Lecture notes in networks and systems, Год журнала: 2024, Номер unknown, С. 375 - 393
Опубликована: Янв. 1, 2024
Язык: Английский
Electronics, Год журнала: 2023, Номер 12(24), С. 4970 - 4970
Опубликована: Дек. 12, 2023
Intelligent traffic systems represent one of the crucial domains in today’s world, aiming to enhance management efficiency and road safety. However, current intelligent still face various challenges, particularly realm target detection. These challenges include adapting complex scenarios lack precise detection for multiple objects. To address these issues, we propose an innovative approach known as YOLOv8-SnakeVision. This method introduces Dynamic Snake Convolution, Context Aggregation Attention Mechanisms, Wise-IoU strategy within YOLOv8 framework performance. Convolution assists accurately capturing object shapes features, especially cases occlusion or overlap. The Mechanisms allow model better focus on critical image regions effectively integrate information, thus improving its ability recognize obscured targets, small objects, patterns. combines dynamic non-monotonic focusing mechanisms, more precisely regress bounding boxes, low-quality examples. We validate our BDD100K NEXET datasets. Experimental results demonstrate that YOLOv8-SnakeVision excels scenarios. It not only enhances but also strengthens targets. provides robust support development holds promise achieving further breakthroughs future applications.
Язык: Английский
Процитировано
27
Alexandria Engineering Journal, Год журнала: 2024, Номер 106, С. 277 - 287
Опубликована: Июль 13, 2024
Язык: Английский
Процитировано
14
Опубликована: Апрель 7, 2024
With the rapid development of autonomous driving technology, demand for real-time and efficient object detection systems has been increasing to ensure vehicles can accurately perceive respond surrounding environment. Traditional models often suffer from issues such as large parameter sizes high computational resource consumption, limiting their applicability on edge devices. To address this issue, we propose a lightweight model called YOLOv8-Lite, based YOLOv8 framework, improved through various enhancements including adoption FastDet structure, TFPN pyramid CBAM attention mechanism. These improvements effectively enhance performance efficiency model. Experimental results demonstrate significant our NEXET KITTI datasets. Compared traditional methods, exhibits higher accuracy robustness in tasks, better addressing challenges fields driving, contributing advancement intelligent transportation systems.
Язык: Английский
Процитировано
9
PeerJ Computer Science, Год журнала: 2025, Номер 11, С. e2519 - e2519
Опубликована: Янв. 21, 2025
Road damage often includes potholes, cracks, lane degradation, and surface shading. Potholes are a common problem in pavements. Detecting them is crucial for maintaining infrastructure ensuring public safety. A thorough assessment of pavement conditions required before planning any preventive repairs. Herein, we report the use transfer learning deep (DL) models to preprocess digital images pavements better pothole detection. Fourteen were evaluated, including MobileNet, MobileNetV2, NASNetMobile, DenseNet121, DenseNet169, InceptionV3, DenseNet201, ResNet152V2, EfficientNetB0, InceptionResNetV2, Xception, EfficientNetV2M. The study introduces modified MobileNetV2 (MMNV2) model designed fast efficient feature extraction. MMNV2 exhibits improved classification, detection, prediction accuracy by adding five-layer pre-trained network framework. It combines learning, neural networks (DNN), which resulted performance compared other models. was tested using dataset 5,000 images. rate 0.001 used optimize model. classified into ‘normal’ or ‘pothole’ categories with 99.95% accuracy. also achieved 100% recall, 99.90% precision, F1-score, 0.05% error rate. uses fewer parameters while delivering results. offers promising solution real-world applications detection assessment.
Язык: Английский
Процитировано
1
Engineering Applications of Artificial Intelligence, Год журнала: 2025, Номер 148, С. 110287 - 110287
Опубликована: Март 9, 2025
Язык: Английский
Процитировано
1
IEEE Open Journal of Intelligent Transportation Systems, Год журнала: 2024, Номер 5, С. 509 - 519
Опубликована: Янв. 1, 2024
Язык: Английский
Процитировано
4
Sustainable Cities and Society, Год журнала: 2024, Номер 116, С. 105882 - 105882
Опубликована: Окт. 10, 2024
Язык: Английский
Процитировано
4
IGI Global eBooks, Год журнала: 2025, Номер unknown, С. 83 - 114
Опубликована: Фев. 21, 2025
There is an effective need to manage the existing traffic systems due rapid increase in production and usage of vehicles. Traffic congestion, crashes delays are some challenges being faced today. Neural networks have emerged as a powerful solution tackle dynamic complex nature systems. This chapter, “Anomaly Detection Systems,” discusses application neural identifying anomalies by highlighting its importance enhancing safety, efficiency, overall management. As urban areas continue grow prevalence such accidents, unexpected patterns poses significant for transportation authorities. The chapter emphasizes role machine learning (ML) deep (DL) techniques highly focuses on networks, within data. also explores how deal with Management System (TMS) make it intelligent.
Язык: Английский
Процитировано
0
Knowledge and Information Systems, Год журнала: 2025, Номер unknown
Опубликована: Май 20, 2025
Язык: Английский
Процитировано
0
Опубликована: Июнь 25, 2024
Язык: Английский
Процитировано
1