Consecutive multiscale feature learning-based image classification model DOI Creative Commons
Bekhzod Olimov, Barathi Subramanian,

Rakhmonov Akhrorjon Akhmadjon Ugli

и другие.

Scientific Reports, Год журнала: 2023, Номер 13(1)

Опубликована: Март 3, 2023

Abstract Extracting useful features at multiple scales is a crucial task in computer vision. The emergence of deep-learning techniques and the advancements convolutional neural networks (CNNs) have facilitated effective multiscale feature extraction that results stable performance improvements numerous real-life applications. However, currently available state-of-the-art methods primarily rely on parallel approach, despite exhibiting competitive accuracy, models lead to poor efficient computation low generalization small-scale images. Moreover, lightweight cannot appropriately learn features, this causes underfitting when training with images or datasets limited number samples. To address these problems, we propose novel image classification system based elaborate data preprocessing steps carefully designed CNN model architecture. Specifically, present consecutive feature-learning network (CMSFL-Net) employs approach usage various maps different receptive fields achieve faster training/inference higher accuracy. In conducted experiments using six datasets, including small-scale, large-scale, data, CMSFL-Net exhibits an accuracy comparable those existing networks. proposed outperforms them terms efficiency speed achieves best accuracy-efficiency trade-off.

Язык: Английский

Spatio-Temporal Graph Neural Networks for Predictive Learning in Urban Computing: A Survey DOI
Guangyin Jin, Yuxuan Liang, Yuchen Fang

и другие.

IEEE Transactions on Knowledge and Data Engineering, Год журнала: 2023, Номер 36(10), С. 5388 - 5408

Опубликована: Ноя. 23, 2023

With recent advances in sensing technologies, a myriad of spatio-temporal data has been generated and recorded smart cities. Forecasting the evolution patterns is an important yet demanding aspect urban computing, which can enhance intelligent management decisions various fields, including transportation, environment, climate, public safety, healthcare, others. Traditional statistical deep learning methods struggle to capture complex correlations data. To this end, Spatio-Temporal Graph Neural Networks (STGNN) have proposed, achieving great promise years. STGNNs enable extraction dependencies by integrating graph neural networks (GNNs) temporal methods. In manuscript, we provide comprehensive survey on progress STGNN technologies for predictive computing. Firstly, brief introduction construction prevalent deep-learning architectures used STGNNs. We then sort out primary application domains specific tasks based existing literature. Afterward, scrutinize design their combination with some advanced Finally, conclude limitations research suggest potential directions future work.

Язык: Английский

Процитировано

143
The effect of choosing optimizer algorithms to improve computer vision tasks: a comparative study DOI Creative Commons

Esraa Hassan,

Mahmoud Y. Shams, Noha A. Hikal

и другие.

Multimedia Tools and Applications, Год журнала: 2022, Номер 82(11), С. 16591 - 16633

Опубликована: Сен. 28, 2022

Optimization algorithms are used to improve model accuracy. The optimization process undergoes multiple cycles until convergence. A variety of strategies have been developed overcome the obstacles involved in learning process. Some these considered this study learn more about their complexities. It is crucial analyse and summarise techniques methodically from a machine standpoint since can provide direction for future work both optimization. approaches under consideration include Stochastic Gradient Descent (SGD), with Momentum, Rung Kutta, Adaptive Learning Rate, Root Mean Square Propagation, Moment Estimation, Deep Ensembles, Feedback Alignment, Direct Adfactor, AMSGrad, Gravity. prove ability each optimizer applied models. Firstly, tests on skin cancer using ISIC standard dataset detection were three common optimizers (Adaptive Moment, SGD, Propagation) explore effect images. optimal training results analysis indicate that performance values enhanced Adam optimizer, which achieved 97.30% second COVIDx CT images, 99.07% accuracy based optimizer. result indicated utilisation such as SGD improved training, testing, validation stages.

Язык: Английский

Процитировано

115
Mobile robot localization: Current challenges and future prospective DOI
Inam Ullah, Deepak Adhikari, Habib Ullah Khan

и другие.

Computer Science Review, Год журнала: 2024, Номер 53, С. 100651 - 100651

Опубликована: Июль 5, 2024

Язык: Английский

Процитировано

30
A systematic survey of air quality prediction based on deep learning DOI Creative Commons
Zhen Zhang, Shiqing Zhang, Caimei Chen

и другие.

