Search Results (6,187)

High-precision detection of rice targets in precision agriculture is crucial for yield assessment and field management. However, existing models still face challenges, such as high rates of missed detections and insufficient localization accuracy, particularly when dealing with small targets and dynamic changes in scale and morphology. This paper proposes an accurate rice detection model for UAV images based on Adaptive Aware Dynamic Convolution, named Adaptive Dynamic Convolution YOLO (ADC-YOLO), and designs the Adaptive Aware Dynamic Convolution Block (ADCB). The ADCB employs a “Morphological Parameterization Subnetwork” to learn pixel-specific kernel shapes and a “Spatial Modulation Subnetwork” to precisely adjust sampling offsets and weights—realizing for the first time the adaptive dynamic evolution of convolution kernel morphology with variations in rice scale. Furthermore, ADCB is embedded into the interaction nodes of the YOLO backbone and neck; combined with depthwise separable convolution in the neck, it synergistically enhances multi-scale feature extraction from rice images. Experiments on public datasets show that ADC-YOLO comprehensively outperforms state-of-the-art algorithms in terms of AP50 and AP75 metrics and maintains stable high performance in scenarios such as small targets at the seedling stage and leaf overlap. This work provides robust technical support for intelligent rice field monitoring and advances the practical application of computer vision in precision agriculture. Full article

Object recognition in remote sensing images is essential for applications such as land resource monitoring, maritime vessel detection, and emergency disaster assessment. However, detection accuracy is often limited by complex backgrounds, densely distributed targets, and multiscale variations. To address these challenges, this study aims to improve the detection of small-scale and densely distributed objects in complex remote sensing scenes. An improved object detection network is proposed, called omnidirectional and adaptive YOLOv8 (OA-YOLOv8), based on the YOLOv8 architecture. Two targeted enhancements are introduced. First, an omnidirectional perception refinement (OPR) network is embedded into the backbone to strengthen multiscale feature representation through the incorporation of receptive-field convolution with a triplet attention mechanism. Second, an adaptive channel dynamic upsampling (ACDU) module is designed by combining DySample, the Haar wavelet transform, and a self-supervised equivariant attention mechanism (SEAM) to dynamically optimize channel information and preserve fine-grained features during upsampling. Experiments on the satellite imagery multi-vehicle dataset (SIMD) demonstrate that OA-YOLOv8 outperforms the original YOLOv8 by 4.6%, 6.7%, and 4.1% in terms of mAP@0.5, precision, and recall, respectively. Visualization results further confirm its superior performance in detecting small and dense targets, indicating strong potential for practical remote sensing applications. Full article

Accurate forecasting of citywide electric vehicle (EV) charging load is critical for alleviating station-level congestion, improving energy dispatching, and supporting the stability of intelligent transportation systems. However, large-scale EV charging networks exhibit complex and heterogeneous spatio-temporal dependencies, and existing approaches often struggle to scale with increasing station density or long forecasting horizons. To address these challenges, we develop a modular spatio-temporal prediction framework that decouples temporal sequence modeling from spatial dependency learning under an encoder–decoder paradigm. For temporal representation, we introduce a global aggregation mechanism that compresses multi-station time-series signals into a shared latent context, enabling efficient modeling of long-range interactions while mitigating the computational burden of cross-channel correlation learning. For spatial representation, we design a dynamic multi-scale attention module that integrates graph topology with data-driven neighbor selection, allowing the model to adaptively capture both localized charging dynamics and broader regional propagation patterns. In addition, a cross-step transition bridge and a gated fusion unit are incorporated to improve stability in multi-horizon forecasting. The cross-step transition bridge maps historical information to future time steps, reducing error propagation. The gated fusion unit adaptively merges the temporal and spatial features, dynamically adjusting their contributions based on the forecast horizon, ensuring effective balance between the two and enhancing prediction accuracy across multiple time steps. Extensive experiments on a real-world dataset of 18,061 charging piles in Shenzhen demonstrate that the proposed framework achieves superior performance over state-of-the-art baselines in terms of MAE, RMSE, and MAPE. Ablation and sensitivity analyses verify the effectiveness of each module, while efficiency evaluations indicate significantly reduced computational overhead compared with existing attention-based spatio-temporal models. Full article

