Search Results (1,823)

Object detection in UAV aerial imagery is challenged by dense small targets, large-scale variation, complex backgrounds, and strict onboard computation limits. To address these issues, this paper proposes SMMNet (Structured-diffusion Mamba Mixture Network), a lightweight plug-and-play detection framework evaluated with YOLO family detectors. SMMNet contains three modules. The Structured Diffusion Feature Extractor (SDFE) uses anisotropic diffusion to preserve boundary-sensitive features during downsampling. The Mamba-driven Receptive-field Context Aggregator (MRCA) performs multi-directional selective state-space scanning to capture long-range context with linear complexity. The Mask-guided Bayesian Box Refinement (MBBR) applies a MAP-inspired confidence-adaptive box update using MobileSAM mask evidence and ELBO-based false-positive filtering. Using YOLOv13-S as the main detector, SMMNet achieves 32.8% mAP@0.5:0.95 and 52.6% mAP@0.5 on VisDrone2019 at 87 FPS on an NVIDIA A800 GPU, improving the YOLOv13-S baseline by 3.6 and 4.5 points, respectively. The added modules reduce throughput compared with the detector-only baseline (168 FPS), but the resulting 87 FPS remains real-time and provides a favorable accuracy–latency trade-off. Three independent-seed runs further show a mean paired gain of 3.60 ± 0.10 mAP on VisDrone2019, 2.53 ± 0.12 mAP on DroneVehicle, and 2.77 ± 0.06 mAP on SeaDronesSee for the YOLOv13-S setting. Additional experiments on DroneVehicle and SeaDronesSee, together with cross-backbone evaluations on YOLOv5/v6/v7/v8/v10/v11/v13 across different UAV benchmarks, show aligned performance trends under matched settings. Edge deployment on an NVIDIA Jetson Orin NX reaches 30 FPS under TensorRT FP16 inference at 15 W TDP, indicating the suitability of SMMNet for resource-constrained UAV perception. Full article

Underwater optical sensor networks are essential for fish monitoring, yet imagery is often affected by illumination variability, low contrast, and complex backgrounds. Attention mechanisms are vital for feature representation in deep networks, yet existing approaches often struggle with spatial information loss and limited multi-scale interaction under such challenging conditions. This paper introduces Convolution to Interactive Capture and Recalibration Enhancement (C2ICARE), a lightweight attention module designed to overcome these challenges. The principal contribution of C2ICARE is the adaptation of memory interaction principles into an edge-oriented attention framework that enhances feature discrimination while maintaining computational efficiency. The architecture employs three core innovations: a 1:3 memory-feature split to preserve context while reducing cost, parallel multi-scale depthwise convolutions (3 × 3 and 7 × 7) for fine-grained and broad feature extraction, and a cross-branch interaction mechanism coupled with a ConvNeXt-style feed-forward network that avoids dimensionality reduction. Experimental results on an underwater fish dataset demonstrate that YOLO26n with C2ICARE achieves a mean average precision (mAP@0.5:0.95) of 0.7033, outperforming Coordinate Attention (+3.8%), FasterBlock (+1.7%), and CBAM (+0.4%) while adding only 0.05M parameters and 0.16 GFLOPs. Multi-objective Pareto Frontier analysis confirms that C2ICARE provides an effective balance between accuracy, efficiency, and generalization for resource-constrained deployment. EigenCAM visualizations further validate that the model focuses on biological morphology rather than background noise. Its lightweight design enables seamless integration with underwater sensor networks and fog platforms for real-time fish detection in aquaculture, commercial fisheries, and scientific research. Future work will investigate broader marine applications and cross-platform deployment scenarios. The code is available on GitHub. Full article

