Binary Classification of Mucormycosis Infection Severity using Transfer Learning with VGG16 DOI

C Sugunadevi,

B. Uma Maheswari

Опубликована: Ноя. 24, 2023

Mucormycosis, one of the significant fungal illnesses caused by fungus Mucormycetes, is a rare and fatal disease. It found all over environment. People with weakened immune systems are more prone to get sick rapidly, including those diabetes or COVID-19 history. Mucormycosis can affect eyes quickly spread brain if it enters through nose, sinuses, lungs. Existing research in medical domain shows that deep learning techniques promising solution for assisting practitioners making quick decisions. This study aims use transfer predefined VGG16 model create classifier distinguish between mild severe illness symptoms. The neural network trained, validated, tested loading black dataset. accuracy calculated varying number images. results show proposed gives an 92% 520 input

Язык: Английский

Deep learning-based defect detection in film-coated tablets using a convolutional neural network DOI

K. M. Pathak,

Prapti Kafle, Ajit Vikram

и другие.

International Journal of Pharmaceutics, Год журнала: 2025, Номер unknown, С. 125220 - 125220

Опубликована: Янв. 1, 2025

Язык: Английский

Процитировано

2
Research on species identification of wild grape leaves based on deep learning DOI
Bowen Pan, Chonghuai Liu, Baofeng Su

и другие.

Scientia Horticulturae, Год журнала: 2024, Номер 327, С. 112821 - 112821

Опубликована: Янв. 6, 2024

Язык: Английский

Процитировано

6
Rapid diagnosis of IgA nephropathy using dual branch network combined with FTIR based on cross attention mechanism DOI
Xueqin Zhang, Chenjie Chang, Peng Chao

и другие.

Microchemical Journal, Год журнала: 2025, Номер unknown, С. 113762 - 113762

Опубликована: Апрель 1, 2025

Язык: Английский

Процитировано

0
Fire image detection and classification analysis based on VGG16 image processing model DOI Creative Commons

Fengjun Hou

Applied and Computational Engineering, Год журнала: 2024, Номер 48(1), С. 225 - 231

Опубликована: Март 18, 2024

Fire image classification technology refers to the and recognition of fire images through computer vision in order take timely countermeasures. With development technology, has been widely studied applied. Deep learning techniques have achieved great success field classification, with researchers classifying identifying by using deep models such as convolutional neural networks. This paper introduces research background application experimental results detection analysis based on vgg16 processing model. The model can classify identify well achieve good prediction effect. accuracy training set test is stable at 99%, reach 98%, recall rate F1 score 99%. significance. By images, it improve efficiency monitoring early warning system, detect control fire, reduce loss caused fire. At same time, continuous also provides a broader space for technology.

Язык: Английский

Процитировано

2
Machine learning-based diagnosis and prognosis of IgAN: A systematic review and meta-analysis DOI Creative Commons
Kaiting Zhuang, Wenjuan Wang, Lei Cheng

и другие.

Heliyon, Год журнала: 2024, Номер 10(12), С. e33090 - e33090

Опубликована: Июнь 1, 2024

Plenty of studies have explored the diagnosis and prognosis IgA nephropathy (IgAN) based on machine learning (ML), but accuracy lacks support evidence-based medical evidence. We aim at this problem to guide precision treatment IgAN.

Язык: Английский

Процитировано

2
Advancing Crayfish Disease Detection: A Comparative Study of Deep Learning and Canonical Machine Learning Techniques DOI Creative Commons
Yasin Atilkan,

Berk Kirik,

Koray Açıcı

и другие.

Applied Sciences, Год журнала: 2024, Номер 14(14), С. 6211 - 6211

Опубликована: Июль 17, 2024

This study evaluates the effectiveness of deep learning and canonical machine models for detecting diseases in crayfish from an imbalanced dataset. In this study, measurements such as weight, size, gender healthy diseased individuals were taken, at least five photographs each individual used. Deep outperformed models, but combining both approaches proved most effective. Utilizing ResNet50 model automatic feature extraction subsequent training RF algorithm with these extracted features led to a hybrid model, RF-ResNet50, which achieved highest performance sample detection. result underscores value integrating algorithms models. Additionally, ConvNeXt-T optimized AdamW, performed better than those using SGD, although its disease detection sensitivity was 1.3% lower model. McNemar’s test confirmed statistical significance differences between AdamW. The model’s improved by 3.2% when combined algorithm, demonstrating potential enhancing accuracy. Overall, highlights advantages leveraging techniques early accurate populations, is crucial maintaining ecosystem balance preventing population declines.

Язык: Английский

Процитировано

1
Real-Time Sign Language Detection: Empowering the Disabled Community DOI Creative Commons
Sumit Kumar, Ruchi Rani,

U.V. Chaudhari

и другие.