Accurate prediction of drug–target interaction (DTI) is crucial for modern drug discovery. However, experimental assays are costly, and many existing computational models still face challenges in capturing multi-scale features, fusing cross-modal information, and modeling fine-grained drug–protein interactions. To address these challenges, We propose Gated-Attention Dual-Fusion Drug–Target Interaction (GADFDTI), whose core contribution is a fusion module that constructs an explicit atom–residue similarity field, refines it with a lightweight 2D neighborhood operator, and performs gated bidirectional aggregation to obtain interaction-aware representations. To provide strong and width-aligned unimodal inputs to this fusion module, we integrate a compact multi-scale dense GCN for drug graphs and a masked multi-scale self-attention protein encoder augmented by a narrow 1D-CNN branch for local motif aggregation. Experiments on two benchmarks, Human and C. elegans, show that GADFDTI consistently outperforms several recently proposed DTI models, achieving AUC values of 0.986 and 0.996, respectively, with corresponding gains in precision and recall. A SARS-CoV-2 case study further demonstrates that GADFDTI can reliably prioritize clinically supported antiviral agents while suppressing inactive compounds, indicating its potential as an efficient in silico prescreening tool for lead-target discovery. Full article

Surface defects in hydraulic concrete structures exhibit extreme topological heterogeneity. and are frequently obscured by unstructured environmental noise. Conventional detection models, constrained by fixed-grid convolutions, often fail to effectively capture these irregular geometries or suppress background artifacts. To address these challenges, this study proposes the Adaptive Local–Global Synergistic Perception Network (ALGSP-Net). First, to overcome geometric constraints, the Defect-aware Receptive Field Aggregation and Adaptive Dynamic Receptive Field modules are introduced. Instead of rigid sampling, this design adaptively modulates the receptive field to align with defect morphologies, ensuring the precise encapsulation of slender cracks and interlaced spalling. Second, a dual-stream gating fusion strategy is employed to mitigate semantic ambiguity. This mechanism leverages global context to calibrate local feature responses, effectively filtering background interference while enhancing cross-scale alignment. Experimental results on the self-constructed SDD-HCS dataset demonstrate that the method achieves an average Precision of 77.46% and an mAP50 of 72.78% across six defect categories. Comparative analysis confirms that ALGSP-Net outperforms state-of-the-art benchmarks in both accuracy and robustness, providing a reliable solution for the intelligent maintenance of hydraulic infrastructure. Full article

Existing unsupervised anomaly detection methods suffer from insufficient parameter precision, poor robustness to noise, and limited generalization capability. To address these issues, this paper proposes an Adaptive Diffusion Adversarial Evolutionary Network (ADAEN) for unsupervised anomaly detection in tabular data. The proposed network employs an adaptive hierarchical feature evolution generator that captures multi-scale feature representations at different abstraction levels through learnable attribute encoding and a three-layer Transformer encoder, effectively mitigating the gradient vanishing problem and the difficulty of modeling complex feature relationships that are commonly observed in conventional generators. ADAEN incorporates a multi-scale adaptive diffusion-augmented discriminator, which preserves scale-specific features across different diffusion stages via cosine-scheduled adaptive noise injection, thereby endowing the discriminator with diffusion-stage awareness. Furthermore, ADAEN introduces a multi-scale robust adversarial gradient loss function that ensures training stability through a diffusion-step-conditional Wasserstein loss combined with gradient penalty. The method has been evaluated on 14 UCI benchmark datasets and achieves state-of-the-art performance in anomaly detection compared to existing advanced algorithms, with an average improvement of 8.3% in AUC, an 11.2% increase in F1-Score, and a 15.7% reduction in false positive rate. Full article

The accurate prediction of the Remaining Useful Life (RUL) of key mechanical equipment in modern industry is crucial for reducing production risks and optimizing maintenance decisions. However, existing Convolutional Neural Network (CNN)-based models lack an inherent memory mechanism, and single convolutional kernel-based CNN models fail to capture multi-scale temporal features effectively. Moreover, some existing methods fail to account for the stability of the model training process, which tends to result in prolonged training time and an elevated risk of overfitting. To overcome these problems, a pre-activated residual parallel convolutional block-based BiGRU model (PRPC-BiGRU) is proposed in this study. First, the residual parallel convolutional block (RPCB) is constructed to simultaneously extract multi-scale temporal features. Subsequently, the pre-activated convolutional structure, which applies normalization and activation function prior to convolution operations, is utilized to improve gradient propagation and training stability. Finally, experimental results using the aero-engine benchmark datasets to verify the effectiveness and superior prediction performance of the proposed PRPC-BiGRU model. Full article