Small object detection is fundamentally constrained by the lack of discriminative fine-grained features. Although introducing higher resolution detection scales can improve performance, it also amplifies background noise. In addition, the independently decoupled design of conventional detection heads is insufficient to address the persistent challenges of missed detections and false positives for small objects. To this end, we propose MSIA-YOLO, a YOLOv11-based detector with multi-scale semantic interaction and alignment, optimized from three complementary perspectives: feature modeling, high resolution semantic compensation, and task coordinated alignment. First, Receptive Field Attention Convolution (RFAConv) is integrated into the backbone to enhance critical local details, such as edge and texture cues, via receptive field aware attention. Second, to alleviate fine detail attenuation caused by repeated downsampling, we construct a CHSP-P2 small object detection framework with an additional P2 branch. A scale sequence fusion mechanism is further introduced to perform high resolution semantic compensation through cross scale hybrid inputs. Finally, we design a DTIA-Head (Dynamic Task Interaction and Alignment Head), which promotes joint optimization of classification and localization through dynamic task interaction and spatial alignment. Extensive experiments on the public datasets VisDrone, TinyPerson, and RSOD show that, compared with the YOLOv11n baseline, MSIA-YOLO improves mAP50 by 7.7%, 10.3%, and 1.0%, respectively, while also outperforming several advanced detectors. These results demonstrate the effectiveness and generalization capability of the proposed method in small object, dense object, and complex scene object detection scenarios. Full article

Respiratory disease classification using lung sound recordings remains challenging due to signal interference, heterogeneous acquisition conditions, and substantial overlap among clinically related acoustic patterns. This study presents a framework for respiratory disease classification using NMF-enhanced log-mel spectrograms and deep neural classifiers. Respiratory sound recordings from two publicly available datasets were harmonized into a unified label space comprising Asthma, Bronchiectasis, Bronchiolitis, COPD, Healthy, Pneumonia and URTI. Following signal standardization and fixed-length segmentation, a non-negative matrix factorization (NMF)-based enhancement stage was applied to increase the salience of respiratory components prior to log-mel spectrogram generation. The proposed classifier was a convolutional recurrent neural network (CRNN) that combined convolutional feature extraction, bidirectional recurrent modelling, and attention-based temporal aggregation. For comparison, RDLINet, a conventional CNN, ResNet, and a YOLO-style backbone were implemented under the same preprocessing and training framework. Experimental results demonstrated that the proposed CRNN achieved the best overall performance, attaining 96.14 ± 0.50% accuracy and 94.05 ± 1.21% Macro-F1 on the unified seven-class cohort. Class-wise analysis, confusion-matrix evaluation, and output-space visualization further showed that the CRNN provided more balanced recognition across disease categories and clearer class separation than competing architectures. These findings indicate that NMF-enhanced spectro-temporal modelling combined with convolutional recurrent learning offers an effective approach for automated multi-class respiratory disease classification. Full article

Crop targets in UAV aerial images are typically characterized by small scale, dense distribution, severe mutual occlusion, and complex backgrounds, which often lead to low detection accuracy and large counting errors for existing deep learning models. To address these issues, this study proposes an improved YOLOv12-based crop detection and counting model, named YOLO-DC. By introducing an attention mechanism (LGCB-AM) and a multi-scale detection head (MS-DH), the proposed model effectively enhances local texture extraction, global modeling, foreground–background contrast, and boundary perception for dense small objects. Subsequently, a series of comparative experiments, ablation studies, and transfer experiments were conducted on the wheat and rice datasets. The results show that YOLO-DC achieves a favorable balance among detection accuracy, counting error, and model efficiency and overall outperforms the other comparison models. Ablation studies further verify the effectiveness of the proposed design, showing that LGCB-AM is the key contributor to the performance improvement, while the boundary branch and repulsion branch play critical roles in dense-target discrimination. In addition, an appropriate module insertion strategy can effectively balance high-level semantic enhancement and feature fusion stability. Transfer experiments demonstrate that pretraining on the wheat dataset and fine-tuning on the rice dataset significantly outperform training from scratch, indicating strong cross-crop transfer potential. Overall, the proposed YOLO-DC provides an effective solution for high-precision crop detection and counting in agricultural scenarios. Full article

