Low-light image enhancement aims to recover clear, natural, and detail-preserving images from scenes with insufficient illumination, low contrast, noise degradation, and color distortion. Existing methods can improve image brightness to some extent, but they often suffer from noise amplification, unstable color restoration, texture loss, and over-enhancement in complex dark scenes. This is mainly because noise and textures may show similar high-frequency responses in the spatial domain, while luminance and chrominance information are highly coupled in the RGB space. To address these problems, this paper proposes SFSNet, a spatial-frequency synergistic network based on color space decomposition. The input image is first transformed into the HVI color space, where the intensity and chrominance components are separated to reduce the mutual interference between brightness recovery and color correction. Before spatial encoding, a frequency denoising module is introduced to suppress noise-related frequency disturbances and provide cleaner features for subsequent restoration. Then, a dual-branch backbone is constructed to process intensity and chrominance features separately, while a multi-scale convolution module is used to capture local textures, directional structures, and contextual information. To improve the consistency between brightness and color restoration, a cross-branch feature collaboration mechanism is designed to promote information interaction between the two branches. In addition, a deep interaction module and an uncertainty-based adaptive weighting strategy are employed to enhance deep feature representation and improve training stability. Experiments are conducted on LOL-v1, LOL-v2-real, and LOL-v2-syn, which provide paired normal-light reference images, and further evaluated on real unpaired datasets including DICM, LIME, MEF, and VV to verify the generalization ability of the proposed method. Compared with representative low-light enhancement methods, the proposed method can better improve dark-region brightness, suppress noise amplification, maintain natural color appearance, and preserve local structural details. Ablation studies further demonstrate the effectiveness of frequency denoising, multi-scale spatial modeling, cross-branch feature collaboration, and deep interaction supervision.

Vision-Language Models (VLMs) have demonstrated exceptional capabilities in cross-modal understanding by aligning Large Language Models (LLMs) with visual encoders. However, cross-modal hallucination remains a critical bottleneck that limits their deployment in accuracy-sensitive scenarios. The fundamental cause lies in the fact that the underlying LLM base, pre-trained on massive text corpora, establishes strong language priors that often override authentic visual features during autoregressive decoding. A detailed analysis of the internal attention mechanism during decoding reveals significant hierarchical disparities in visual feature evolution: shallow layers exhibit highly diffuse attention containing substantial redundant noise, which easily triggers attention drift, while deep layers suffer from severe attention decay due to language dominance, effectively severing the link between text generation and visual facts. Existing decoding-time intervention methods typically adopt global, coarse-grained strategies, failing to systematically address the contradiction between shallow-layer diffusion and deep-layer decay. Consequently, this paper proposes a Multi-layer Cumulative Visual Prior (MCVP) method, which reconstructs feature distributions at the decoding stage through three synergistic mechanisms. First, a sparsified focusing strategy is introduced to dynamically evaluate shallow-layer responses and filter background noise to extract high-confidence features, fundamentally preventing attention drift. Second, a layer-wise recursive accumulation mechanism is designed to construct a temporally robust global visual prior by applying weighted aggregation to preceding shallow-layer sparse features via a sliding window. Finally, an adaptive soft injection strategy is proposed to calculate the real-time visual dependency of the current generation step, dynamically incorporating the robust prior into deep-layer cross-attention modules to rebalance modal weights and enforce strict alignment with visual facts. Extensive empirical evaluations on LLaVA-1.5, MiniGPT4, and DeepseekVL, based on CHAIR and POPE benchmarks, demonstrate that MCVP significantly outperforms baseline sampling and state-of-the-art intervention methods. In the CHAIR benchmark, MCVP achieves a relative improvement of 18.8% in hallucination suppression performance; specifically, it reduces the sentence-level hallucination rate of LLaVA-1.5 from 48.4% to 38.2% while maintaining a high recall of 77.7%, successfully overcoming the tendency of traditional exploratory strategies to over-fit language probabilities and achieving an optimal balance between precision and coverage. In POPE testing, MCVP achieves a state-of-the-art F1 score of 88.45% on MiniGPT4 and effectively corrects the over-prediction issues caused by shallow-layer diffusion in DeepseekVL. Comprehensive ablation and visualization analyses confirm that these mechanisms effectively prevent deep-layer feature decay and refocus attention onto target regions. In summary, MCVP accurately captures the evolutionary patterns of hierarchical attention and provides an efficient, training-free approach to suppressing cross-modal hallucinations without compromising general reasoning capabilities, establishing a new theoretical and practical paradigm for white-box hierarchical intervention in large model optimization.

Brain tumor MRI images usually exhibit large variations in tumor scale, strong morphological heterogeneity, and blurred lesion boundaries that are often intertwined with normal brain tissues. Existing detection models still suffer from insufficient local detail extraction, global semantic modeling, and cross-scale feature fusion, which may lead to inaccurate boundary localization, increased missed and false detections, and high deployment costs. To address these issues, this paper proposes GDC-DETR, a high-accuracy and lightweight brain tumor detection model based on RTDETR-r18. The proposed model aims to improve the detection accuracy of complex lesions while reducing the number of parameters and computational cost, making it more suitable for clinical computer-aided diagnosis and lightweight deployment. GDC-DETR is improved from three aspects: backbone enhancement, feature fusion, and efficient feature interaction. First, a gated enhancement module with contextual awareness, named GLCA-CG, is designed. It extracts edge and texture details and long-range contextual information through local and global dual branches, respectively, and adaptively selects discriminative features using a convolutional gating mechanism. This enhances the model’s ability to distinguish small lesions, weak boundaries, and complex backgrounds. Second, a hierarchical local-global attention fusion module, named HLGA-FM, is constructed. By employing parallel patch-aware attention at local and global scales, this module achieves semantic alignment and adaptive fusion between high-level and low-level features, alleviating information dilution, channel redundancy, and semantic deviation caused by traditional feature concatenation, thereby improving cross-scale lesion localization. Finally, a low-resolution lightweight attention-based intra-scale feature interaction module, named LR-AIFI, is proposed. High-resolution features are first mapped into a low-resolution space through anti-aliasing downsampling for attention interaction, and then spatial details are compensated through upsampling and depthwise convolution refinement. As a result, the computational complexity of the main attention term is reduced to approximately 1/16 of the original strategy, achieving a better balance between global modeling capability and computational efficiency. Experiments are conducted on a public brain tumor MRI dataset containing three types of tumors: glioma, meningioma, and pituitary tumor. Under unified training parameters, evaluation metrics, and experimental environments, GDC-DETR is compared with RTDETR series models, YOLO series models, and several brain tumor detection methods. The experimental results show that GDC-DETR achieves an mAP50 of 92.9%, which is 2.7 percentage points higher than the RTDETR-r18 baseline. The precision and recall reach 91.4% and 91.1%, respectively. Meanwhile, the number of parameters is reduced to 15.6 M, representing a 21.6% decrease compared with the baseline, and the computational cost is 47.8 GFLOPs, maintaining a relatively low computational overhead among the mainstream comparison models. Visualization results indicate that, for challenging samples with blurred boundaries, irregular shapes, and high interference between lesions and surrounding tissues, GDC-DETR can more accurately focus on the tumor core region, reduce invalid responses in background areas, and obtain more stable bounding-box localization results. Ablation experiments demonstrate the effectiveness of each proposed module. Introducing GLCA-CG alone improves mAP50 by 1.2 percentage points and significantly reduces the parameter scale. Combining HLGA-FM with GLCA-CG further improves mAP50 to 92.4%. LR-AIFI reduces the computational burden while maintaining stable detection accuracy. The best overall performance is achieved when the three modules are jointly integrated, indicating that detail enhancement, multi-scale semantic fusion, and low-resolution efficient interaction are complementary to each other. Generalization experiments further show that the proposed model achieves mAP50 values of 94.9% and 92.9% on two external MRI datasets, respectively. In the zero-annotation transfer scenario, the mAP50 reaches 73.5%, and after fine-tuning with only 10% and 20% of target-domain annotations, it increases to 86.5% and 88.6%, respectively, demonstrating favorable cross-domain adaptability and few-shot transfer capability. In summary, GDC-DETR effectively alleviates the trade-off between high accuracy and low computational cost in brain tumor MRI detection, improves the sensitivity and localization ability for complex lesions, and reduces the dependence on hardware resources, providing a lightweight reference solution for clinical auxiliary brain tumor detection.

With the rapid development of Large Language Model (LLM), LLM has shown promising application prospects in text classification tasks. Existing text classification methods based on prompt engineering heavily rely on the generative capability of LLM, and often suffer from insufficient generation ability and poor reasoning stability in lightweight LLM scenarios, making it difficult to achieve satisfactory performance. In addition, different hidden layers of LLM exhibit significant differences in representation capability, and intermediate layer representations often contain richer semantic information, even outperforming the last layer outputs in certain tasks. To address these issues, a multi-level semantic fusion for text classification based on LLM is proposed. By extracting and fusing text embedding representations from different hidden layers of lightweight LLM, the proposed method achieves efficient and stable classification performance without fine-tuning LLM. Firstly, prompt engineering is introduced into the construction process of text embedding representations. By designing prompt templates suitable for text classification tasks, original sentences are inserted into specific contextual semantics to guide the model in adjusting the attention distribution when generating hidden layer representations, thereby enhancing the discriminative capability of text embedding representations. Secondly, curvature is employed to characterize the geometric smoothness of hidden layer representations. Hidden layer representations with lower curvature correspond to smoother semantic trajectories and are more suitable as text embedding representations for text classification tasks. By calculating the average curvature of each hidden layer, several hidden layer representations with lower curvature are selected as feature sources for subsequent fusion. Thirdly, intra-layer position-weighted pooling and inter-layer multi-scale convolution fusion are used to obtain more expressive sentence embeddings. The intra-layer position-weighted pooling assigns different weights according to the relative positions of tokens in the input sequence, preserving important semantic information at the tail while also considering contextual information at the beginning of the sequence, thereby generating more complete and stable text embedding representations. The inter-layer multi-scale convolution fusion models relationships among different hidden layers through convolution operations to achieve the fusion of multi-level semantic information. Finally, the classifier is trained using a joint loss function, simultaneously improving classification accuracy, the learning capability for hard samples, and the discriminative of the representation space, thereby optimizing model parameters from multiple perspectives to obtain the final prediction results. Experimental results demonstrate that the proposed method achieves accuracies of 96.11%, 93.48%, 89.92%, 74.15%, and 96.33% on the SST-2, MR, 20NG, Ohsumed, and R52 datasets, respectively, outperforming existing mainstream baseline methods. In addition, Macro-F1 is introduced as a supplementary evaluation metric on the relatively imbalanced Ohsumed and R52 datasets, where the proposed method achieves 62.93% and 73.12%, respectively, demonstrating its effectiveness under complex data distribution conditions. Furthermore, ablation studies, different intra-layer pooling methods analysis, different prompt templates analysis, and different numbers of layers analysis are conducted to verify the rationality and effectiveness of each module design. Sensitivity analysis of loss weights is performed to evaluate the influence of different loss weight variations on classification performance. Different layers performance analysis is performed to explore the relationship between different hidden layer representations and classification performance. In conclusion, the proposed method can effectively extract and fuse LLM embedding representations for text classification, significantly reducing computational overhead while improving the representation quality and discriminative capability, demonstrating strong generalizability and practical applicability.