The quality of tea bud harvesting directly affects the final quality of the tea; however, due to the small size of tea buds and the complex natural background, accurately detecting them remains challenging. To address this issue, this paper proposes a lightweight and efficient tea bud detection model named TSF-Net. This model adopts the P2-enhanced bidirectional feature pyramid network (P2A-BiFPN) to enhance the recognition ability of small objects and achieve efficient multi-scale feature fusion. Additionally, coordinate space attention (CSA) is embedded in multiple C3k2 blocks to enhance the feature extraction of key regions, while an A2C2f module based on self-attention is introduced to further improve the fine feature representation. Extensive experiments conducted on the self-built WYTeaBud dataset show that TSF-Net increases mAP@50 by 2.0% and reduces the model parameters to approximately 85% of the baseline, achieving a good balance between detection accuracy and model complexity. Further evaluations on public tea bud datasets and the VisDrone2019 small object benchmark also confirm the effectiveness and generalization ability of the proposed method. Moreover, TSF-Net is converted to the RKNN format and successfully deployed on the RK3588 embedded platform, verifying its practical applicability and deployment potential in intelligent tea bud harvesting. Full article

Remote sensing image segmentation requires balancing global context and local detail across multi-scale objects. However, convolutional neural network (CNN)-based methods struggle to model long-range dependencies, while transformer-based approaches suffer from quadratic complexity and become inefficient for high-resolution remote sensing scenarios. In addition, the semantic gap between deep and shallow features can cause misalignment during cross-layer aggregation, and information loss in upsampling tends to break thin continuous structures, such as roads and roof edges, introducing pronounced structural noise. To address these issues, we propose lightweight Lite-BSSNet (Blueprint-Guided State Space Network). First, a Structural Blueprint Generator (SBG) converts high-level semantics into an edge-enhanced structural blueprint that provides a topological prior. Then, a Visual State Space Bridge (VSS-Bridge) aligns multi-level features and projects axially aggregated features into a linear-complexity visual state space, smoothing high-gradient edge signals for sequential scanning. Finally, a Structural Repair Block (SRB) enlarges the effective receptive field via dilated convolutions and uses spatial/channel gating to suppress upsampling artifacts and reconnect thin structures. Experiments on the ISPRS Vaihingen and Potsdam datasets show that Lite-BSSNet achieves the highest segmentation accuracy among the compared lightweight models, with mIoU of 83.9% and 86.7%, respectively, while requiring only 45.4 GFLOPs, thus achieving a favorable trade-off between accuracy and efficiency. Full article

Mikania micrantha is one of the most detrimental invasive plant species in the southeastern coastal region of China. To accurately predict the invasion pattern of Mikania micrantha and offer guidance for production practices, it is essential to determine its precise location and the driving factors. Therefore, a design of the wavelet convolution and dynamic feature fusion module was studied, and WaveEdgeNet was proposed. This model has the abilities to deeply extract image semantic features, retain features, perform multi-scale segmentation, and conduct fusion. Moreover, to quantify the impact of human and natural factors, we developed a novel proximity factor based on land use data. Additionally, a new feature selection framework was applied to identify driving factors by analyzing the relationships between environmental variables and Mikania micrantha. Finally, the MaxEnt model was utilized to forecast its potential future habitats. The results demonstrate that WaveEdgeNet effectively extracts image features and improves model performance, attaining an MIoU of 85% and an overall accuracy of 98.62%, outperforming existing models. Spatial analysis shows that the invaded area in 2024 was smaller than that in 2023, indicating that human intervention measures have achieved some success. Furthermore, the feature selection framework not only enhances MaxEnt’s accuracy but also cuts down computational time by 82.61%. According to MaxEnt modeling, human disturbance, proximity to forests, distance from roads, and elevation are recognized as the primary factors. In the future, we will concentrate on overcoming the seasonal limitations and attaining the objective of predicting the growth and reproduction of kudzu before they happen, which can offer a foundation for manual intervention. This study lays a solid technical foundation and offers comprehensive data support for comprehending the species’ dispersal patterns and driving factors and for guiding environmental conservation. Full article