In electronic training scenarios, the safety risk of a soldering iron cannot be determined by object detection alone, as its state must be further distinguished among hand-held, stand-supported, desk-exposed, and uncertain interactions. To address this problem, this paper proposes RISNet, the Relation-aware Interaction State Network, which establishes a two-stage instance-level interaction understanding framework for soldering iron safety monitoring. In the first stage, YOLO is used to generate candidate instances of soldering irons and related environmental objects, and dual-layer feature fusion is adopted to jointly exploit shallow details and deep semantics. In the second stage, the soldering iron is treated as the interaction subject. The Pointer-Head models associations between the subject and contextual objects, and the State-Head predicts the safety state conditioned on subject-object relational constraints. To reduce false alarms from false detections and weak interactions, RISNet introduces a Quality-Head that estimates the reliability of each interaction conclusion and filters low-quality predictions during inference. The unknown label is used during training as conservative supervision for weak, unreliable, or indeterminate interaction evidence, with semantics close to the no-interaction label in HOI. This paper also constructs the Soldering Iron Safety Interaction Dataset (SISID) to support detection, interaction modeling, and state evaluation of slender metallic tools in training scenarios. On the SISID validation split, RISNet achieves an Overall F1 of 95.38%, an Overall Precision of 96.73%, and an inference speed of 57.1 FPS, satisfying the centralized single-frame polling requirement considered in this work. Full article

Given the issues in natural orchard environments, such as large-scale variations of defective citrus fruits, weak texture boundaries, strong illumination changes, branch and leaf occlusion, and significant background interference, this paper constructs a lightweight detection model, RT-DETR-DCEA, based on RT-DETR-R18. This model is improved through four aspects: “fine-grained defective feature extraction—multi-scale feature fusion—up-sampling detail recovery—global feature interaction for noise suppression”. First, a Dynamic Hybrid Convolution Module (DIMB) is introduced into the backbone network, drawing on the ideas of Inception-style multi-branch depthwise convolution and MetaFormer residual mixing. It extracts local textures of various forms through square convolution, horizontal strip convolution, and vertical strip convolution, and utilizes dynamic branch weights to enhance the model’s adaptability to irregular defects such as lesions, mildew, and external damage. Second, a Content-Guided Attention Feature Fusion Network (CGAFN) is designed in the neck network, which achieves adaptive fusion of low-level detail features and high-level semantic features through channel attention, spatial attention, and pixel-level fusion weights. Next, a lightweight upsampling enhancement module called EUCB-SC is constructed, which introduces channel rearrangement and Shift spatial offset into the efficient upsampling convolutional structure to enhance the local spatial interaction capability of upsampled features with low parameter overhead. Finally, adaptive sparse self-attention is introduced into the AIFI module to form AIFI-ASSA, which suppresses irrelevant background interactions through a sparse attention branch and retains necessary contextual information through a dense attention branch. The experimental results demonstrate that on a dataset containing four categories of citrus images—healthy, diseased, moldy, and severely externally damaged—RT-DETR-DCEA achieves 92.1% Precision, 86.1% Recall, and 91.8% mAP@50, with a parameter count of 1.477 × 10⁷ and an inference speed of 81 FPS. Compared with the original RT-DETR-R18 and various YOLO series models, this method strikes a favorable balance among detection accuracy, recall capability, and model lightweightness. This paper also discusses limitations such as data scale, ratio of private data, single training result, and insufficient validation on edge devices, providing a basis for subsequent cross-regional data validation and real-world deployment testing. Full article

Vehicle-to-Vehicle (V2V) cooperative object detection enhances environmental perception capabilities in complex traffic scenarios by sharing sensory information among vehicles, but limited transmission bandwidth and wireless channel noise can significantly affect the reliable transmission of cross-vehicle semantic features and lead to a degradation in detection performance at the receiver. Although existing semantic communication methods based on DeepJSCC can alleviate the cliff effect of traditional separated source–channel coding under low signal-to-noise ratio conditions, they typically rely on additional external autoencoder structures, which increase model complexity and the deployment burden on vehicular edge computing platforms. Meanwhile, under high compression ratios, these methods struggle to adequately preserve detection-related fine-grained information, such as object boundaries, spatial locations, and local structures. Motivated by these challenges, we develop Mamba-KGSC as a lightweight knowledge-guided semantic communication framework for robust V2V cooperative object detection. At the transmitter, Mamba-KGSC utilizes the internal time-scale parameters of the Mamba-YOLO-T backbone network to generate spatial semantic masks, realizing the sparse encoding and transmission of task-relevant features while avoiding the introduction of complex external codec networks. At the receiver, a multi-source knowledge base constraint verification module is constructed to refine the initial detection results by combining physical consistency screening with visual–physical spatial joint redundancy suppression, thereby suppressing physically inconsistent misdetections and repeated detections induced by channel noise. The experimental evaluation indicates that, under a 50% compression ratio, multiple SNR settings, and different channel models, the front-end semantic communication branch of Mamba-KGSC improves mAP@0.5:0.95 by an average of 1.90 percentage points over the DeepJSCC baseline. The multi-source knowledge base constraint verification module further reduces abnormal and duplicate candidate bounding boxes. Overall, Mamba-KGSC provides a balanced solution in terms of transmission cost, detection accuracy, model complexity, and physical consistency, offering a lightweight implementation scheme for robust V2V cooperative detection in challenging communication environments. Full article