To meet the demand for sharing massive and continuous time-series data in Internet of Things (IoT) scenarios, traditional certificateless signcryption (CLSC) schemes based on bilinear pairings or elliptic curve cryptography face security limitations under the threat of quantum computing. Moreover, repeated signcryption for multiple data segments incurs considerable computational and communication overhead. To address these challenges, this paper proposes a lightweight post-quantum certificateless hybrid signcryption scheme for time-series data sharing. First, a bidirectional hash-chain-driven symmetric key generation and management mechanism is designed to enable rapid key derivation for multiple consecutive data segments while providing forward and backward security. In addition, a lightweight hybrid signcryption framework with conditional privacy protection is constructed over module lattices. By further introducing an offline/online paradigm that shifts part of the computation to the precomputation stage, the proposed scheme achieves post-quantum security while significantly reducing both the computational burden on end devices and the transmission overhead over communication links. Security analysis in the random oracle model demonstrates that the scheme satisfies IND-CCA confidentiality and EUF-CMA unforgeability. Prototype implementation and evaluation are further conducted on resource-constrained devices. Comparative experimental results show that, compared with representative schemes, the proposed scheme offers significant advantages, reducing the overall end-side latency and transmission cost to about 40% of those of comparable post-quantum schemes on average. In long-duration and continuous time-series data scenarios, these advantages become even more pronounced. The proposed scheme strikes a balance among post-quantum security, lightweight design, and scalability, making it well suited for time-series data sharing applications in next-generation intelligent IoT systems.

In scenarios related to human-agent collaboration, the explicit/legible representation and accurate recognition of agent intent serve as vital means to enhance collaborative safety and efficiency, and constitute the foundation for establishing and improving human trust in agents. The core lies in enhancing the legibility of agent behaviors to facilitate effective human understanding. Existing research primarily conveys intent by introducing differentiated behaviors that deviate from the optimal path. While effective in certain contexts, such schemes inherently sacrifice partial execution efficiency. In highly confusing scenarios, they face two critical challenges: 1) In environments characterized by overlapping and dense routes, sacrificing efficiency yields diminishing returns in intent conveyance; 2) Overlapping routes obscure the informational value of the agents' original behavioral sequences, making it difficult for observers to discern their true intentions. Addressing these issues, this study investigates key factors influencing behavioral legibility in complex human collaboration, focusing on action posture, velocity, and distance. Without altering the agents' original action sequences, this paper proposes a multi-dimensional fusion representation mechanism through the multi-dimensional modeling and integration of key state information, including posture, velocity, and distance. Specifically, we design a posture mapping function aligned with human reasoning, utilize Gaussian kernel functions to transform and map velocity and distance representations, and integrate these features via fusion parameters to enhance behavioral legibility. Furthermore, corresponding to this mechanism, we propose an efficient intent recognition scheme based on Bayesian update. By fully leveraging the fused state information, the scheme achieves rapid and accurate identification of agent intent. Consequently, the proposed method—which integrates the mechanism and the scheme—not only enhances the explicit expressive capability of overall behaviors regarding their intent but also improves task execution efficiency by reducing the number of differentiated behaviors required. The method is comprehensively validated across three experimental scenarios: unconstrained, complex obstacle-laden, and highly confusing environments. Experimental results demonstrate that, compared to baseline methods, our approach incorporates additional dimensions of state information, improving the accuracy of agent intent recognition by an average of 44.47%. Moreover, by conveying sufficient intent information through shorter behavioral sequences, the method significantly boosts task execution efficiency, reducing agent execution costs by an average of 42.86%. Ablation studies verify the positive contributions of posture, velocity, and distance to legibility enhancement, with posture identified as the most significant factor. Sensitivity analysis determines the optimal value ranges for color and velocity factor parameters, ensuring stable performance. Additionally, experiments on continuous control tasks confirm the method's applicability in complex engines and real-world engineering. On the algorithmic side, complexity analysis reveals that the inference time increases by only 2.89 milliseconds compared to the base method, satisfying real-time interaction requirements. This study provides a novel pathway for reducing redundant behaviors while enhancing legibility, balancing belief transmission and execution efficiency without disrupting original action sequences. The method can be directly embedded into various baseline approaches to achieve varying degrees of improvement in recognition accuracy. Future wor

Efficient cross-domain resource allocation is the foundation for achieving optimized resource distribution and ensuring service quality in the Space-Air-Ground Integrated Network (SAGIN). However, constrained by the inherent multi-layer highly complex topology of SAGIN and a massive number of differentiated network slicing service requests, traditional flat architectures and static mathematical optimization models struggle to achieve efficient and dynamic cross-domain resource allocation in the face of dynamic and complex problems. Therefore, this paper proposes a hierarchical cross-domain resource allocation method based on deep reinforcement learning (DLCDS), aiming to maximize system utility while ensuring the efficient utilization of underlying physical resources. Firstly, this paper designs a hierarchical collaborative cross-domain framework, establishing a dual-layer collaborative mechanism between the global domain layer and the local domain layer. Secondly, at the global domain layer, a service level agreement (SLA) decomposition algorithm based on the Twin Delayed Deep Deterministic Policy Gradient is proposed. This algorithm leverages deterministic policies within continuous action spaces to achieve the planning and decomposition of SLA constraints across subdomains. At the local domain level, a resource allocation algorithm based on Proximal Policy Optimization is designed. This algorithm constructs an SLA mapping model to achieve the conversion from service objectives to resource requirements, and employs the exploratory advantage of stochastic policies to enhance resource allocation efficiency within the domain. Finally, experimental results demonstrate that, compared to other benchmark algorithms, this method achieves an average 12% improvement in resource utilization and a 19% increase in system utility.

To address the problems of low detection accuracy and high missed detection rate caused by the small size,weak texture features,and severe background interference of unmanned aerial vehicle (UAV) targets in low-altitude complex environments,a lightweight UAV object detection method named EF-YOLO (Efficient Fusion YOLO) based on YOLOv11 is proposed. Considering the limitations of existing models in complex scenarios,including insufficient deep semantic representation, multi-scale feature fusion mismatch, and serious background interference, the YOLOv11 network architecture is systematically optimized. First, a C3K2-MSFE module is introduced into the deep layers of the backbone network. By combining grouped convolution with multi-scale convolution kernels, the module enhances the discriminative ability of weak deep target features while maintaining the stability of the main semantic flow through a linear branch, thereby improving the feature representation capability of small UAV targets in complex backgrounds. Second, an improved multi-scale feature fusion structure named PSG-FPN is proposed. A Soft Nearest Neighbor Interpolation mechanism is adopted during the upsampling stage to alleviate semantic coverage problems in cross-scale feature fusion through adaptive soft coefficients. Meanwhile,GSConvE is employed to optimize the downsampling process,and an additional P2 detection layer is introduced to enhance the perception ability for small-scale UAV targets, thereby improving multi-scale feature alignment. Furthermore, to reduce parameter redundancy and computational overhead in traditional detection heads, a lightweight shared-feature detection head named SFE-Head is designed. By adopting a shared convolutional stem and grouped convolution structure, classification and regression branches can share features,which significantly reduces model parameters and computational complexity while maintaining detection accuracy. Finally,the original C2PSA module in YOLOv11 is improved by introducing a frequency-domain enhanced feedforward structure. Through frequency-domain feature modeling, redundant texture noise in complex backgrounds can be effectively suppressed, thereby improving the response capability and anti-interference performance for weak UAV targets. Experiments are conducted on the DUT Anti-UAV and Det-Fly datasets. The results show that, on the DUT Anti-UAV dataset, EF-YOLO achieves Precision, Recall, mAP50, and mAP50-95 values of 96.76%, 88.16%, 92.59%, and 62.41%, respectively, representing improvements of 4.01%, 3.44%, 3.04%, and 6.18% compared with the original YOLOv11. Meanwhile, the number of model parameters is reduced from 2.58M to 2.27M, and the model size decreases from 5.3 MB to 4.8 MB, while maintaining favorable real-time detection performance under relatively low computational complexity. On the Det-Fly dataset, EF-YOLO achieves mAP50 and mAP50-95 values of 96.4% and 68.06%, respectively, which are 2.39% and 2.16% higher than those of YOLOv11, demonstrating excellent generalization ability and stability under different low-altitude complex scenarios. In addition, ablation experiments verify that the proposed modules provide significant collaborative benefits. Specifically, the C3K2-MSFE module effectively enhances weak deep feature representation, PSG-FPN improves cross-scale semantic alignment, SFE-Head reduces model complexity while maintaining detection accuracy, and the frequency-domain enhancement structure further suppresses complex texture noise. Compared with conventional object detection models, EF-YOLO exhibits more stable detection performance in complex backgrounds, long-distance target scenarios, and low-contrast environments, effectively reducing missed detections and false detections while preserving model lightweight characteristics. Overall, EF-YOLO achieves a favorable balance among detection accuracy, lightweight design, and real-time performance, showing strong engineering deployment potential and providing reliable technical support for low-altitude airspace supervision, edge intelligent monitoring, and UAV security applications.

This paper studies operator-level reconstruction for the structural mismatch between YOLOv11 and Ascend NPUs. The SPPF, C3K2, and C2PSA modules are optimized without changing network semantics or model scale. Three Ascend C operators are designed: the SPPF operator uses an on-chip data loop and halo cache to reduce redundant global-memory traffic in multi-stage pooling; the C3K2 operator integrates multi-core task assignment and multi-queue asynchronous pipelining to reduce fine-grained kernel launch overhead; and the C2PSA operator reconstructs attention communication through a parallel reduction-broadcast primitive. On an Ascend 910B NPU, the complete reconstruction reduces the COCO training time per epoch by 27.6%. The results show that matching Ascend on-chip memory, asynchronous queues, and multi-core synchronization mechanisms improves the training execution efficiency of key YOLOv11 modules and keeps inference performance stable.

Reinforcement Learning from Human Feedback (RLHF) is a key technology for achieving professional value alignment in large legal language models. However, existing reviews lack systematic analysis of the specific adaptation mechanisms and challenges of legal scenarios. This paper aims to fill this gap, systematically construct a panoramic view of the research and application of RLHF in legal large language models, and provide a reference for method selection in engineering practice. This article takes the six-dimensional framework of "technical foundation - algorithm evolution - task migration - evaluation system - application challenges - future path" as the main analytical thread, systematically reviewing 80 core papers from top conferences and journals such as NeurIPS, ICML, ICLR, and ACL from 2017 to 2025. Analyze the complete evolution logic of RLHF in the legal field from algorithm adaptation to evaluation system. Analysis shows that the core predicament of legal RLHF lies in the inaccurate reward modeling caused by the conflict of multi-dimensional rewards and the sparsity of expert preferences. The PPO algorithm has insufficient stability in the generation of long text laws, while new algorithms such as DPO and GRPO simplify the process but face the bottleneck of preference data quality. Further research reveals that different legal tasks vary significantly in terms of risk levels, reasoning complexity, and feedback acquisition costs. Therefore, there is no unified optimal solution for the RLHF algorithm. For structured tasks such as the generation of legal documents and summaries of regulations, DPO has good applicability due to its advantages such as a simple training process and relatively high stability. For tasks such as legal reasoning and case analysis that require multi-path argumentation, GRPO uses intra-group relative ordering to construct learning signals and shows great potential in handling complex legal arguments. In high-risk scenarios such as judicial auxiliary decision-making and sentencing recommendations, it is necessary to combine expert feedback, rule constraints, and knowledge retrieval mechanisms to achieve reliable alignment. The evaluation system is evolving from automated metrics to process interpretability and bias auditing. RLHF provides a key evolutionary path for legal large language models from usability to credibility. However, its implementation relies on endogenous rewards and hybrid feedback mechanisms to reduce annotation costs, and the development of reward traceability and process evaluation to meet judicial transparency requirements. Meanwhile, with the development of new methods such as direct preference optimization and group relative strategy optimization, legal RLHF is gradually evolving from the traditional reward model-driven mode to lightweight preference optimization and relative ranking optimization, providing a new technical path for reducing training complexity and enhancing the adaptability to legal scenarios. Existing research indicates that relying solely on a single reward model is insufficient to fully depict the complex trade-off relationships within the legal value system. In the future, it is necessary to further integrate rule knowledge, expert experience, and process feedback information to construct a more stable and interpretable legal alignment mechanism. This paper not only constructs a six-dimensional analysis framework for legal RLHF research, but also further proposes a legal task-oriented RLHF adaptability analysis system, systematically comparing the applicable boundaries and failure mechanisms of algorithms such as PPO, DPO, and GRPO in different tasks such as legal document generation, legal reasoning, and compliance review. In addition, this paper conducts a horizontal comparison of different RLHF methods by combining the risk levels of legal tasks and the characteristics of algorithms, providing a reference basis for algorithm selection and engineering deployment in legal intelligent systems. Research shows that the core contradiction of legal RLHF is not merely the insufficiency of algorithm performance, but rather a structural contradiction among multi-dimensional reward conflicts in legal scenarios, inconsistent expert preferences, and the difficulty in quantifying judicial value. This paper, for the first time, comprehensively reveals the entire process logic of legal RLHF from algorithm adaptation to engineering implementation from four aspects: technological evolution, reward modeling, evaluation system, and judicial governance, providing a systematic research framework and practical guidance for the trustworthy construction of legal large language models.

