Jul 2026· Italian National Conference on Sensors· Vol 26, pp. 4299· 0 citations· 46 references
Medicine
TL;DR
This study proposes an optimized YOLOv8-based detection paradigm that decouples multi-scale feature enhancement from dynamic focused bounding box regression, and maintains a real-time inference speed of 86 FPS on an NVIDIA RTX 3090 GPU, far exceeding the 30 FPS threshold for real-time traffic monitoring.
Abstract
Conventional object detection models suffer from significant performance degradation in dense urban traffic scenarios. To address these critical limitations and enable accurate real-time road congestion decision making, this study proposes an optimized YOLOv8-based detection paradigm that decouples multi-scale feature enhancement from dynamic focused bounding box regression. Specifically, a multi-scale feature enhancement (MFE) module is designed to extract high-resolution shallow features directly from the P2 layer of the YOLOv8 backbone. Then, a convolutional block attention module (CBAM) is embedded into the feature fusion neck to adaptively filter complex urban background noise and recalibrate channel–spatial feature responses for vehicle target saliency. Furthermore, the standard CIoU loss is replaced with the Wise-IoU (WIoU) dynamic focusing loss function, which suppresses gradient interference from low-quality, occluded samples and stabilizes bounding box regression for dense vehicle targets. The high-precision vehicle detection outputs are fed into a quantitative congestion index (CI) model, which fuses vehicle density and average speed to realize real-time congestion-level classification. Extensive experiments on the public UAVDT benchmark dataset demonstrate that the proposed model achieves an mAP@0.5 of 83.1% (3.8 percentage points higher than the YOLOv8 baseline), an mAP_S (small target) of 23.2% (a 4.3 percentage point improvement), and a real-time congestion decision accuracy of 83.8%. Ablation studies verify the independent and synergistic effectiveness of the MFE, CBAM, and WIoU modules, with the MFE module making the greatest contribution to small-target detection performance (+1.7% mAP@0.5). The proposed model maintains a real-time inference speed of 86 FPS (frames per second) on an NVIDIA RTX 3090 GPU, far exceeding the 30 FPS threshold for real-time traffic monitoring.
To address the problems of weak small target features, target occlusion, severe weather interference, and bounding box regression bias in vehicle detection under complex traffic scenarios, a multi-scale feature fusion vehicle detection method is proposed. An improved YOLOv11 model is used for vehicle target detection. First, data augmentation techniques are used to expand the sample size for complex weather and lighting distortion scenarios to improve the model's environmental adaptability. Second, an efficient multi-scale attention mechanism (EMA) is embedded in the backbone network to enhance target feature extraction and suppress background redundancy. Third, a neck network using PSConv replaces the traditional downsampling convolution to optimize small target feature preservation and multiscale receptive field configuration (MSRF). Finally, an Inner-EIoU loss function is introduced during the training phase to improve bounding box regression accuracy and training stability. Experimental results show that on the UA-DETRAC benchmark dataset and the self-built complex traffic scene dataset, the mAP50 and mAP50-90 values of the vehicle detection model in complex scenes are 64.8% and 46.5%, respectively, which are improved by 4.5% and 4.2% by YOLOv11. From the perspective of computational efficiency, the improved model has a GFLOPs of 6.60, which is lower than YOLOv11's 6.90, and has better detection performance.
Real-time vehicle detection in edge-based intelligent transportation systems is essential for enhancing traffic efficiency and safety. However, vehicle detection in complex urban environments faces challenges (cluttered backgrounds, significant scale variations among targets), necessitating lightweight models that achieve both high accuracy and low computational cost for resource-constrained edge devices. To address these challenges and tackle key engineering problems in traffic scenarios, we proposed YOLOv8n-MMSP, a lightweight real-time vehicle detection algorithm optimized for edge deployment. To mitigate fine-grained feature loss from downsampling, the C2f-StarBlocks module was integrated to enhance multiscale feature fusion, beneficial for long-range vehicle perception in traffic monitoring systems. To reduce the distant small-object miss rate, a P2 small-object detection layer was introduced to strengthen shallow feature representation, improving detection reliability. To handle significant scale variations and complex background interference, the Multi-Scale Convolutional Attention mechanism was incorporated, enabling robust perception under challenging conditions (nighttime and adverse weather). To improve bounding box regression accuracy in densely occluded scenarios, the Multi-Precision Distance Intersection over Union loss function was adopted, facilitating more precise localization in congested traffic environments. Furthermore, to meet the computational constraints of edge devices, an L1 regularization–based channel pruning strategy was applied to batch normalization layers, reducing model complexity while maintaining detection performance. Results demonstrated 81.2% mAP and 61.4% mAP50-95, consistent improvements over the baseline model, while reducing model size, parameter count, and computational load to 61.89%, 59.12%, and 58.40% of the original, respectively. Edge-device tests further demonstrate a real-time inference speed of 73.6 fps.