To address the issues of low detection accuracy and high miss rates in long-range small traffic sign detection, which are caused by insufficient feature information and susceptibility to background interference, this paper proposes an improved real-time traffic sign detection algorithm based on YOLO11n. First, a cross-guided feature extraction module, C3k2_CGPEMA, is designed within the neck network. By embedding the Efficient Multi-Scale Attention (EMA) mechanism into the feature extraction branch of Partial Convolution (PConv), this module utilizes the spatial attention mask generated by the convolutional branch to provide cross-branch guidance and filter out complex background noise from the identity branch. This achieves precise fine-grained feature focusing while preserving high-frequency spatial details. Furthermore, a joint bounding box regression loss function combining Complete Intersection over Union (CIoU) and Gaussian Combined Distance (GCD) is adopted. This preserves the stable convergence properties of CIoU while leveraging the scale invariance of GCD to enhance the regression accuracy for small targets. Finally, the detection layers are reconstructed by removing the P5 layer and introducing a high-resolution P2 layer (160 × 160), significantly strengthening the localization capability for distant, tiny targets. Experimental results demonstrate that the proposed algorithm achieves improvements of 5.4, 7.4, and 6.6 points in precision, recall, and mAP@0.5, respectively, on the TT100K dataset compared to the baseline YOLO11n. While boosting detection accuracy, the model maintains an inference speed of 114.5 frames per second (FPS), fully satisfying the requirements for real-time detection in in-vehicle environments. Generalization experiments conducted on the CCTSDB dataset further validate the robustness of the proposed algorithm in complex environments. Full article

In modern farm management, accurate detection of estrus behavior in dairy cows is essential for improving reproductive efficiency and enabling intelligent decision-making. Although fisheye lenses offer a wider field of view, they often introduce image distortion. This leads to geometric and scale deformation of cow mounting behavior features, which reduces detection accuracy. To address this issue, a lightweight model called Boundary-Enhanced Frequency-Domain Feature Pyramid Network YOLO (BFFPN-YOLO) was developed. It is designed for detecting dairy cow mounting behavior under fisheye imaging, incorporating boundary enhancement and frequency-domain compensation. Initially, the backbone network was equipped with the multi-scale dilated fusion structure SPPELAN. This structure expands the receptive field and preserves detailed information, thereby enhancing boundary modeling for targets with scale variations. Subsequently, a boundary-enhanced frequency-domain feature pyramid network (BFFPN) module was designed for reconstructing the top-down transmission path in the Neck. The module is composed of the frequency-domain detail compensation FreqFusion and the spatial attention enhancement SEAM. By strengthening boundary responses, compensating for high-frequency details, and replacing the traditional upsampling and concatenation operations, it effectively mitigates blurred target boundaries in images of dairy cow mounting behavior. The improved algorithm demonstrates strong detection performance, achieving a Precision of 88%, a Recall of 84.5%, and an mAP@0.5 of 92.7%. Compared with the original YOLOv11, these metrics were increased by 3.8, 2.3, and 4.6 percentage points, respectively. The model parameter count was reduced by 1.10 × 10⁶. In complex scenarios, edge features and high-frequency details of dairy cow mounting behavior are more accurately captured by the improved model. These improvements provide a reliable technical basis for the intelligent detection of estrus behavior. Full article