Globally escalating marine ecological pollution necessitates efficient and precise automated intelligent monitoring and cleaning technologies for underwater litter leveraging autonomous underwater vehicles (AUVs) or underwater robots, which has become an urgent academic consensus in the field of marine environmental protection. To address the inherent physical degradation of underwater visible light imaging—such as color casts, blurring, and low contrast—as well as the bottlenecks of cross-domain heterogeneity caused by varying optical properties of water bodies across different sea areas and the severe scarcity of high-quality fine-grained annotated samples, this study aims to overcome the feature shift and negative transfer phenomena induced by deploying models in unknown water environments. To this end, it proposes an underwater litter object detection framework that merges pixel-level image enhancement with a two-stage robust cross-domain adaptive learning scheme. This framework adopts a design strategy of "step-by-step construction and dynamic alignment," decoupling the pipeline construction into two mutually collaborative training stages. The first stage aims to construct an interference-resistant baseline perception architecture, establishing a solid feature foundation of deep representations for the entire network. To maintain high computational efficiency while circumventing the limitations of complex deep generative networks—which easily introduce spurious edges and grid-like artifacts under small-sample conditions—a low-cost pixel-level physical alignment pipeline is implemented at the data loading layer. Specifically, the input blue-green (BGR) raw images are first transformed into the LAB color space, where the L channel is extracted and subjected to adaptive pixel-level linear stretching to maximize the restoration of edge contrast and fine details. Subsequently, a Channel Stabilization Module compensates for color casts in the A and B channels to robustly suppress underwater-specific cyan-blue artifacts, before inversely mapping the calibrated feature maps back to the standard red-green-blue (RGB) space. On this basis, 30,000 iterations of supervised perceptual training are executed on the source dataset (TrashCan Instance), enabling the RetinaNet detector to acquire high-quality low-level general geometric textures and high-level robust semantic features, thereby establishing a solid perception baseline for multi-source scenarios. The workflow then naturally proceeds to the second stage, which focuses on domain alignment and precision enhancement for cross-domain deployment. To eliminate deep feature shift and response distortion triggered by unknown new water domains, a lightweight Feature Calibration Module (FCM) is symmetrically injected after the multi-scale output layers of the pre-trained backbone network. A differential initial assignment based on the Kaiming normal distribution is directly applied to the internal parameters of the FCM at the very beginning of the fine-tuning process to explicitly induce adaptive dynamic calibration. During the subsequent 10,000 iterations of cross-domain adaptive fine-tuning, to avoid catastrophic forgetting under low-data regimes, this stage completely freezes the feature extractor parameters of the backbone network, serving as a locked low-level feature stabilizer. Concurrently, a differential asymmetrical dual-learning-rate mechanism is deployed, assigning high gradient sensitivity (initial learning rate 0.01) to the newly injected FCM while allocating a smaller step size (initial learning rate 0.002) to the top-level Feature Pyramid Network (FPN) and detection head. Furthermore, an Exponential Moving Average (EMA) technique with a fixed decay rate (λ = 0.9999) is introduced throughout the entire fine-tuning process. It automatically switches to the smoothed EMA weights during the single complete evaluation performed every 1,000 iterations, leveraging a smooth weight loss plane to shield the model from training noise brought by extremely small samples and ensuring a lossless migration of feature space alignment. To guarantee the scientific rigor of the experimental design and its practical deployment reference value, this study introduces completely independent third-party cross-domain underwater datasets for blind generalization validation. Quantitative results demonstrate that while maintaining a lightweight and compact architecture with only 55.86 million parameters, the proposed method achieves a real-time inference speed of up to 44.61 FPS on a single NVIDIA RTX 4080 SUPER GPU, showcasing immense potential for edge-side practical deployment. In in-domain closed-loop testing, the framework significantly improves the bounding box Average Precision (AP) by 5.33 percentage points compared to the traditional RetinaNet baseline model, with a notable jump of 8.97 percentage points in the highly challenging task of detecting small targets characterized by blurred edges and low visibility underwater. In the most rigorous cross-water and cross-dataset generalization blind tests, the proposed method demonstrates remarkable performance gains: on the completely unseen independent J_Litter dataset, the AP_50 of the baseline model is as low as 0.31% due to multi-source interference, whereas the proposed method catapults the AP_50 to 28.66% through dynamic feature calibration; on the DeepPlastic deep underwater dataset, the method successfully elevates the baseline AP_50 from 3.30% to a stable 12.81%. These outcomes robustly validate the scientific validity and synergetic efficiency of the "physical foundation construction" and "dynamic feature calibration" within the two-stage architecture, proving its capability to effectively overcome negative transfer in cross-water detection and highlighting its strong practical generalization potential for real-world engineering deployment in automated marine litter removal.

As a core research task within the fields of artificial intelligence and multimodal understanding, Visual Question Answering (VQA) requires models to jointly comprehend the visual content of images and natural language questions in order to generate accurate answers. Addressing the limitations of existing VQA methods—which typically lack large-scale pre-training and suffer from interference by weakly correlated background elements and redundant regions during multimodal feature modeling—this paper proposes a multi-module collaborative VQA method based on multimodal filtering and spatial-channel attention. Specifically, the proposed method first introduces a Multimodal Filtering Module. This module extracts question feature vectors from specific positions to serve as a global semantic representation; it then performs a Hadamard product interaction with aligned visual features and utilizes a perceptron combined with an activation function to generate normalized filtering gate weights. By element-wise scaling the image features prior to deep-level interaction, this process explicitly suppresses redundant background information, thereby providing a "cleaner" visual input for subsequent deep interactions. Subsequently, a Spatial-Channel Self-Attention Unit is introduced to replace the traditional self-attention architecture. Within a six-layer stacked encoder-decoder framework, this unit reshapes features into 2D maps along the spatial dimension; by combining multi-scale convolutional kernels with global pooling, it generates a spatial attention distribution that effectively captures spatial dependencies between different image regions. Simultaneously, along the channel dimension, it employs pooling and fully connected networks to dynamically adjust the semantic importance of individual channels. Finally, a gating mechanism is utilized to perform a weighted fusion of the spatial-channel-modulated local features with the global context captured by multi-head self-attention, thereby effectively enhancing the discriminative power of the visual features. Building upon this foundation, a Dynamic Modality Fusion Module is constructed to move beyond static fusion mechanisms. After multiple layers of interaction, the model generates attention maps for both visual and textual features, concatenates them, and feeds them into a fully connected classification network (incorporating dropout) to adaptively learn 2D dynamic fusion weight factors. This process ultimately achieves a dynamic weighted integration of multimodal features, with the overall model parameters optimized using a binary cross-entropy loss function. Extensive quantitative experiments conducted on the VQA-v2 and GQA datasets demonstrate that the proposed method achieves significant performance improvements, even without relying on massive-scale cross-modal pre-training data. Specifically, on the VQA-v2 test-dev split, the proposed method achieved an overall accuracy of 71.85%, representing a significant improvement of 1.22% and 2.33% over comparable baseline models—MCAN and BAN—respectively. Notably, the accuracy rates for binary "Yes/No" questions and "Other" question types reached as high as 88.03% and 62.39%, respectively. On the GQA test set—which emphasizes compositional reasoning and relational understanding—the model achieved an overall accuracy of 57.92%, with binary "Yes/No" questions attaining an accuracy of 76.68%. Furthermore, rigorous ablation studies quantitatively validated the efficacy of each core module; the introduction of the multi-modal filtering, spatial-channel attention, and dynamic modality fusion modules yielded absolute performance gains of 0.63%, 0.69%, and 0.52%, respectively. Additionally, an analysis of the impact of model depth confirmed that feature interactions are most comprehensive—and overall model performance reaches its peak—when both the encoder and decoder consist of six stacked layers. In summary, by employing fine-grained feature modeling and dynamic interactions, the proposed method successfully mitigates issues such as severe noise interference, isolated spatial-channel processing, and the poor adaptability of static fusion mechanisms, thereby offering an improved solution for visual-language multi-modal understanding and deep feature representation within specific vertical domains characterized by limited computational resources.

Generative face-swapping technology poses a severe threat to personal privacy. Existing proactive defense methods suffer from poor generalization in black-box scenarios and insufficient robustness under image processing distortions. To address these issues, this paper proposes a Robust Black-box Proactive Defense (RBPD) framework. The framework aims to inject imperceptible adversarial perturbations into the source image to disrupt the identity feature extraction process of unknown face-swapping models, thereby achieving reliable identity protection.The framework employs a two-stage generation mechanism. In the first stage, a Semantic-Aware Encoder (SAE) and a Texture-Guided Decoder (TGD) are utilized to generate initial semantic perturbations. The SAE uses facial semantic masks for guidance and incorporates a Convolutional Block Attention Module to accurately focus on identity-critical regions. The TGD fuses shallow texture features through skip connections, constraining the perturbations to adapt to the local texture distribution and gradient intensity of the source image. This ensures effective attack performance while significantly reducing visual artifacts and improving visual quality.In the second stage, a Dual-Stream Fusion Encoder (DFE) and a Multi-Scale Aggregation Decoder (MAD) are introduced. The DFE extracts deep features from both the source image and initial perturbations and performs nonlinear fusion, deeply embedding adversarial information into the image’s semantic feature space. The MAD employs three parallel dilated convolutions to capture multi-scale contextual information and integrates a Squeeze-and-Excitation module for adaptive channel recalibration, thereby enhancing the robustness of the perturbations against image distortions. Additionally, a Meta-learning Adaptive Attack (MAA) strategy is designed, which integrates gradient feedback from four heterogeneous identity feature extractors (ArcFace, FaceNet, MagFace, and AdaFace), dynamically adjusts optimization weights, and achieves precise breakthroughs against strongly robust extractors, thereby improving the cross-model generalization of the perturbations on unknown black-box models.Evaluations on the CelebA-HQ and RaFD datasets, targeting three mainstream face-swapping models (SimSwap, E4S, and DiffSwap) as well as commercial facial recognition APIs from Baidu and Tencent, yield the following results: the protected images achieve average Top-1 and Top-5 identity matching rates of 0.311 and 0.396, respectively, representing decreases of 63.28% and 54.79% compared to unprotected images, and further reductions of 16.17% and 19.02% compared to the best baseline methods. On the unseen RaFD dataset, the method maintains stable performance with average Top-1 and Top-5 matching rates of 0.349 and 0.394. In black-box face-swapping tests, the cosine similarity between swapped images and source images mostly drops below 0.3, with the lowest reaching 0.185, achieving reliable identity mismatch. When facing common social network distortions such as JPEG compression, Gaussian blur, noise, and resizing, the Distortion Defense Volatility (DDV) averages only 3.4%, significantly outperforming the baseline methods Saliency (55.5%), DF-RAP (27.1%), NullSwap (19.70%), and ID-Eraser (13.90%). In terms of visual quality, the protected images attain a PSNR of 37.38 dB, SSIM of 0.976, and LPIPS of 0.0065, demonstrating good perceptual naturalness. In commercial API tests, the face matching pass rate on Baidu API decreases from 89.92% to 0.40%, and on Tencent API from 90.37% to 4.25%. The total processing time for a single image is only 8.23 ms, indicating strong practical deployment potential.The proposed RBPD framework effectively addresses the deficiencies of existing methods in black-box generalization and distortion robustness. Through the synergistic design of semantic-texture guidance and multi-scale deep fusion, it achieves a strong balance among attack effectiveness, visual quality, and robustness, providing an efficient and practical proactive defense solution for personal privacy protection in complex social network environments. This work holds significant theoretical importance and application value.