High-speed jet-in-crossflow (JICF) configurations are central to several aerospace applications, including turbine-blade film cooling, thrust vectoring, and fuel or hydrogen injection in combusting or reacting flows. This study employs high-fidelity direct numerical simulations (DNS) to investigate the dynamics of a supersonic jet (Mach 3.73) interacting with a subsonic crossflow (Mach 0.8) at low Reynolds numbers. Three momentum-flux ratios (J = 2.8, 5.6, and 10.2) are considered, capturing a broad range of jet–crossflow interaction regimes. Turbulent inflow conditions are generated using the Dynamic Multiscale Approach (DMA), ensuring physically consistent boundary-layer turbulence and accurate representation of jet–crossflow interactions. Modal decomposition via proper orthogonal decomposition (POD) and spectral POD (SPOD) is used to identify the dominant spatial and spectral features of the flow. Across the three configurations, near-wall mean shear enhances small-scale turbulence, while increasing J intensifies jet penetration and vortex dynamics, producing broadband spectral gains. Downstream of the jet injection, the spectra broadly preserve the expected standard pressure and velocity scaling across the frequency range, except at high frequencies. POD reveals coherent vortical structures associated with shear-layer roll-up, jet flapping, and counter-rotating vortex pair (CVP) formation, with increasing spatial organization at higher momentum ratios. Further, POD reveals a shift in dominant structures: shear-layer roll-up governs the leading mode at high J, whereas CVP and jet–wall interactions dominate at lower J. Spectral POD identifies global plume oscillations whose Strouhal number rises with J, reflecting a transition from slow, wall-controlled flapping to faster, jet-dominated dynamics. Overall, the results demonstrate that the momentum-flux ratio (J) regulates not only jet penetration and mixing but also the hierarchy and characteristic frequencies of coherent vortical, thermal, and pressure and acoustic structures. The predominance of shear-layer roll-up over counter-rotating vortex pair (CVP) dynamics at high J, the systematic upward shift of plume-oscillation frequencies, and the strong analogy with low-frequency shock–boundary-layer interaction (SBLI) dynamics collectively provide new mechanistic insight into the unsteady behavior of supersonic jet-in-crossflow flows. Full article

In clinical CT imaging, high-density metallic implants often induce severe metal artifacts that obscure critical anatomical structures and degrade image quality, thereby hindering accurate diagnosis. Although deep learning has advanced CT metal artifact reduction (CT-MAR), many methods do not effectively use frequency information, which can limit the recovery of both fine details and overall image structure. To address this limitation, we propose a Hybrid-Frequency-Aware Mixture-of-Experts (HFMoE) network for CT-MAR. The proposed method synergizes the spatial-frequency localization of the wavelet transform with the global spectral representation of the Fourier transform to achieve precise multi-scale modeling of artifact characteristics. Specifically, we design a hybrid-frequency interaction encoder with three specialized branches, incorporating wavelet-domain, Fourier-domain, and cascaded wavelet–Fourier modulation, to distinctively refine local details, global structures, and complex cross-domain features. Then, they are fused via channel attention to yield a comprehensive representation. Furthermore, a Frequency-Aware Mixture-of-Experts (MoE) mechanism is introduced to dynamically route features to specific frequency experts based on the degradation severity, thereby adaptively assigning appropriate receptive fields to handle varying metal artifacts. Evaluations on synthetic (DeepLesion) and clinical (SpineWeb, CLINIC-metal) datasets show that HFMoE outperforms existing methods in both quantitative metrics and visual quality. Our method demonstrates the value of explicit frequency-domain adaptation for CT-MAR and could inform the design of other image restoration tasks. Full article