Chang-Qin Huang, Lixing Zhang, Wanting Duan et al.· Transportation Research Reco...· 0 citations
A YOLO11n-based traffic light detection algorithm, named YOLO11n-PRE, which replaces the original C3k2 module in the backbone network with the C3k2-RCB module, which enhances deep feature extraction capability while maintaining lightweight via efficient residual connection and feature recalibration mechanism.
Ce Zheng, Xiao-Qiang Yu, Wenguo Li· International Conference on...· 0 citations
To address the poor adaptability to complex scenes, blurred boundary details, and the difficult trade-off between accuracy and computational cost in road segmentation from unmanned aerial vehicle (UAV) imagery, this paper proposes DACS-Net, a lightweight road segmentation model based on U-Net and enhanced by dynamic attention and cross-scale semantic information. The model adopts EfficientNetV2-S as a lightweight and efficient encoder, using compound scaling and MBConv bottleneck structures to capture multi-scale features accurately. A dynamic global-local attention module (GLA) is designed to jointly model global context and local neighborhood details through a dynamic global attention branch and an adaptive local branch. Furthermore, a cross-scale semantic feature fusion module (CSF) is constructed to perform intelligent selection and efficient fusion of multi-stage encoder features through semantic-guided weight allocation, edge-enhanced feature alignment, and lightweight channel compression. Experiments on the public AeroScapes dataset show that the proposed method outperforms mainstream models in Dice, Kappa, and other metrics while maintaining relatively low computational complexity, demonstrating its effectiveness and practical value.
Wen-Wei Xiao, Yan Luo, Junlian Liao et al.· 2026 8th International Confe...· 0 citations
Pedestrian detection constitutes a critical task within advanced driver assistance systems (ADAS), where reliable identification of pedestrians is essential for ensuring vehicular safety. Although deep learning has substantially improved detection performance, existing state-of-the-art models continue to exhibit notable degradation in adverse weather conditions and low-light. To mitigate these challenges, this study introduces an enhanced pedestrian detection framework based on you only look one version 5 (YOLOv5s), retrained on an augmented common object in context (COCO) dataset focused on the person class. Additionally, a novel, lightweight, and adaptive attention mechanism called: the weighted fusion efficient channel attention (WF-ECA) module is incorporated into the detection architecture. The WF-ECA module selectively focusses on important features without compromising computational efficiency or inference speed. Comparative experiments demonstrate a 5% increase in mean average precision (mAP) in comparison to the baseline model, thereby demonstrating the efficacy of the proposed attention module in improving detection robustness under challenging environmental conditions. These findings highlight the potential of attention-based mechanisms to enhance pedestrian detection performance in real-world ADAS applications.
Dense small traffic object detection is essential for intelligent transportation systems but remains challenging because distant targets contain limited visual details, densely distributed objects frequently overlap, and complex road backgrounds introduce substantial interference. To address these limitations, this study proposes GAD-YOLO, a multi-level feature enhancement network based on YOLOv9. Ghost-MSConv performs lightweight multi-receptive-field feature extraction in the backbone, Mixed Local Channel Attention combines local spatial relationships with global channel dependencies during feature refinement, and DySample performs content-adaptive point sampling during feature upsampling. In the primary experiments on a six-class traffic object dataset derived from UA-DETRAC, GAD-YOLO achieved a precision of 78.9%, a recall of 76.4%, an mAP50 of 82.8%, and an mAP50:95 of 65.5%. Compared with YOLOv9c, precision, recall, mAP50, and mAP50:95 increased by 5.4, 0.5, 3.1, and 4.8 percentage points, respectively. Under the complexity statistics used in the primary experiments, GAD-YOLO contains 25.455 M parameters and requires 102.4 GFLOPs, compared with 25.442 M parameters and 103.2 GFLOPs for YOLOv9c. Additional experiments on the public VisDrone2019-DET benchmark were conducted to evaluate cross-dataset applicability, small-object performance, scene-density sensitivity, and standardized inference efficiency. On the VisDrone2019-DET test-dev set, GAD-YOLO improved mAP50 and mAP50:95 from 26.5% and 15.7% to 27.1% and 16.3%, respectively. A COCO-style analysis further showed that APS increased from 6.72% to 7.31%, while the dense-subset mAP50:95 increased from 13.95% to 14.44%. Under an RTX 4090, batch-size-one, 640×640, FP32 inference protocol, GAD-YOLO achieved a mean latency of 9.98 ms and a throughput of 100.20 FPS. These results show that GAD-YOLO improves the primary traffic object detection task and yields modest positive performance differences on an independent public benchmark under the fixed experimental setting, while maintaining real-time inference capability.