Question answering models aim to enable machines to understand natural language questions and automatically reason over text or knowledge bases to generate accurate answers. Existing retrieval-augmented generation (RAG) based QA methods improve generation quality by incorporating external knowledge, yet they still face two challenges: first, insufficient semantic understanding makes it difficult to capture the relationship between the question and the retrieved knowledge; second, the retrieval stage inevitably introduces noisy information with low relevance to the question, which is then amplified during the generation process, easily leading to answer deviation or hallucination. To alleviate these issues, this paper proposes a Semantic Retrieval-Augmented Generation Question Answering model based on a two-stage filtering mechanism, named BiS-RQA. By constructing a progressive framework of explicit and implicit filtering, the model improves answer accuracy in QA scenarios while maintaining efficient inference. The model adopts a two-stage progressive filtering architecture: the first stage filters out obviously irrelevant knowledge entities, and the second stage performs noise suppression on the retained candidate knowledge, allowing the large model to focus on the most valuable pieces of information. The first stage is explicit filtering, which introduces a Semantic-guided Personalized PageRank algorithm (SPPR). In implementation, for an input natural language question, a pre-trained language model is used to extract the question embedding vector; entities and relations in the knowledge graph also obtain their respective embedding vectors through the same model. First, the similarity between the question vector and each entity vector is computed and used as a personalization vector to guide the initial preferences in a random walk with restart. Then, the similarity between the question vector and each relation vector is computed, and this similarity is employed to adjust the transition matrix from the current node to its neighbors during the walk, i.e., the jump probabilities are modified according to the question-relation similarity. On this basis, the algorithm iterates until convergence, yielding stationary distribution scores for all nodes. Finally, the nodes are sorted in descending order of scores, the top-K entities and their associated relations are selected, and a corresponding subgraph is constructed. This process explicitly filters out irrelevant nodes and edges from the original knowledge graph, thereby effectively improving the relevance of the retrieved information. The second stage is implicit filtering, which proposes a Multi-Attention Information Fusion method (MAIF). The MAIF method utilizes self-attention and cross-attention mechanisms to weight entities and relations, respectively. First, entities and relations are pooled to serve as the subgraph embedding; subsequently, the entities, relations, and subgraph pass through a self-attention module to capture internal dependencies, and then cross-attention is further performed to allow them to interact with each other to obtain the correlation degree between information at different granularities, thus achieving implicit filtering and weighting of the candidate information. Finally, the weights output by the cross-attention mechanism serve as soft prompts, while the entities and relations obtained from the first stage serve as hard prompts, and together with the question they are fed into the large language model for answer generation. On the dataset WebQSP, BiS-RQA achieves an accuracy of 77.21%, a relative improvement of 4.6% over the best baseline model G-Retriever (73.79%); on the dataset CWQ, BiS-RQA achieves an F1 score of 35.38%, a relative improvement of 8.1% over the second-best baseline (32.70%). In terms of memory consumption, BiS-RQA uses 31.5 GB of memory, an increase of 1.5% compared to the baseline G-Retriever; in terms of inference time, BiS-RQA requires 0.9427 seconds, an increase of 56% compared to GRAG. Experimental results on multiple public QA datasets validate the effectiveness of the proposed method in semantic retrieval and answer generation, and demonstrate strong competitive advantages.

While punctuation restoration of standardized text has made significant progress, real-time punctuation restoration within Automatic Speech Recognition (ASR) systems remains a challenging task. The core challenge lies in the colloquial nature of ASR transcripts and the severe imbalance of punctuation categories. Moreover, despite strong semantic understanding capabilities, large language models (LLMs) suffer from high inference latency and deployment costs. To alleviate these problems, this paper proposes a real-time punctuation restoration method based on progressive iterative optimization. First, multiple public Chinese corpora are integrated to construct a large-scale training dataset covering diverse text domains. In addition, a punctuation-aware data weighting strategy is introduced to mitigate the learning bias caused by imbalanced punctuation distributions. By assigning different weights to training samples, the proposed strategy improves the alignment between the training corpus and real-world ASR outputs. Then, by introducing dynamic masking to restrict the attention range, only the context information within a limited window near the current position is retained, ensuring the ability to obtain key semantic information and effectively reducing long-distance noise interference, thereby achieving a balance between accuracy and latency. Finally, a progressive iterative optimization mechanism is introduced. The sliding-window average loss is used to estimate sample difficulty, and a dynamic weighting strategy is designed accordingly. For difficult samples that continuously generate high prediction losses, a multi-sentence concatenation method is used to generate new samples with more complex context dependencies, and their training weights are increased, enabling the model to gradually focus on complex semantic boundaries, achieving the collaborative evolution of training data and model capabilities, and thereby continuously improving the model's ability to capture complex semantic structures. This paper conducts systematic experiments on the real punctuation restoration test set and compares it with representative models such as CT-transformer, Qwen2.5-7B, Llama3.1-8B, Gemma2-9B, and DeepSeek-V3. The experimental results show that the F1 value of this method is 4.92 percentage points higher than that of the traditional real-time punctuation restoration baseline CT-transformer model. Moreover, when compared with large language models with parameter scales much larger than this model, this method still achieves better restoration performance. In terms of inference efficiency, the average inference time per sentence of the model is only 26 ms, which is nearly 200 times faster than DeepSeek-V3, meeting the requirements of low latency response in real-time speech processing scenarios. Ablation experiments further verify the effectiveness of the controllable delay mechanism, data weighting strategy, and progressive iterative optimization module in improving model performance. Among them, the iterative optimization process gradually increases the F1 value from 38.78% to 43.70%. Compared with general large models relying on large-scale parameters, the data and model collaborative optimization scheme designed based on task characteristics can achieve better punctuation restoration effects while maintaining low resource consumption and low-latency inference, demonstrating strong practical value for real-time ASR post-processing applications.

To address the issues of severe background interference, large variations in defect target scales, inconspicuous features of small-scale defects, and progressive loss of edge information in deep networks during insulator defect detection in UAV inspection scenarios, and given the favorable balance between real-time performance and parameter count of YOLOv11n, this paper proposes an improved insulator defect detection method based on YOLOv11n to enhance detection accuracy and robustness under complex background conditions. This method performs collaborative optimization of the network structure from three aspects, namely feature enhancement, feature fusion, and feature extraction, thereby improving the model's perception and recognition capabilities for insulator defect targets in complex scenes while maintaining lightweight characteristics.The proposed model incorporates three core optimization modules. First, to alleviate complex background interference and enhance fine-grained defect representation capability, a Dynamic Dual-Domain Feature Enhancement Module (DDFEM) is designed in the feature extraction stage of the backbone network, establishing a collaborative working mechanism between a global semantic branch and a local detail branch. The global branch establishes long-range spatial dependencies through bidirectional global pooling and cross-dimensional matrix interaction, achieving global semantic modeling with low computational complexity. The local branch extracts fine-grained texture features such as flashover traces and damage using multi-branch depthwise separable convolutions, combined with a dynamic attention fusion mechanism for adaptive recalibration of local features, thereby achieving complementary enhancement of global semantic information and local detail information. Second, to address the progressive weakening of edge detail information during deep semantic feature enhancement, a Sobel-Edge-Guided Weighted Fusion Module (SEGWF) is designed. This module explicitly extracts edge information from shallow features using the Sobel operator and dynamically fuses edge structural information with deep semantic features through a channel-wise weighted fusion mechanism, thereby enhancing the model's perception of key structural features such as damaged edges and contours of insulators, and improving small target detection performance under complex backgrounds. Finally, to enhance multi-scale target feature extraction capability and address the fixed receptive field of traditional convolutions that struggles to adapt to defects of different scales, a Receptive-Field Attention Convolution (RFAConv) mechanism is introduced into the backbone network to replace traditional convolutional structures. This mechanism adaptively adjusts the receptive field response range according to the feature distribution of different spatial regions, thereby improving the model's feature extraction capability for insulator defects at different scales and enhancing multi-scale target detection performance.To verify the effectiveness of the proposed method, experimental studies are conducted on a self-constructed insulator defect dataset, with comparisons against mainstream object detection models. Experimental results show that the proposed method achieves 92.2% mAP@0.5, representing a 4.9 percentage point improvement over the original YOLOv11n.Precision increases from 89.7% to 93.0% with a 3.3 percentage point improvement, while Recall increases from 79.6% to 88.3% with an 8.7 percentage point improvement, demonstrating that the proposed method effectively reduces the miss detection rate and improves defect detection accuracy in complex scenes. Furthermore, to validate the generalization capability of the model, tests are conducted on the public insulator defect dataset IDID. Experimental results show that the proposed method achieves a Precision of 89.6%, a Recall of 79.8% and an mAP@0.5 of 88.9%. It obtains the optimal mAP@0.5 among all comparative models, with an improvement of 1.9 percentage points over the original YOLOv11n, which verifies its strong cross-dataset generalization ability and stable detection performance. Meanwhile, the proposed model only has 3.12M parameters and a computational cost of 7.6 GFLOPs, maintaining low computational complexity while boosting detection accuracy. In conclusion, the improved YOLOv11n algorithm proposed in this paper can effectively complete the task of insulator defect detection under complex backgrounds.