A problem was formulated, and methods and algorithms for identifying, recognizing, and classifying micro-objects were developed using problem-oriented image processing systems utilizing the characteristics of image structural components, dynamic models, and neural networks. The micro-objects studied were pollen grains and unicellular microorganisms, the application of which is in demand in palynology, environmental protection, ecology, and medicine. Models, algorithms, and software were developed based on the statistical, dynamic, morphometric, textural, and brightness characteristics of point image signals. The main contribution of the work is a hybrid model combining Daubechies wavelet functions (orders 4 and 8) with convolutional neural networks, which provides an average pollen grain recognition and classification accuracy of up to 97.7% and an overall classification accuracy of 98.2%. For comparative evaluation, the YOLO11m model was used as an independent baseline deep learning model, achieving 88.6% overall accuracy on a validation set of 280 images. The software includes tools for Gaussian filtering, median filtering, Sobel and Canny operators, and threshold correction of defective points with hard and soft control rules. The software package includes modules for training a convolutional neural network. The study was conducted on real-world databases of micro-object images for crop breeding and environmental pollution assessment. Full article

Insulators are critical insulation components in power transmission lines; however long-term exposure to adverse environmental conditions may threaten the safety and stability of power delivery. Existing studies primarily emphasize detection accuracy, while deployment efficiency and inference speed have received insufficient attention, limiting their applicability to CPU-based edge computing devices. To address these limitations, this paper proposes PEBL-YOLO, a lightweight model for insulator defect detection. The proposed model retains the external C3k2 structure of YOLOv11 while simplifying its internal bottleneck module, in which PConv is embedded to improve spatial feature extraction and fusion efficiency. In the neck, the original Path Aggregation Feature Pyramid Network (PAFPN) is reconstructed by integrating a Bidirectional Feature Pyramid Network (BiFPN) with Efficient Channel Attention (ECA), enabling more effective aggregation of multi-scale features and stronger focus on defect-related regions with minimal parameter increase. Moreover, a lightweight shared decoupled detection head is designed to decouple classification and regression branches. By combining parameter sharing with Group Normalization (GN) the detection head further reduces model complexity while maintaining accurate localization capability. Experimental results show that PEBL-YOLO contains only 1.68 M parameters. It achieves Precision, Recall, mAP@0.5, and mAP@0.5:0.95 of 95.0%, 92.1%, 94.4%, and 53.6%, respectively. These results demonstrate that PEBL-YOLO achieves a favorable trade-off between detection accuracy and parameter efficiency, providing a practical solution for lightweight insulator defect detection in edge computing scenarios. Full article

Background/Objectives: Automated cephalometric landmark detection using deep learning has the potential to streamline routine orthodontic diagnosis. However, the clinical relevance of artificial intelligence (AI) localisation accuracy depends on how detection errors propagate into derived angular measurements and skeletal classifications. We retrospectively evaluated 14 YOLO-based model configurations and quantified the agreement between AI-derived and expert-derived ANB-based skeletal classifications. Methods: Twelve working YOLO-based models (YOLOv5xu, YOLOv11 nano/small/medium/large variants) were trained on a single-centre dataset of 120 lateral cephalograms and evaluated on an independent test set of 11 cephalograms (stratified across skeletal Classes I, II, III). The four ANB-defining landmarks (Sella, Nasion, A-point, B-point) were the focus of the analysis. Each test cephalogram had been annotated by four orthodontists (44 measurements per image), yielding the expert reference. We assessed the effects of architecture, bounding-box size (40/100/150 px), training dataset scale (235–4255 images) and training epochs on localisation accuracy (mean radial error, MRE; Successful Detection Rate, SDR) and on the downstream ANB-based skeletal classification. Diagnostic concordance was quantified by classification agreement, Cohen’s $\kappa$ with bootstrap 95% confidence intervals (10,000 iterations), an exact one-sided binomial test for discordance, and Wilson exact CIs per class. Results: The best-performing model (Model 2; YOLOv11l, 40 × 40 px bounding box, 1175 training images) achieved an MRE of $3.10\pm 1.00$ mm and a SDR@4 mm of 87.2% for S, N, A, and B. ANB-based skeletal classification demonstrated 96.9% concordance with expert assessments (95% bootstrap CI: 93.8–99.2%; Cohen’s $\kappa$ = 0.946 [95% CI 0.89–0.99]; exact binomial test against a 90% concordance threshold $p=0.003$). Per-class concordance was Class I 95.8% (23/24), Class II 94.9% (56/59), and Class III 100% (47/47). Three of four discordant cases clustered near the Class I/II diagnostic threshold (expert ANB $\approx 4.5°$). Bounding-box size dominated localisation accuracy, with a 3.5-fold increase in MRE from 40 × 40 to 150 × 150 px configurations and SDR@4 mm collapsing from 82.8% to 0%. Conclusions: Within the constraints of a retrospective single-centre design with a small (n = 11) independent test set, YOLO-based AI landmark detection demonstrated promising diagnostic concordance with expert consensus for ANB-based skeletal classification. These findings warrant prospective, multi-centre external validation before clinical deployment and support a confidence-aware workflow in which AI predictions for borderline ANB values undergo mandatory clinician verification. Bounding-box calibration emerged as the single most impactful preprocessing decision. Full article

