Convolutional Neural Network-based Architecture for Detecting Face Mask in Crowded Areas DOI
Jad Abou Chaaya,

Batoul Zaraket,

Hassan Harb

et al.

Published: July 2, 2023

After the invasion of Covid-19 virus, governments started containing spread virus by forcing people to wear face masks in public places. Therefore, automatic mask detection has become very important limit spread. Unfortunately, existing methods present limited performance accurately detecting crowded areas due significant number faces per scene. In order tackle this challenge, we propose a two-stage neural network-based architecture that can detect environments. Several simulations have been conducted investigate efficiency proposed and results show high accuracy reach up 96.5%.

Language: Английский

A Review on Deep Learning Techniques for IoT Data DOI Open Access
Kuruva Lakshmanna, Rajesh Kaluri,

Nagaraja Gundluru

et al.

Electronics, Journal Year: 2022, Volume and Issue: 11(10), P. 1604 - 1604

Published: May 18, 2022

Continuous growth in software, hardware and internet technology has enabled the of internet-based sensor tools that provide physical world observations data measurement. The Internet Things(IoT) is made up billions smart things communicate, extending boundaries virtual entities further. These intelligent produce or collect massive daily with a broad range applications fields. Analytics on these huge critical tool for discovering new knowledge, foreseeing future knowledge making control decisions make IoT worthy business paradigm enhancing technology. Deep learning been used variety projects involving mobile apps, encouraging early results. With its data-driven, anomaly-based methodology capacity to detect developing, unexpected attacks, deep may deliver cutting-edge solutions intrusion detection. In this paper, increased amount information gathered produced being further develop intelligence application capabilities through Learning (DL) techniques. Many researchers have attracted various fields IoT, both DL techniques approached. Different studies suggested as feasible solution manage by because it was intended handle large amounts, requiring almost real-time processing. We start discussing introduction generation also discuss approaches their procedures. surveyed summarized major reporting efforts region datasets. features, challenges uses empower applications, which are discussed promising field, can motivate inspire developments.

Language: Английский

Citations

105
Face Mask Detection in Smart Cities Using Deep and Transfer Learning: Lessons Learned from the COVID-19 Pandemic DOI Creative Commons
Yassine Himeur, Somaya Al‐Maadeed, Iraklis Varlamis

et al.

Systems, Journal Year: 2023, Volume and Issue: 11(2), P. 107 - 107

Published: Feb. 17, 2023

After different consecutive waves, the pandemic phase of Coronavirus disease 2019 does not look to be ending soon for most countries across world. To slow spread COVID-19 virus, several measures have been adopted since start outbreak, including wearing face masks and maintaining social distancing. Ensuring safety in public areas smart cities requires modern technologies, such as deep learning transfer learning, computer vision automatic mask detection accurate control whether people wear correctly. This paper reviews progress research, emphasizing techniques. Existing datasets are first described discussed before presenting recent advances all related processing stages using a well-defined taxonomy, nature object detectors Convolutional Neural Network architectures employed their complexity, techniques that applied so far. Moving on, benchmarking results summarized, discussions regarding limitations methodologies provided. Last but least, future research directions detail.

Language: Английский

Citations

43
A lightweight network for traffic sign recognition based on multi-scale feature and attention mechanism DOI Creative Commons
Wei Wei, Lili Zhang,

Kang Yang

et al.

Heliyon, Journal Year: 2024, Volume and Issue: 10(4), P. e26182 - e26182

Published: Feb. 1, 2024

Traffic sign recognition is an important part of intelligent transportation system. It uses computer vision and traffic technology to detect recognize signs on the road automatically. In this paper, we propose a lightweight model for based convolutional neural networks called ConvNeSe. Firstly, feature extraction module constructed using Depthwise Separable Convolution Inverted Residuals structures. The extracts multi-scale features with strong representation ability by optimizing structure fusing features. Then, introduces Squeeze Excitation Block (SE Block) improve attention features, which can capture key information images. Finally, accuracy in German Sign Recognition Benchmark Database (GTSRB) 99.85%. At same time, has good robustness according results ablation experiments.

Language: Английский

Citations

5
Intelligent Risk Prediction System in IoT-Based Supply Chain Management in Logistics Sector DOI Open Access
Ahmed Alzahrani, Muhammad Zubair Asghar

Electronics, Journal Year: 2023, Volume and Issue: 12(13), P. 2760 - 2760

Published: June 21, 2023

The Internet of Things (IoT) has resulted in substantial advances the logistics sector, particularly storage management, communication systems, service quality, and supply chain management. goal this study is to create an intelligent (SC) management system that provides decision support SC managers order achieve effective (IOT)-based logistics. Current research on predicting risks shipping operations sector during natural disasters produced a variety unexpected findings utilizing machine learning (ML) algorithms traditional feature-encoding approaches. This prompted concerns regarding research’s validity. These previous attempts, like many others before them, used deep neural models gain features without requiring user maintain track all sequence information. paper offers hybrid (DL) approach, convolutional network (CNN) + bidirectional gating recurrent unit (BiGRU), lessen impact by addressing question, “Can goods be shipped from source location destination?”. suggested DL methodology divided into four stages: data collection, de-noising or pre-processing, feature extraction, prediction. When compared baseline work, proposed CNN BiGRU achieved accuracy up 94%.