Multi-view subspace clustering aims to learn latent shared structures by exploiting the consistency, complementarity, and discrepancy information among multiple feature sources. How to simultaneously preserve global structural relationships, local geometric characteristics, and representation stability during feature fusion remains a critical issue in this field. Existing methods are mainly built upon low-rank representation, sparse constraints, or graph learning mechanisms. Although they can exploit cross-view shared information to some extent, they still lack sufficient collaborative modeling of global structure preservation, local smoothness constraints, and robust representation learning, making them prone to local structure distortion, unstable representations, and degraded clustering performance under noise contamination, outlier perturbations, and complex data distributions. To address these issues, this paper proposes a Structured Local Smooth Multi-View Subspace Clustering (SLS-MVSC) method. Under a unified self-representation learning framework, the proposed method jointly incorporates low-rank constraints, Total Variation (TV) regularization, graph regularization, and robust error modeling to achieve collaborative optimization of global information learning, local relationship preservation, and stable representation learning. Specifically, low-rank constraints are employed to learn shared representations across multiple views, thereby capturing the latent consistent global subspace structure among different views and enhancing the modeling capability for cross-view common information. Considering that noise interference may cause severe fluctuations in the self-representation matrix, TV regularization is introduced into the self-representation learning process to preserve representation continuity and boundary structures by restricting abrupt local variations, thereby improving the smoothness and stability of learned representations. Furthermore, graph regularization is adopted to maintain neighborhood relationship consistency among samples. By explicitly modeling local manifold structures, the proposed method strengthens the characterization of data geometric relationships and enables the learned representations to better conform to the intrinsic distribution characteristics of the original data. Meanwhile, a robust norm is introduced into the reconstruction error term to enhance the model’s adaptability to outliers and complex noise, thereby improving representation quality and clustering robustness from multiple perspectives. To optimize the proposed model, an iterative algorithm based on the Alternating Direction Method of Multipliers (ADMM) is developed. By introducing auxiliary variables, the complex objective function is decomposed into several independently solvable subproblems, and the update rules for all variables together with the overall optimization procedure are derived, ensuring training stability and computational efficiency. Extensive experiments are conducted on six publicly available datasets to evaluate the effectiveness of the proposed method and compare it with several representative multi-view clustering approaches. Experimental results demonstrate that the proposed method achieves superior performance in terms of Clustering Accuracy (ACC), Normalized Mutual Information (NMI), Purity (PUR), Adjusted Rand Index (AR), and F-score, attaining optimal or near-optimal results on multiple datasets and exhibiting strong clustering capability and cross-dataset adaptability. Further ablation studies verify the positive contribution of each module to performance improvement. Among them, low-rank constraints and graph regularization play important roles in global structure learning and local geometric relationship preservation, respectively, while TV regularization is particularly effective in enhancing representation smoothness, alleviating noise-induced representation fluctuations, and improving model stability. Robustness experiments further demonstrate that the proposed method maintains stable performance under different noise intensities, indicating its effectiveness in mitigating the adverse influence of noise on feature representation learning and clustering results. In summary, the proposed SLS-MVSC method effectively improves clustering performance, representation stability, and noise robustness, providing an effective solution for complex multi-view data clustering tasks.

Existing salient object detection (SOD) methods generally follow the principle of passive visual stimulation. These methods rely on low-level features such as color, texture, and contrast to determine salient regions. They treat the object with the strongest visual features in the scene as the focus of user attention. However, they neglect the decisive role of active user demands in determining salient targets. In real scenarios such as human-computer interaction and robot inspection, target objects are often submerged in complex backgrounds. In these cases, user attention has a strong intent orientation. Recently, the user-demand-driven salient object detection (UserSOD) task is introduced. It shifts the perception paradigm from passive visual response to active intent matching. This task also provides corresponding benchmark datasets and baseline models. Existing methods lack deep fusion and dynamic calibration between visual features and user demand semantics. Consequently, models only capture shallow semantic associations of keywords. Moreover, the downsampling process in hierarchical backbone networks easily loses spatial details. Therefore, models fail to accurately locate targets that match abstract demands in complex backgrounds. The detection performance remains heavily limited. To address these issues, this paper proposes a depth calibration and attribute-aware network (DCA-Net) within the UserSOD task framework. This network adopts the Swin-Transformer model as the visual backbone. It combines a pre-trained Contrastive Language-Image Pre-training (CLIP) text encoder as the text branch. The network designs a cascaded semantic recalibration encoder (CSRE) as its core. The CSRE contains four independent semantic recalibration modules (SRM). Each SRM utilizes a cross-modal cross-attention mechanism to achieve step-by-step semantic alignment between visual features and user intent at the source of feature extraction. Meanwhile, each SRM introduces a gated feature flow mechanism (ScGate) based on a multilayer perceptron structure. The ScGate uses full-connection transformation and ReLU-Tanh activation to generate adaptive weights. These weights dynamically adjust the intensity of semantic flows in each channel. This mechanism accurately locks the target region in the early stage of the feature flow. Meanwhile, the network constructs a convolutional cross-scale interaction module (CCIM) to achieve cross-level feature compensation. The CCIM aggregates global features using a pyramid pooling aggregation operator and a 1×1 convolution. It captures multi-scale context information through three parallel 3×3 depthwise separable convolution branches with different dilation rates. Then, the module applies weights by combining serial channel and spatial attention modules. Finally, it uses bilinear interpolation upsampling to restore the feature size. This module significantly enhances the perception ability of the model for spatial details of multi-scale targets.Furthermore, the network introduces a fine-grained attribute-aware decoder (AGD) to provide fine-grained constraints. The AGD adopts a top-down path to step-by-step upsample and merge high-level features. It utilizes the CLIP text encoder to initialize four types of preset attribute prompts, including category, color, appearance, and functional requirements. This initialization provides semantic priors and avoids the random convergence of attribute queries. Next, the AGD extracts explicit semantic constraint vectors of the four attribute dimensions from user instructions via multi-head cross-attention. This operation decouples abstract intent into four independent fine-grained attribute constraints. It avoids semantic confusion caused by multi-attribute demands. Then, the decoder feeds visual features into four parallel attribute-specific feature branches to perceive the visual patterns of corresponding attributes. These features then fuse with the constraint vectors. Finally, global text features dynamically predict the attribute existence probability weights. The decoder adaptively adjusts the contribution intensity of each branch based on the attribute of the four categories. This process ensures the precise correspondence between fine-grained constraints and user intent. It guides the model to generate detection masks with clear boundaries and consistent semantics.Experimental results on the UserSOD benchmark dataset demonstrate the effectiveness of the model. Compared with the state-of-the-art methods, DCA-Net improves the S-measure, F-measure, and E-measure by 4.5%, 6.8%, and 3.8%, respectively. It also reduces the mean absolute error (MAE) to 0.028. These results effectively validate the superiority and strong robustness of the proposed architecture in complex background interference, multi-scale target capture, and abstract intent alignment.

Image deraining aims to restore high-quality clean images from rain-degraded inputs and is a key technique for improving the robustness of outdoor vision systems. Existing convolutional neural network based methods are limited by local receptive fields and thus struggle to effectively model long-range dependencies. Although some Transformer-based methods enhance global modeling capability, they still show limited ability in handling multi-scale and directional rain streaks as well as recovering local high-frequency details. Moreover, existing frequency-enhanced methods often treat the frequency domain merely as an auxiliary representation, lacking fine-grained collaborative modeling of the complementary relationship between the spatial and frequency domains. To address these issues, this paper proposes a synergizing frequency and spatial network (SynFSNet) for image deraining. Unlike existing methods that mainly use the frequency domain as a global enhancement tool, the proposed method starts from the directional, density-related, and scale-varying characteristics of rain streaks in the frequency domain. Specifically, a Multi-Scale Fourier Fusion Module (MSFFM) is designed to progressively suppress rain streaks in a coarse-to-fine manner, while a Fourier Shaped Attention Module (FSAM) is introduced, in which Rectangular Filter Attention (RFA) and Square Filter Attention (SFA) are used to enhance direction-sensitive and local-structure-sensitive frequency representations, respectively. Furthermore, a dual-domain loss is employed to jointly constrain spatial structural restoration and frequency-domain consistency, thereby improving restoration quality in complex scenes. Experimental results show that SynFSNet achieves superior performance on multiple synthetic and real-world datasets, verifying the effectiveness of collaborative spatial-frequency modeling for complex image deraining tasks.

Physiological time series classification plays a critical role in various healthcare tasks, including sleep monitoring, electrocardiogram diagnosis, and epilepsy detection. However, in practical applications, severe class imbalance makes it challenging to learn effective representations for minority classes. Although these minority samples are scarce, they often contain highly informative signals, and accurate identification of their patterns is crucial for timely disease intervention, precise health assessment, and optimized decision-making in clinical and other application scenarios. Due to the limited number of minority samples and the predominance of background or normal waveforms, models struggle to learn discriminative features from entire sequences. Within these minority samples, a small number of critical local segments exist, whose unique structures or dynamic patterns carry essential class-discriminative information. Recognizing these segments is vital for effective modeling of minority classes. Contrastive learning offers strong generalization and feature distribution awareness and has been widely used to address class imbalance. Yet, most existing approaches rely on global sample representations or originate from image-based methods, making it difficult to capture sparse yet discriminative local waveform segments in minority samples. Moreover, current methods often underutilize frequency-domain information, and data augmentation typically lacks class differentiation, ignoring distributional differences between minority and majority classes. To address these challenges, this study proposes KWave-CL, a key waveform-aware contrastive learning method for class-imbalanced physiological time series classification. Hard-to-learn key segments in minority samples typically exhibit large reconstruction errors. KWave-CL employs a variational autoencoder jointly modeling time- and frequency-domain information to reconstruct minority waveforms. By computing reconstruction errors in both domains, the method identifies key and non-key waveform segments, providing critical guidance for subsequent contrastive learning. To fully exploit local discriminative information in minority samples, a key waveform-aware contrastive loss is designed to pull key segments closer while pushing key and non-key segments apart in the representation space, enhancing the discriminability of minority class embeddings. To increase minority diversity while maintaining the stability of majority features, class-differentiated data augmentation is applied, imposing stronger perturbations on minority samples and weaker perturbations on majority samples, thereby mitigating class distribution shifts in the augmented space. The method adopts a joint optimization strategy, integrating self-supervised contrastive loss, key waveform-aware contrastive loss, and time-frequency variational autoencoder reconstruction loss, enabling cooperative learning of global and local features. KWave-CL is also highly flexible and can be embedded into various time series contrastive learning frameworks. Experiments on three publicly available physiological datasets show that KWave-CL outperforms multiple baseline methods for class-imbalanced learning under two representative contrastive learning frameworks. On the PhysioNet 2017 dataset using an instance-level framework, the overall F1 score improves by up to 6.69%, while the minority class F1 score increases by up to 11.67%. Ablation studies further demonstrate that the key waveform-aware contrastive loss, time-frequency variational autoencoder, and class-differentiated data augmentation all play crucial roles in enhancing minority class performance. These results indicate that KWave-CL effectively mitigates class imbalance and provides reliable decision support for healthcare applications.