Defect detection on Printed Circuit Boards (PCBs) constitutes a pivotal component of the quality control system in electronics manufacturing. However, owing to the intricate circuitry structures on PCB surfaces and the characteristics of defects—specifically their minute scale, irregular morphology, and susceptibility to background texture interference—existing generic deep learning models frequently fail to achieve an optimal equilibrium between detection accuracy and inference speed. To address these challenges, this study proposes MDEB-YOLO, a lightweight real-time detection network tailored for PCB micro-defects. First, to enhance the model’s perceptual capability regarding subtle geometric variations along conductive line edges, we designed the Efficient Multi-scale Deformable Attention (EMDA) module within the backbone network. By integrating parallel cross-spatial channel learning with deformable offset networks, this module achieves adaptive extraction of irregular concave–convex defect features while effectively suppressing background noise. Second, to mitigate feature loss of micro-defects during multi-scale transformations, a Bidirectional Residual Multi-scale Feature Pyramid Network (BRM-FPN) is proposed. Utilizing bidirectional weighted paths and residual attention mechanisms, this network facilitates the efficient fusion of multi-view features, significantly enhancing the representation of small targets. Finally, the detection head is reconstructed based on grouped convolution strategies to design the Lightweight Grouped Convolution Head (LGC-Head), which substantially reduces parameter volume and computational complexity while maintaining feature discriminability. The validation results on the PKU-Market-PCB dataset demonstrate that MDEB-YOLO achieves a mean Average Precision (mAP) of 95.9%, an inference speed of 80.6 FPS, and a parameter count of merely 7.11 M. Compared to baseline models, the mAP is improved by 1.5%, while inference speed and parameter efficiency are optimized by 26.5% and 24.5%, respectively; notably, detection accuracy for challenging mouse bite and spur defects increased by 3.7% and 4.0%, respectively. The experimental results confirm that the proposed method outperforms state-of-the-art approaches in both detection accuracy and real-time performance, possessing significant value for industrial applications. Full article

Background/Objectives: Accurate sleep stage classification is essential for evaluating sleep quality and diagnosing sleep disorders. Despite recent advances in deep learning, existing models inadequately represent complex brain dynamics, particularly the time-lag effects inherent in neural signal propagation and regional variations in cortical activation patterns. Methods: We propose the MFST-GCN, a graph-based deep learning framework that models these neurobiological phenomena through three complementary modules. The Dynamic Dual-Scale Functional Connectivity Modeling (DDFCM) module constructs time-varying adjacency matrices using Pearson correlation across 1 s and 5 s windows, capturing both transient signal transmission and sustained connectivity states. This dual-scale approach reflects the biological reality that neural information propagates with measurable delays across brain regions. The Multi-Scale Morphological Feature Extraction Network (MMFEN) employs parallel convolutional branches with varying kernel sizes to extract frequency-specific features corresponding to different EEG rhythms, addressing regional heterogeneity in neural activation. The Adaptive Spatio-Temporal Graph Convolutional Network (ASTGCN) integrates spatial and temporal features through Chebyshev graph convolutions with attention mechanisms, encoding evolving functional dependencies across sleep stages. Results: Evaluation on ISRUC-S1 and ISRUC-S3 datasets demonstrates F1-scores of 0.823 and 0.835, respectively, outperforming state-of-the-art methods. Conclusions: Ablation studies confirm that explicit time-lag modeling contributes substantially to performance gains, particularly in discriminating transitional sleep stages. Full article

Identifying floating residual feed is a critical technology in recirculating aquaculture systems, aiding water-quality control and the development of intelligent feeding models. However, existing research is largely based on ideal indoor environments and lacks adaptability to complex outdoor scenarios. Moreover, current methods for this task often suffer from high computational costs, poor real-time performance, and limited recognition accuracy. To address these issues, this study first validates in outdoor aquaculture tanks that instance segmentation is more suitable than individual detection for handling clustered and adhesive feed residues. We therefore propose LMFF–YOLO, a lightweight multi-scale fusion feed segmentation model based on YOLOv8n-seg. This model achieves the first collaborative optimization of lightweight architecture and segmentation accuracy specifically tailored for outdoor residual feed segmentation tasks. To enhance recognition capability, we construct a network using a Context-Fusion Diffusion Pyramid Network (CFDPN) and a novel Multi-scale Feature Fusion Module (MFFM) to improve multi-scale and contextual feature capture, supplemented by an efficient local attention mechanism at the backbone’s end for refined local feature extraction. To reduce computational costs and improve real-time performance, the original C2f module is replaced with a C2f-Reparameterization vision block, and a shared-convolution local-focus lightweight segmentation head is designed. Experimental results show that LMFF–YOLO achieves an mAP50 of 87.1% (2.6% higher than YOLOv8n-seg), enabling more precise estimation of residual feed quantity. Coupled with a 19.1% and 20.0% reduction in parameters and FLOPs, this model provides a practical solution for real-time monitoring, supporting feed waste reduction and intelligent feeding strategies. Full article