Language: Английский

Citations

12
Fast detection of face masks in public places using QARepVGG-YOLOv7 DOI

Chuying Guan,

Jiaxuan Jiang, Zhong Wang

et al.

Journal of Real-Time Image Processing, Journal Year: 2024, Volume and Issue: 21(3)

Published: May 19, 2024

Language: Английский

Citations

4
Ensemble of deep transfer learning models for real-time automatic detection of face mask DOI Open Access
Rubul Kumar Bania

Multimedia Tools and Applications, Journal Year: 2023, Volume and Issue: 82(16), P. 25131 - 25153

Published: Feb. 1, 2023

Language: Английский

Citations

10
A Chinese medical named entity recognition method considering length diversity of entities DOI
Hongyu Zhang,

Long Lyu,

W. L. Chang

et al.

Engineering Applications of Artificial Intelligence, Journal Year: 2025, Volume and Issue: 150, P. 110649 - 110649

Published: March 26, 2025

Language: Английский

Citations

0
Mushroom species classification and implementation based on improved MobileNetV3 DOI

Jun Peng,

Song Li, Hui Li

et al.

Journal of Food Science, Journal Year: 2025, Volume and Issue: 90(4)

Published: April 1, 2025

Current methods for mushroom species classification face limitations in generalization ability and lack exploration of model deployment. To address these issues, this study systematically compares five models, including Transformer common convolutional neural networks. MobileNetV3 was chosen as the study, combining transfer learning with adaptive hybrid optimizer (AHO) dynamic cyclic rate strategies proposed research. The AHO merges Adam's fast convergence stochastic gradient descent's stable fine-tuning. It adjusts dynamically based on training progress, enabling quick early precise adjustments later. optimized trained, validated, deployed a dataset constructed which includes 3633 images covering three types mushrooms. achieved validation accuracy 98.13% an average test 97.98%, smallest standard deviation loss fluctuation (0.0343), confirming model's stability. Notably, due to slightly larger number Matsutake subset (1412 images) compared other two categories (1148 1073 images), (99.28%) higher than that Red (96.97%) Beefsteak (97.69%), highlighting minor limitation. However, recall F1 scores each class are balanced, suggesting exhibits robust performance addressing interclass similarities, corroborated by t-SNE visualization Grad-CAM analysis. Additionally, confirmed feasibility practical application through deployment PC, Android, embedded platforms, providing guiding solution laboratory research, wild picking, automated sorting. PRACTICAL APPLICATION: This provides AI lightweight network identifying different species. can be widely applied scenarios such harvesting, sorting, helping farmers, consumers, researchers easily accurately identify varieties, thereby contributing development industry.

Language: Английский

Citations

0
SERS with Flexible β-CD@AuNP/PTFE Substrates for In Situ Detection and Identification of PAH Residues on Fruit and Vegetable Surfaces Combined with Lightweight Network DOI Creative Commons
Mengqing Qiu,

Le Tang,

Jinghong Wang

et al.

Foods, Journal Year: 2023, Volume and Issue: 12(16), P. 3096 - 3096

Published: Aug. 17, 2023

The detection of polycyclic aromatic hydrocarbons (PAHs) on fruit and vegetable surfaces is important for protecting human health ensuring food safety. In this study, a method the in situ identification PAH residues was developed using surface-enhanced Raman spectroscopy (SERS) based flexible substrate lightweight deep learning network. SERS fabricated by assembling β-cyclodextrin-modified gold nanoparticles (β-CD@AuNPs) polytetrafluoroethylene (PTFE) film coated with perfluorinated liquid (β-CD@AuNP/PTFE). concentrations benzo(a)pyrene (BaP), naphthalene (Nap), pyrene (Pyr) could be detected at 0.25, 0.5, 0.25 μg/cm2, respectively, all relative standard deviations (RSD) were less than 10%, indicating that β-CD@AuNP/PTFE exhibited high sensitivity stability. network then used to construct classification model identifying various residues. ShuffleNet obtained best results accuracies 100%, 96.61%, 97.63% training, validation, prediction datasets, respectively. proposed realised vegetables simplicity, celerity, sensitivity, demonstrating great potential rapid, nondestructive analysis surface contaminant food-safety field.

Language: Английский

Citations

7
IoT Solutions and AI-Based Frameworks for Masked-Face and Face Recognition to Fight the COVID-19 Pandemic DOI Creative Commons
Jamal Al-Nabulsi, Nidal Turab, Hamza Abu Owida

et al.

Sensors, Journal Year: 2023, Volume and Issue: 23(16), P. 7193 - 7193

Published: Aug. 15, 2023

A global health emergency resulted from the COVID-19 epidemic. Image recognition techniques are a useful tool for limiting spread of pandemic; indeed, World Health Organization (WHO) recommends use face masks in public places as form protection against contagion. Hence, innovative systems and algorithms were deployed to rapidly screen large number people with faces covered by masks. In this article, we analyze current state research future directions masked-face recognition. First, paper discusses importance applications facial mask recognition, introducing main approaches. Afterward, review recent frameworks based on Convolution Neural Networks, deep learning, machine MobilNet techniques. detail, critically discuss scientific works which employ learning (ML) tools promptly recognizing masked faces. Also, Internet Things (IoT)-based sensors, implementing ML DL algorithms, described keep track persons donning notify proper authorities. challenges open issues that should be solved studies discussed. Finally, comparative analysis discussion reported, providing insights outlining next generation systems.

Language: Английский

Citations

6