To address the problems of extremely small target scale, weak texture information, low signal-to-noise ratio, and severe background interference in infrared UAV aerial images, while satisfying the lightweight and real-time deployment requirements of UAV platforms, this paper proposes a lightweight infrared small object detection algorithm named ACFF-YOLOv12n based on YOLOv12n. Existing infrared small object detection methods generally suffer from insufficient fine-grained feature modeling, inefficient multi-scale feature fusion, and poor adaptability of detection heads to low-contrast targets, resulting in feature dilution, missed detections, and false alarms during deep feature propagation. To improve the perception capability of weak infrared targets in complex scenes, the proposed method performs collaborative optimization from three aspects: backbone feature extraction, feature fusion strategy, and detection head design. First, an A2C2f-ACmix++ feature enhancement module is constructed in the backbone network. By integrating global self-attention and local enhanced convolution through a dual-path collaborative mechanism, together with a Lightweight Channel Attention (LCA) module, the network can adaptively strengthen weak target channel responses and enhance the extraction capability of edge and texture information under limited computational cost. Second, a Semantic-Gated Dynamic Fusion Module (SGDFM) is designed in the neck network. High-level semantic features are used as dominant guidance, while shallow detail features are treated as auxiliary information. Through dynamically generated gating weights, adaptive cross-level feature fusion is achieved, effectively alleviating the feature dilution problem of infrared small targets during deep propagation and improving robustness under complex backgrounds and scale variations. Finally, a CondConv-based Lightweight Detection Head (CLD-DET) is proposed for infrared small object detection. The detection head introduces Conditionally Parameterized Convolution (CondConv) to dynamically adjust convolution kernels according to input features and incorporates a Small Object Enhancement (SOE) module to strengthen weak edges and low-contrast spatial details, thereby improving the localization and classification capability for infrared weak targets. Experiments are conducted on the HIT-UAV infrared dataset, and additional generalization evaluations are performed on the SIRST and NUDT-SIRST datasets. Experimental results demonstrate that ACFF-YOLOv12n achieves an mAP@0.5 of 87.1% on the HIT-UAV dataset, which is 5.7 percentage points higher than YOLOv12n, while the mAP@0.5:0.95 reaches 56.8%, improving by 2.8 percentage points. Meanwhile, the number of parameters is reduced to 2.13M and the computational complexity is reduced to 5.0 GFLOPs, achieving an effective balance between detection accuracy and lightweight deployment. On the SIRST dataset, the proposed method achieves an mAP@0.5 of 78.6%, outperforming YOLOv12n by 5.6 percentage points. On the NUDT-SIRST dataset, the mAP@0.5 reaches 76.7%, improving by 3.4 percentage points over the baseline model, demonstrating strong cross-scene generalization and robustness. Furthermore, visualization and heatmap analyses show that the proposed model can generate more accurate and balanced feature attention regions in complex backgrounds, long-distance weak target scenes, and dense small-target scenarios, effectively reducing missed detections and false alarms. Overall, ACFF-YOLOv12n achieves a superior balance among detection accuracy, computational complexity, and real-time performance, making it highly suitable for infrared UAV small object detection tasks and practical applications such as UAV inspection, border surveillance, and complex environment monitoring.

raffic flow prediction is a key core technology of intelligent transportation systems, which has significant value in improving the efficiency of urban traffic management. In traditional methods, time series prediction and machine learning models are widely used, but spatiotemporal graph neural networks have become a research hotspot due to their outstanding advantages in traffic flow representation learning. However, existing methods have significant limitations: firstly, they rely on static graph structures, making it difficult to model long-range spatial dependencies and regional differences; The second issue is the failure to capture the time delay effect between road segments, resulting in insufficient modeling of spatiotemporal dependencies. In response to these issues, this paper proposes a dynamic graph convolutional traffic flow prediction model based on time-delay aware attention mechanism (TLAA-SDGC). The encoder decoder architecture is adopted, and the mechanism is combined with spatial dynamic graph convolution to mitigate the adverse effects of the time-delay phenomenon on prediction accuracy. Specifically, gated causal convolution incorporates attention mechanisms to explicitly model the asynchronous temporal dependencies of spatial nodes by learning delay parameters; The dynamic adaptive spatial correlation matrix can perceive real-time changes in traffic status and accurately depict the dynamic propagation process of traffic flow in the road network. Meanwhile, the introduction of spatiotemporal embedding enables the model to accurately depict the short-term fluctuations, long-term periodic patterns, and spatiotemporal heterogeneity of traffic flow under network topology constraints.Experiments on public benchmark datasets demonstrate that this method effectively addresses challenges related to time lag and dynamic behavior. Compared to existing state-of-the-art baseline models, it achieves significant reductions in key prediction error metrics (MAE, RMSE) ranging from 6.5% to 11.9%.

Simultaneous localization and mapping (SLAM) in dynamic scenes is highly susceptible to moving objects, occlusions, and illumination variations, which often lead to pose drift and reconstruction artifacts. Existing methods remain limited in identifying unreliable dynamic observations and constraining map updates. They frequently depend on category priors, fixed thresholds, or scene-specific distributions, and may incorporate moving objects into the map as spurious static structures. To address these issues, a Multi-Source Consistency Residual-Enhanced Gaussian SLAM method, termed MCRGS-SLAM, is proposed for dynamic scenes. It models dynamic interference from the perspective of observation reliability in a continuous and interpretable manner, and embeds reliability constraints into both front-end pose estimation and back-end Gaussian mapping to improve the localization accuracy and static map reconstruction quality of monocular SLAM. The method is built on the physical constraint that static regions should satisfy multi-view consistency. It constructs four complementary consistency residuals, namely appearance, geometric, motion, and structural residuals, thereby transforming dynamic observation identification into a quantifiable physical measurement. Specifically, the appearance residual characterizes cross-view brightness and texture consistency, the geometric residual measures the stability of depth projection relationships, the motion residual detects independent displacement that cannot be explained by camera ego-motion, and the structural residual describes variations in local edges and geometric patterns. Unlike strategies that directly remove dynamic regions using binary masks or fixed thresholds, these residuals are modeled as continuous physical constraints to represent the degree to which pixel observations deviate from the static-scene assumption. On this basis, a semantic–geometric dual-stream reliability inference network, named Multi-source Consistency Residual Network (MCR-Net), is designed. The semantic stream extracts high-level semantic features to provide category-level dynamic priors, while the geometric stream encodes residual evidence to represent multi-view consistency. The two streams are fused through an attention mechanism to generate a pixel-wise reliability map. This map is introduced into both the front-end and back-end of the SLAM system as soft weights. In front-end pose optimization, appearance, motion, geometric, and structural constraints are weighted by reliability to reduce the interference of dynamic outliers in camera pose estimation. In back-end 3D Gaussian Splatting (3DGS) mapping, the reliability map guides the initialization, update, and removal of Gaussian primitives, thereby adaptively suppressing map contamination caused by dynamic observations. In this way, dynamic observation handling is transformed from discrete removal into reliability-based continuous weighting, which preserves boundary regions and weakly reliable static observations while reducing the cumulative influence of dynamic outliers on pose estimation and Gaussian map updates. In addition, MCRGS-SLAM establishes a self-supervised closed-loop optimization mechanism based on reprojection errors and rendering consistency, enabling the network to adapt online to dynamic variations in unknown scenes. Experiments on dynamic-scene datasets, including Bonn and TUM, show that MCRGS-SLAM achieves competitive performance in both localization accuracy and reconstruction quality. In localization evaluation, MCRGS-SLAM obtains an average ATE RMSE of 2.35 cm on the Bonn Dynamic dataset, outperforming several representative methods. These results indicate that reliability-weighted optimization effectively reduces the impact of dynamic observations on pose estimation. In the reconstruction and rendering quality evaluation, the proposed method achieves average PSNR, SSIM, and LPIPS values of 17.99 dB, 0.730, and 0.272 on the TUM dynamic dataset, respectively. Compared with Dy3DGS-SLAM, a representative dynamic 3DGS-SLAM method, it improves PSNR by 0.14 dB and SSIM by approximately 1.1%, while reducing LPIPS by approximately 4.6%, effectively mitigating reconstruction artifacts caused by dynamic objects. Results on real-world complex scene sequences further demonstrate that MCRGS-SLAM maintains stable performance under scene distribution shifts and unstructured dynamic interference, indicating good cross-scene applicability.

The anonymity inherent in blockchain provides a concealed environment for transactions, and Bitcoin address type detection, which aims to categorize addresses with distinct behavioral patterns, is of great significance for blockchain de-anonymization analysis. Existing methods face three major challenges: (1) the Bitcoin transaction network is large and complex, leading to low training efficiency when directly using graph neural networks; (2) deep graph neural networks are prone to over-smoothing, causing node representations to converge and discriminative ability to decline; (3) most methods rely solely on static graph structures or simple statistical features, ignoring temporal transaction patterns and failing to capture dynamic address behaviors. To address these issues, this paper proposes F-BAC (Filter-enhanced Bitcoin Address Classifier), a Bitcoin address type detection method based on a temporal graph feature network. The method consists of four core steps. First, transaction modeling: each transaction of a labeled address is constructed as an independent homogeneous graph, where nodes are addresses involved in the transaction and edges represent fund flows. To handle the large variance in transaction scales, a fine-grained address aggregation compression strategy is proposed, which divides nodes into four tiers based on their total occurrence frequency in all transaction records and aggregates them accordingly, transforming the original heterogeneous graph into a small-scale homogeneous graph with regular structure. This significantly simplifies topological complexity and lays an effective foundation for subsequent graph representation learning. Second, address feature extraction and augmentation: local features are computed for each address, and statistical features are computed for each aggregated address group. Then, a graph feature network (GFN) is applied to propagate and augment the features through three layers while preserving the original node information at each layer, effectively mitigating the over-smoothing problem. Third, temporal feature optimization: the time-ordered sequence of transaction graph representations of the same address is fed into a filter-enhanced MLP (F-MLP), transformed to the frequency domain via discrete Fourier transform, where a trainable linear filter suppresses obfuscation behaviors and noise, and finally reconstructed to an enhanced temporal signal via inverse Fourier transform. This design significantly improves the model's ability to capture periodic transaction patterns. Fourth, address classification: the filtered and optimized temporal graph representation sequence is input into a classifier to output the address type label, trained under the supervision of the cross-entropy loss function. This paper also constructs a dataset incorporating temporal features, providing an important data foundation for subsequent research. Systematic experiments are conducted on our self-constructed dataset and three public datasets. Ablation studies show that removing the address aggregation module, the GFN module, or the temporal filtering module degrades model performance to varying degrees, validating the necessity of the three strategies. On our self-constructed dataset, F-BAC achieves 96.7% precision, 95.9% recall, 96.3% F1-score, and a macro-averaged F1 of 91.1%. On the BAC dataset, F-BAC achieves an F1-score of 98.4%; on the BATC dataset, 98.0%; and on the Google/CvdxBp dataset, 94.6%. Comparative experiments show that F-BAC outperforms existing methods including BAClassifier, Multi-hop GAT, Balanced-BiEGCN, and MDST-GNN. Furthermore, F-BAC reduces data collection volume by 89.3% compared to the BABD method, and the GFN module saves approximately 30%–45% of training time compared to GCN, significantly reducing training costs. In summary, the proposed F-BAC method effectively alleviates the problems of large graph scale, over-smoothing, missing temporal patterns, and obfuscation interference in Bitcoin address type detection through fine-grained address aggregation, graph feature network enhancement, and temporal frequency-domain filtering. Experiments on multiple datasets validate that F-BAC has good generalization performance for this task.

A Multi-distribution and Multi-scale Adaptive Prior Transformer for Anomaly Detection (MMAPT-AD) is proposed to address the limitations of existing multivariate time series anomaly detection methods in modeling complex temporal dependencies, multi-scale dynamic patterns, and prior information utilization. The model employs a data embedding module to obtain unified representations of multivariate time series from both temporal and variable dimensions, enhancing the representation of variable coupling relationships and dynamic temporal patterns. A multi-distribution and multi-scale prior generation mechanism is further introduced by jointly incorporating Gaussian, Laplace, and Cauchy distributions under different temporal scales to model latent correlation structures within the sequence. Specifically, the Gaussian distribution characterizes stationary variation patterns, the Laplace distribution captures local abrupt changes, and the Cauchy distribution improves the adaptability to long-tail anomalies and complex fluctuation patterns. To describe heterogeneous temporal dependencies across different scales, prior correlation matrices are generated at multiple temporal granularities for joint modeling of local dependencies and global temporal structures. Learnable fusion weights are adopted to integrate heterogeneous prior information from different distributions and scales, improving the representation capability for complex statistical characteristics and dynamic patterns. Based on the generated priors, a prior-guided anomaly attention mechanism is designed by incorporating multi-scale prior information into the attention weight computation process. The proposed mechanism introduces structural prior constraints while learning temporal correlation features and guides the model to focus on anomaly-related temporal segments and variable channels, thereby improving the detection capability for local anomalies, sparse anomalies, and complex structural anomalies. To enhance anomaly discrimination capability, a joint optimization objective combining reconstruction error, attention-prior discrepancy constraints, and multi-scale inconsistency constraints is constructed. In addition, a comprehensive anomaly scoring strategy integrating reconstruction error, attention distribution discrepancy, and multi-scale structural deviation is developed for time-step-level anomaly detection. Experiments on five public datasets, including SMD, MSL, SWaT, SMAP, and PSM, show that MMAPT-AD achieves F1 scores of 93.40%, 94.99%, 96.06%, 96.67%, and 98.06%, respectively. On the SMD dataset, the proposed method improves the F1 score by 7.18 percentage points and 1.07 percentage points compared with InterFusion and Anomaly Transformer, respectively. On the MSL dataset, MMAPT-AD achieves an F1 score of 94.99%, outperforming TransDe by 0.63 percentage points. On the SMAP dataset, the Recall reaches 99.35%, indicating strong anomaly coverage capability. Ablation studies demonstrate that multi-distribution prior modeling, multi-scale structural constraints, and the joint optimization strategy all contribute to performance improvement. Robustness experiments further verify the stability and generalization capability of the model under different input perturbation conditions. Experimental results demonstrate that MMAPT-AD effectively integrates multi-distribution statistical characteristics and multi-scale temporal dependencies, exhibiting strong anomaly detection capability and structural adaptability in complex dynamic scenarios.