MethodsX, Год журнала: 2024, Номер 13, С. 102901 - 102901

Опубликована: Авг. 8, 2024

Interaction and communication for normal human beings are easier than a person with disabilities like speaking hearing who may face problems other people. Sign Language helps reduce this gap between disabled person. The prior solutions proposed using several deep learning techniques, such as Convolutional Neural Networks, Support Vector Machines, K-Nearest Neighbors, have either demonstrated low accuracy or not been implemented real-time working systems. This system addresses both issues effectively. work extends the difficulties faced while classifying characters in Indian Language(ISL). It can identify total of 23 hand poses ISL. uses pre-trained VGG16 Convolution Network(CNN) an attention mechanism. model's training Adam optimizer cross-entropy loss function. results demonstrate effectiveness transfer ISL classification, achieving 97.5 % 99.8 plus mechanism.•Enabling quick accurate sign language recognition help trained model mechanism.•The does require any external gloves sensors, which to eliminate need physical sensors simplifying process reduced costs.•Real-time processing makes more helpful people disabilities, making it them communicate humans.

Язык: Английский

Процитировано

1
Automatic Classification of Fungal-Fungal Interactions Using Deep Leaning Models DOI Creative Commons
Marjan Mansourvar, Jonathan Funk, Søren D. Petersen

и другие.

Computational and Structural Biotechnology Journal, Год журнала: 2024, Номер 23, С. 4222 - 4231

Опубликована: Ноя. 14, 2024

Fungi provide valuable solutions for diverse biotechnological applications, such as enzymes in the food industry, bioactive metabolites healthcare, and biocontrol organisms agriculture. Current workflows identifying new fungi often rely on subjective visual observations of strains' performance microbe-microbe interaction studies, making process time-consuming difficult to reproduce. To overcome these challenges, we developed an AI-automated image classification approach using machine learning algorithm based deep neural network. Our method focuses analyzing standardized images 96-well microtiter plates with solid medium fungal-fungal challenge experiments. We used our model categorize outcome interactions between plant pathogen

Язык: Английский

Процитировано

1
Optimizing VGG16 deep learning model with enhanced hunger games search for logo classification DOI Creative Commons
Mohammed Hussain, Thaer Thaher, Mohamed Basel Almourad

и другие.

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

Опубликована: Дек. 30, 2024

Accurate classification of logos is a challenging task in image recognition due to variations logo size, orientation, and background complexity. Deep learning models, such as VGG16, have demonstrated promising results handling tasks. However, their performance highly dependent on optimal hyperparameter settings, whose fine-tuning both labor-intensive time-consuming. Swarm intelligence algorithms been widely adopted solve many nonlinear, multimodal problems succeeded significantly. The Hunger Games Search (HGS) recent swarm algorithm that has shown good across various applications. the standard HGS still faces limitations, restricted population diversity tendency get trapped local optima, which can hinder its effectiveness. In this paper, we propose an optimized deep architecture called EHGS-VGG16 designed based VGG16 model boosted by enhanced (EHGS) for tuning. proposed enhancement involves modified search strategies, incorporating concepts "local best" escaping mechanism" improve exploration capability. To validate our approach, evaluation conducted three folds. First, EHGS evaluated through 30 real-valued benchmark functions from IEEE CEC2014 suite. Second, custom-developed tested Flickr-27 dataset compared against state-of-the-art models ResNet50V2, InceptionV3, DenseNet121, EfficientNetB0, MobileNetV2. Finally, integrated into optimize hyperparameters. experimental show outperformed other counterparts with accuracy 0.956966, precision 0.957137, recall 0.956966. Moreover, integration further improved quality 3%. These findings highlight potential combining evolutionary optimization techniques log

Язык: Английский

Процитировано

1
Comparative Analysis VGG16 Vs MobileNet Performance for Fish Identification DOI Creative Commons
Djarot Hindarto

International Journal Software Engineering and Computer Science (IJSECS), Год журнала: 2023, Номер 3(3), С. 270 - 280

Опубликована: Дек. 10, 2023

This research aims to conduct a comparative evaluation of the efficacy two neural network architectures in field fish identification through utilization supervised learning techniques. The VGG16 and MobileNet, which are prominent deep architectures, has been conducted about their speed, accuracy, efficiency resource utilization. To assess classification performance both we employed dataset encompassing diverse categories. findings indicated that model demonstrated superior accuracy classification, albeit due increased computational time On contrary, MobileNet exhibits enhanced speed efficiency, at marginal cost its accuracy. this study have potential inform selection models for recognition scenarios, considering specific requirements task, such as prioritizing or efficiency. mentioned above offer significant insights can be utilized advancement Artificial Intelligence (AI)-based applications within domains fisheries management environmental monitoring. These specifically necessitate precise effective capabilities. comparison indicate achieved by was 0.99, whereas also attained an 0.99.

Язык: Английский

Процитировано

2