuisine, fish holds a prominent position, with seafood restaurants, markets, and eateries serving popular venues for consumption. However, it imperative preserve freshness improper storage can lead rapid spoilage, posing risks potential foodborne illnesses. To address this concern, artificial intelligence techniques have been utilized evaluate freshness, introducing deep learning machine approach. Leveraging dataset 4476 images, study conducted feature extraction using three transfer models (MobileNetV2, Xception, VGG16) applied four algorithms (SVM, LR, ANN, RF) classification. The synergy Xception MobileNetV2 SVM LR achieved 100% success rate, highlighting the effectiveness preventing illness preserving taste quality products, especially mass production facilities.

Язык: Английский

---

Machine Learning-Based Classification of Mulberry Leaf Diseases

Proceedings of international conference on intelligent systems and new applications., Год журнала: 2024, Номер unknown

Опубликована: Апрель 28, 2024

This research examines the potential of machine learning methods in classification Mulberry leaf diseases. By applying SqueezeNet's deep feature extraction, study aimed to identify disease patterns efficiently. The dataset used consisted ten distinct classes diseases, which was divided into an 80% training set and a 20% testing set. Support Vector Machine (SVM) supervised algorithm classify model achieved accuracy 77.5%. results demonstrate effectiveness approaches aiding detection management can contribute advancements agricultural monitoring mitigation strategies.

Язык: Английский

---