To address the problems of decreased accuracy in causal structure learning and unreliable causal direction determination among observed variables under latent-variable interference, a latent variable causal discovery method based on non-Gaussian information is proposed. In real-world data, there often exist latent variables that cannot be directly observed or completely recorded. These latent variables may simultaneously affect multiple observed variables, thereby inducing statistical correlations among observed variables due to common causes. Causal discovery methods that do not consider the existence of latent variables tend to misinterpret such latent-variable-induced correlations as causal relationships, resulting in spurious causal edges, incorrect causal directions, and incomplete structure recovery. To improve the accuracy of causal structure recovery in scenarios with latent variables, the proposed method utilizes the non-Gaussian information contained in data under a linear non-Gaussian acyclic causal model with latent variables, and identifies both direct causal relationships among observed variables and undirected causal relationships affected by latent variables. Specifically, in the first stage, the method takes a complete undirected graph over observed variables as the initial structure, and gradually removes statistically independent variable pairs through conditional independence tests based on regression residuals, thereby obtaining an initial skeleton structure among observed variables. This stage can preferentially determine some reliable direct causal relationships among observed variables without explicitly modeling latent variables, providing a structural basis for subsequent latent variable detection and orientation of remaining edges. In the second stage, for variable relationships that remain unoriented and may be affected by latent variables, higher-order cumulants are introduced to characterize the higher-order statistical information in non-Gaussian distributions. Fourth-order joint cumulants are used to estimate the effect strength of latent variables on observed variables, and whether multiple observed variables are affected by a common latent variable is determined by testing whether the shared single latent component condition is satisfied. On this basis, the remaining unoriented edges among observed variables affected by latent variables are further oriented, and a causal structure containing both observed variables and latent variables is finally output. Theoretical analysis shows that, when the variables are generated by a linear non-Gaussian acyclic causal model, the noise variables are mutually independent, and the relevant identifiability conditions hold, the proposed method can identify causal structures with latent variables by using independence constraints and higher-order cumulant information. To verify the effectiveness of the method, the proposed method is compared with several representative methods under five simulated causal graphs and three sample-size settings. Experimental results show that the proposed method maintains relatively high precision in most scenarios, with precision at least 30% higher than that of existing methods. Meanwhile, the proposed method achieves the best results in both recall and F1-score, and its F1-score is greater than or equal to 75% under different sample sizes. In particular, in simulated scenarios with strong latent-variable effects and complex structures, the proposed method can effectively reduce the omission of true causal edges and improve the overall structure recovery ability. Ablation experiments show that both the local structure identification stage without latent-variable interference and the stage of latent variable detection and unoriented causal edge identification play important roles in final causal structure recovery. In the real-world financial data experiment, return data of multiple constituent stocks in the Hong Kong stock market are used to further verify the applicability of the proposed method in practical scenarios. Experimental results show that the proposed method can identify latent variable structures with clear economic meanings while maintaining reasonable sparsity of the causal graph, and can characterize a multi-level causal transmission relationship from market-wide macro factors to industry sectors and then to individual stocks. Compared with several representative methods, the proposed method avoids the problem of overly dense causal graphs and compensates for the limitations of some methods that lack global driving factors or do not explicitly characterize latent variable structures. Overall, the results of simulation experiments, ablation experiments, and real-world data experiments demonstrate that the proposed method can effectively utilize higher-order statistical information and conditional independence constraints, improve the accuracy, completeness, and interpretability of causal structure recovery in scenarios with latent variables, and provide an effective approach for causal analysis and intelligent decision-making in complex data environments.

Road waterlogging is one of the most direct risk forms affecting traffic operation and emergency dispatch during urban rainstorm flooding. Timely recognition of waterlogging levels is therefore important for road control, rescue scheduling, and public travel guidance. However, waterlogging level recognition from real road surveillance videos remains challenging. First, heavy rainfall, nighttime low illumination, vehicle occlusion, and water-surface reflection cause large intra-class variations. Second, high-risk categories such as heavy and severe waterlogging occur rarely, resulting in a long-tailed data distribution. Third, video clips from the same camera, road segment, or rainfall event may be highly similar, which can bias the evaluation of generalization if not handled properly. To address these problems, this paper proposes a few-shot road waterlogging level recognition method based on multimodal label constraints. The proposed method first extends the I-JEPA-style joint-embedding predictive architecture to video clips. Cross-frame spatiotemporal blocks in consecutive frames are used as context and prediction targets, and waterlogging-related target-block features are predicted in the representation space. This avoids redundant pixel-level reconstruction and encourages the encoder to focus on semantic cues related to waterlogging levels, such as road water layers, vehicle wheel submergence, reflective textures, and water-surface disturbances. Second, the CLIP text encoder is used to extract semantic features from waterlogging-level descriptions. Textual constraints are introduced during support prototype construction so that visual features are aligned with category semantics such as slight, mild, moderate, heavy, and severe waterlogging in a shared space, thereby alleviating unstable decision boundaries between adjacent water-depth levels under few-shot conditions. Third, dynamic time warping is introduced in the matching stage between query videos and support prototypes. By searching the optimal temporal alignment path through a frame-level cumulative distance matrix, the model can adapt to differences in video length, vehicle-water interaction moments, and key-frame locations, improving matching stability across different shooting rhythms. Two waterlogging level datasets are constructed from real road surveillance videos in Nanjing and Xiamen. Before splitting the data into training, validation, and testing sets, near-duplicate clips from the same source are removed, and camera IDs, road locations, and rainfall-event times are cross-checked to reduce the risk of leakage caused by highly similar clips appearing across different sets. For comparison, ResNet50, VGG-16, ViT-B, SwinT-B, HCL, OTAM, CPEA, CLIP-FSAR, Qwen2-VL, and Qwen2.5-VL are selected as baselines, and the year and publication venue of each method are added to the tables. In addition, AVF-MAE++ and TEAM, two officially published and open-source methods after 2025, are included for method-level comparison to clarify the difference between the proposed method and recent open-source video representation learning and few-shot video recognition approaches. Experimental results show that the overall accuracy of the proposed method reaches 96.4% and 95.0% on the Nanjing and Xiamen datasets, respectively, outperforming CNN, Transformer, few-shot video recognition, and multimodal large-model baselines. Considering the limited number of minority-class test samples, P/R=100% under a single split is not used as independent evidence of stability. Instead, five independent stratified re-splitting experiments are added so that heavy and severe waterlogging samples are covered multiple times in different test subsets. The standard deviation of overall accuracy is below 0.8% over five runs, indicating that the proposed method has good robustness for long-tailed few-shot categories and cross-city scenarios.

With the continuous progress in drone technology, the detection of small targets in drone aerial photography has garnered increasing scholarly attention. To tackle the challenges of low detection accuracy and high missed detection rates caused by small target scales, dense distributions, and complex backgrounds in drone aerial images, we propose PK-YOLO, a drone aerial small target detection model based on an improved YOLOv11n. Using YOLOv11n as the baseline model, we introduce targeted enhancements in three key areas: feature extraction, feature fusion, and loss function, with the objective of improving the detection accuracy and overall robustness of small targets in drone aerial scenarios. In the feature extraction phase, recognizing that YOLOv11n's feature pyramid structure, spanning layers P3 to P5, exhibits limited responsiveness to extremely small targets (less than 32×32 pixels), leading to missed and false detections, we incorporate an additional small target detection layer, namely the P2 detection layer. This layer leverages higher-resolution feature maps to preserve richer spatial position and edge detail information, facilitating a gradual fusion from high-resolution shallow features to semantically rich deep features, thereby significantly enhancing the network's capacity for extracting features from small targets. Furthermore, we design a C3k2_KSFA module that integrates the Kernel Selective Fusion Attention (KSFA) mechanism to replace the original C3k2 module in the backbone network. The KSFA mechanism draws inspiration from dynamic convolution and selective kernel networks, extracting multi-scale features by employing convolution kernels with various dilation rates and sizes in parallel. It then adaptively assigns optimal kernel weights to different spatial locations using a spatial-spectral selection mechanism, enabling the model to dynamically adjust its receptive field based on the actual size of the input target, thus improving feature representation and discrimination for targets of varying scales. During the feature fusion stage, traditional fixed upsampling methods struggle to accommodate the dramatic variations in target scales in drone images, often resulting in the loss of small target features and susceptibility to complex background interference. To address this, we replace the fixed upsampling module in the neck network with DySample dynamic upsampling. Utilizing a point sampling strategy, DySample adaptively adjusts sampling point positions according to the feature map content, effectively suppressing background interference while enlarging the feature map and reducing the loss rate of small target features. In terms of loss function optimization, acknowledging that the CIoU loss function employed by YOLOv11n is prone to interference from low-quality prediction boxes, leading to subpar localization accuracy, we introduce the Inner-IoU mechanism to enhance Wise-IoU v3, proposing the Inner-WIoU loss function. This function improves the fine-grained description of target location and shape while retaining the ability to focus on challenging samples through a dynamic non-monotonic frequency modulation mechanism and auxiliary bounding boxes, thereby effectively enhancing the localization accuracy of small targets. To validate the model's effectiveness, we conduct comparative experiments on three publicly available datasets: VisDrone2019, TinyPerson, and RSOD. The experimental results on the VisDrone2019 dataset reveal that, compared to the baseline model, the PKD-YOLO model achieves improvements of 5.0% and 3.3% in mAP@0.5 and mAP@0.5:0.95, respectively, along with a 5.0% increase in precision. On the more demanding TinyPerson dataset, these three metrics improve by 5.7%, 1.9%, and 7.4%, respectively, fully demonstrating the model's superior detection capabilities for small targets in complex environments. On the RSOD dataset, mAP@0.5 and mAP@0.5:0.95 reach 96.1% and 68.8%, respectively, representing improvements of 2.7% and 3.6% over the baseline model, indicating that the model maintains high localization accuracy while enhancing its ability to discriminate target detection confidence. Combining the experimental results from these three datasets, PKD-YOLO exhibits exceptional small target detection performance across various drone aerial scenarios, confirming the effectiveness and generalizability of our proposed improvement methods.