Severe environmental pollution and complex weather changes significantly hinder the effectiveness of vehicle traffic flow statistics and traffic monitoring technologies. Therefore, achieving fast and accurate vehicle detection in complex environments has become one of the key tasks in the new era. This paper proposes a vehicle detection method for complex environments based on YOLOv11, named YOLO-FTG. First, the neck network of the YOLOv11 baseline model is improved by adding a P2 detection layer and the corresponding detection head. Second, a Spatial-Frequency Hybrid Convolution (SFHC) module is designed. Third, a Global-Local Adaptive Module (GLAM) is proposed. To verify the effectiveness of the proposed model, experiments were conducted on three self-constructed datasets and the public BDD100K dataset. The experimental results demonstrate that compared with existing methods, the YOLO-FTG model achieves higher accuracy, with mAP50 scores of 75.3%, 98.71%, 51.00%, and 64.59% on the four datasets, respectively. These scores represent improvements of 3.24%, 0.34%, 5.64%, and 3.51% over the baseline model, respectively. While maintaining real-time inference speed, these results indicate the effectiveness and robustness of the proposed model in complex environments. Full article

Human pose detection is fundamental to quantitative sports training analysis in college general education courses, enabling an objective assessment of college students’ movement quality and the early identification of sports injury risks among non-professional athletes. At present, those detectors based on YOLO have encountered difficulties in capturing the continuous movement patterns of college athletes in routine training, maintaining the regression accuracy of different size posture targets, and maintaining the real-time calculation speed in the campus sports environment. Furthermore, most existing pose-estimation frameworks are optimized for general scenes and fail to address the unique challenges of college physical education settings, including non-standard student movements, diverse skill levels, and strict cost constraints for large-scale deployment. In order to solve these problems, we put forward PPLCNet-YOLOv11, which is a simplified human posture-estimation framework designed for college physical education. This model is optimized by three key improvements: (1) replacing the original backbone network with PPLCNet to enhance feature extraction, while strictly observing the strict FLOPs and parameter restrictions; (2) an enhanced Multi-Scale Attention Mechanism (MSAM) that combines adaptive scale perception, hierarchical channel attention, and pose-sensitive spatial attention to better represent elongated anatomical structures and multi-scale pose cues; and (3) an improved enhanced IoU loss function that incorporates scale-aware and aspect-ratio-aware penalty terms to refine the bounding box adjustment for atypical and sports-specific gestures. Experiments on both a dedicated college student sports pose dataset and two public benchmark datasets (COCO Keypoints 2017 and MPII Human Pose) demonstrate that PPLCNet-YOLOv11 achieves 77.8% mAP@0.5 and 37.09% mAP@0.95 based on the campus dataset, with 82.34% precision and 75.00% recall, while requiring only 2.62 M parameters and 6.38 GFLOPs. Extensive inference speed tests show that the model achieves 127 FPS on an NVIDIA RTX 4090 GPU, 38 FPS on an Intel i7-12700 CPU, and 16 FPS on a Jetson Nano edge device, meeting the real-time requirements of campus sports monitoring. Compared with mainstream lightweight YOLO variants and state-of-the-art specialized pose-estimation models, our proposed method improves mAP@0.5 by 4.93–12.6 percentage points based on the campus dataset. All experiments were repeated five times with different random seeds, and we report mean values with standard deviations and statistical significance tests to ensure result reliability. These results indicate that PPLCNet-YOLOv11 provides an accurate and resource-efficient solution for real-time pose evaluation in college physical training. Full article