Alexandria Engineering Journal, Год журнала: 2024, Номер 93, С. 128 - 141

Опубликована: Март 16, 2024

The impact of air pollution on public health is substantial, and accurate long-term predictions quality are crucial for early warning systems to address this issue. Air prediction has drawn significant attention, bridging environmental science, statistics, computer science. This paper presents a comprehensive review the current research status advances in methods. Deep learning, novel machine learning approach, demonstrated remarkable proficiency identifying complex, nonlinear patterns data, yet its application still relatively nascent. also conducts systematic analysis summarizes how cutting-edge deep models applied prediction. Initially, historical evolution methods datasets presented. followed by an examination conventional techniques. A thorough comparative progress made with both traditional learning-based provided. particularly focuses three aspects: temporal modeling, spatiotemporal attention mechanisms. Finally, emerging trends field identified discussed.

Язык: Английский

Процитировано

17
Double fuzzy clustering-driven context neural network for intrusion detection in cloud computing DOI Creative Commons

S. Anu Velavan,

C. Sureshkumar

Wireless Networks, Год журнала: 2025, Номер unknown

Опубликована: Янв. 20, 2025

Язык: Английский

Процитировано

4
Unsupervised graph-level representation learning with hierarchical contrasts DOI
Wei Ju, Yiyang Gu, Xiao Luo

и другие.

Neural Networks, Год журнала: 2022, Номер 158, С. 359 - 368

Опубликована: Ноя. 26, 2022

Язык: Английский

Процитировано

46
Cancer detection and segmentation using machine learning and deep learning techniques: a review DOI
Hari Mohan

Multimedia Tools and Applications, Год журнала: 2023, Номер 83(9), С. 27001 - 27035

Опубликована: Авг. 22, 2023

Язык: Английский

Процитировано

33
Traffic forecasting with graph spatial–temporal position recurrent network DOI
Yibi Chen, Kenli Li, Chai Kiat Yeo

и другие.

Neural Networks, Год журнала: 2023, Номер 162, С. 340 - 349

Опубликована: Март 15, 2023

Язык: Английский

Процитировано

30
A review on action recognition for accident detection in smart city transportation systems DOI Creative Commons
Victor Adewopo, Zag ElSayed,

Zag ElSayed

и другие.

Journal of Electrical Systems and Information Technology, Год журнала: 2023, Номер 10(1)

Опубликована: Ноя. 27, 2023

Abstract Accident detection and public traffic safety is a crucial aspect of safe better community. Monitoring flow in smart cities using different surveillance cameras plays role recognizing accidents alerting first responders. In computer vision tasks, utilizing action recognition (AR) has contributed to high-precision video surveillance, medical imaging, digital signal processing applications. This paper presents an intensive review focusing on accident autonomous transportation systems for city. focused AR that use diverse sources video, such as static intersections, highway monitoring cameras, drone dash-cams. Through this review, we identified the primary techniques, taxonomies, algorithms used detection. We also examined datasets utilized identifying features datasets. provides potential research direction develop integrate cars by emergency personnel law enforcement event road minimize human error reporting provide spontaneous response victims.

Язык: Английский

Процитировано

28
Recognition of human activity using GRU deep learning algorithm DOI Creative Commons
Saeed Mohsen

Multimedia Tools and Applications, Год журнала: 2023, Номер 82(30), С. 47733 - 47749

Опубликована: Май 11, 2023

Abstract Human activity recognition (HAR) is a challenging issue in several fields, such as medical diagnosis. Recent advances the accuracy of deep learning have contributed to solving HAR issues. Thus, it necessary implement algorithms that high performance and greater accuracy. In this paper, gated recurrent unit (GRU) algorithm proposed classify human activities. This applied Wireless Sensor Data Mining (WISDM) dataset gathered from many individuals with six classes various activities – walking, sitting, downstairs, jogging, standing, upstairs. The tested trained via hyper-parameter tuning method TensorFlow framework achieve Experiments are conducted evaluate GRU using receiver operating characteristic (ROC) curves confusion matrices. results demonstrate provides achieves testing 97.08%. rate loss for 0.221, while precision, sensitivity, F1-score 97.11%, 97.09%, 97.10%, respectively. Experimentally, area under ROC (AUC S ) 100%.

Язык: Английский

Процитировано

27