In recent years, generative models have rapidly advanced, demonstrating strong capabilities in image synthesis, artistic creation, and digital portrait generation. However, while improving generation performance, these models also pose significant privacy risks, as private information in the training data may be leaked in the generated content. To address this issue, machine unlearning has been proposed to reduce a model’s memory of specific data, preventing the disclosure of private information. In particular, ID unlearning for face generation aims to prevent the model from generating images of specific identities, thereby protecting personal privacy. Nevertheless, existing ID unlearning methods still suffer from insufficient robustness. Attackers can exploit a small amount of unlearned identity data to perform limited steps of retraining, thereby recovering the unlearned identity, a process known as relearning attack. Experiments on the CelebAHQ dataset show that, although existing ID unlearning methods can effectively unlearn specific identity, the unlearned effect can be easily reversed under relearning attacks, indicating limited robustness in such scenarios. Therefore, while existing methods achieve reasonable unlearning in initial evaluations, they remain vulnerable in practical attack settings. To address this problem, this work has two objectives: first, to construct a relearning attack method to evaluate the robustness of existing ID unlearning methods; second, to propose a robust ID unlearning method that can maintain effective unlearning under such attacks. Technically, we first design a relearning attack method for generative adversarial networks (GANs), which updates the model with a small amount of unlearned identity data using limited gradient steps, causing the generated images to approach the unlearned identity representations and thereby assessing the robustness of ID unlearning. Second, we propose robust ID unlearning via smoothness optimization (RIDU), a robust ID unlearning method based on smooth optimization. During training, RIDU applies random perturbations to the model parameters, allowing optimization to not only achieve unlearning at the current parameter point but also maintain stability across neighboring regions, forming a smooth and stable unlearning zone. In this way, the unlearning effect does not rely on a fragile local optimum, making it difficult for relearning attacks to restore the unlearned identity. Furthermore, RIDU incorporates appropriate loss functions to balance unlearning objectives with generation quality. Experiments on multiple public datasets demonstrate the effectiveness of RIDU. On the CelebAHQ dataset, RIDU significantly reduces the similarity between specific identity and generated images under non-attack conditions, outperforming existing methods. Under relearning attacks, existing methods are easily reversed, whereas RIDU maintains strong unlearning, effectively suppressing identity recovery. Additional experiments indicate that RIDU preserves the model’s generative capability while unlearning specific identities. In summary, our work introduces a relearning attack to evaluate the robustness of existing ID unlearning methods and proposes RIDU, which integrates smooth optimization with ID unlearning to enhance robustness. RIDU also ensures effective unlearning under relearning attacks while simultaneously maintaining the model’s generative capability.

To address bottlenecks limiting LLMs in high-knowledge-density, logically-constrained domains like healthcare—namely the lack of self-correction in homogeneous multi-agent systems due to cognitive uniformity, and diminishing returns of computational investment—this study proposes a multi-agent validation fusion architecture based on heterogeneous model interaction. The architecture builds a differentiated hierarchical collaborative framework via prompt engineering, disrupting logical blind spots of single-model lineages. A high-performance large-scale model serves as the core decision-making agent, parsing medical records and generating preliminary diagnostic logic. Concurrently, heterogeneous-source models—with distinct output stylistics, conservative logical biases, and divergent training distributions—form an independent verification layer. This layer does not generate answers but conducts multi-dimensional audits through a “verification–issue enumeration–response” paradigm, scrutinizing inference pathways, factual consistency, and logical soundness. The theoretical foundation leverages cognitive divergences among heterogeneous models to enable productive logical conflict and rigorous cross-verification, enhancing output diversity, robustness, and medical factual rigor. In diabetes care evaluations, the system improves accuracy by 10% on multiple-choice questions, 8% on fill-in-the-blank tasks, and 22% on complex reasoning problems versus a single-agent baseline. Compared to Colacare, it shows superior logical consistency and inference stability. Notably, it surpasses Diabetica-7B (a deeply fine-tuned specialized model) without costly annotated data or fine-tuning. Experimental records show average inference cost of 61,881 tokens/query and latency of 71.74 seconds/query. Relative to specialized models, this system avoids prohibitive training costs and long data annotation cycles, while reducing architecturalcomplexity via a modular plug-in design. We conclude that for medical decision-making demanding high knowledge integration, introducing inter-model heterogeneity—not merely scaling homogeneous capacity—is critical for reliability. Optimizing heterogeneous-agent verification remediates logical vulnerabilities via inherent robustness without compromising large-model inference bounds. These findings offer a low-cost technical framework for medical AI and support the paradigm shift from “general intelligence” to “reliable professional intelligence.” Reduction achieved by removing phrases like “fundamental bottlenecks constraining the application,” “attributable to underlying,” “characterizing computational investment relative to performance gains,” “elaborately engineered,” “endowed with massive parameter counts and advanced generalized reasoning capabilities,” “innovatively,” “crucially,” “systematically,” “instantiate productive,” “concurrently improving model adaptability and stability across diverse clinical scenarios,” “specialized experimental evaluations targeting,” “through comprehensive comparative analysis with the state-of-the-art policy model,” “more prominently,” “professionally specialized,” “subjected to deep domain-specific fine-tuning,” “regarding the trade-off between engineering feasibility and resource consumption,” “protracted,” “overall,” “our research conclusions establish that,” “the successful empirical validation of this model demonstrates that,” “these findings furnish not only...but also,” and some redundant adjectives.

Anchor-based bipartite graphs can approximate sample relationships with low computational cost, and have been widely used for graph construction in large-scale multi-view clustering. However, most existing methods generate anchors by random sampling, K-Means centers, or static dictionaries. The anchor positions are usually fixed before graph learning. For data with non-spherical clusters, ring-shaped clusters, elongated clusters, and large density differences among clusters, static anchors are difficult to approach local density peaks and curved cluster boundaries. This may lead to anchor mismatch. The sample-anchor bipartite graph constructed in this way may contain redundant or unreliable connections, which affects cross-view structure fusion and reduces the stability of clustering results. To address these problems, a Multi-view Bipartite Graph Clustering method guided by Mean-Shift (MBGC-MS) is proposed. The aim of the method is to maintain the high efficiency of anchor graphs, improve the adaptive representation ability of anchors for complex data distributions, and integrate graph learning and clustering assignment into a unified optimization process. First, an anchor set is initialized in each view. Mean Shift is then used to iteratively update the anchors along the kernel density gradient, so that the anchors are changed from Euclidean center representatives to local density modes. After that, the local bandwidth is estimated according to the distance from each sample to its neighboring anchors. Only the neighboring anchor relationships of each sample are retained to construct a sparse and density-aware sample-anchor bipartite graph. Second, the bipartite graph of each view is decomposed into a shared consistency graph and view-specific complementary graphs. The consistency graph is used to describe the clustering structure jointly supported by different views, and a nuclear norm constraint is imposed to enhance its low-rank property and structural compactness. The complementary graph is used to preserve the discriminative local information that deviates from the consensus in each single view, and a sparsity constraint is imposed to suppress noise and redundant connections. Third, view weights are adaptively learned according to the reconstruction error, so as to avoid the information interference caused by equal-weight fusion of all views. On this basis, the consistency graph and complementary graphs are fused to construct a unified sample-anchor bipartite graph and an augmented Laplacian matrix. A spectral trace constraint is further imposed to make the unified graph tend to form a given number of connected components. The proposed model adopts a block coordinate descent strategy to alternately update the clustering indicator matrix, consistency graph, complementary graphs, and view weights. Each subproblem can be solved by eigen-decomposition, soft-thresholding operator, singular value thresholding, or quadratic programming. In this way, one-step optimization of graph structure learning and clustering assignment is achieved. Experiments are conducted on four multi-view datasets, including Handwritten, BBC-sport, MSRC_v1, and Caltech101-7. ACC, NMI, and F-score are used as evaluation metrics. The results show that the proposed method obtains the best performance on 8 out of 12 metrics. On Handwritten, the highest NMI is obtained, reaching 0.949. On BBC-sport, the ACC and F-score reach 0.971 and 0.933, respectively. On MSRC_v1, the ACC and F-score reach 0.966 and 0.931, respectively. On Caltech101-7, the ACC, NMI, and F-score reach 0.711, 0.521, and 0.484, respectively, and all three metrics are the best. Ablation experiments further show that Mean Shift anchor updating, consistency-complementarity decomposition, and Laplacian spectral trace constraint all contribute to performance improvement. On Caltech101-7, the ACC, NMI, and F-score of the complete model are higher than those of the model using only the Mean Shift module, whose corresponding values are 0.534, 0.362, and 0.348. The comparison of different anchor strategies shows that the three metrics of Mean Shift anchors are 0.711, 0.521, and 0.484, which are clearly higher than those of K-Means anchors, namely 0.576, 0.395, and 0.456, and also higher than those of random sampling anchors, namely 0.308, 0.260, and 0.320. The convergence experiments show that the objective function usually decreases rapidly in the first 5–10 iterations and becomes stable within 20 iterations. Complexity analysis and running time experiments show that, when the number of anchors and iterations are fixed, the running time of the algorithm grows approximately linearly with the sample size. In summary, the proposed method can calibrate anchor positions by using density information and enhance the representation ability of bipartite graphs for complex cluster structures. It also considers multi-view consistency, complementarity, and one-step clustering optimization. Therefore, it shows good clustering performance, stability, and scalability on non-spherical and uneven-density data.

This paper proposes a distributed task allocation method based on overlapping coalition formation game, to address the collaborative task allocation problem of heterogeneous multi-functional cleaning robot clusters in photovoltaic (PV) power stations under multiple constraints. First, this paper constructs a multi-constraint cooperative task optimization model of heterogeneous robot clusters for PV operation and maintenance (O&M). It systematically analyzes the resource adaptation mechanism between heterogeneous robots and diverse O&M tasks. It designs a matching mechanism between PV module pollution levels and robot cleaning capabilities. It establishes a comprehensive utility function integrating comprehensive resource matching degree, recovered power generation loss, total O&M cost, maximum completion time and redundancy penalty mechanism. The objective function of this task allocation problem is constructed to maximize the total coalition utility. Combined with engineering practice, it introduces constraint conditions including task priority, task uniqueness and safety distance. It completes the refined mathematical modeling of robot cluster O&M tasks. It significantly improves the resource matching degree and O&M scheduling adaptability between robots and tasks. Second, this paper constructs an overlapping coalition formation game framework based on the bilateral mutual benefit criterion. This framework converts the global combinatorial optimization problem into a distributed coalition partition problem. Relying on the defined preference relation and exchange operation, it supports robots to realize cross-coalition dynamic cooperation and independently form an overlapping coalition structure. This paper further proposes a bilateral mutual benefit exchange criterion to describe the mapping relationship between individual selfish profit-seeking and global system optimization. The criterion significantly improves the flexibility of resource scheduling. It effectively coordinates individual rationality and overall system utility. This game is proven to be a potential game. When robots change the resource allocation structure to maximize their own utility, the difference of the objective function is consistent with the difference of the utility function. And there exists at least one Nash equilibrium, namely a stable coalition structure, in the game framework. Finally, this paper designs an overlapping coalition formation algorithm integrating preference gravity guidance and disturbance mechanism. The algorithm introduces a preference gravity mechanism to provide directional guidance for robots to select coalitions. It uses the marginal revenue evaluation mechanism and the proposed double tabu list to eliminate inefficient coalitions and avoid invalid searches. Combined with the co-designed dynamic resource adjustment strategy and random disturbance mechanism, it effectively gets rid of local optimum. This algorithm is proven to converge to a T-stable state within a finite number of iterations. Its complexity is analyzed to be within a controllable range. When solving the task allocation mathematical model composed of the task optimization model and the game framework, the algorithm can meet the task allocation requirements of heterogeneous robot clusters in complex and diverse task cooperation scenarios. It can quickly converge to a high-quality Nash equilibrium solution. Simulation experimental results show that the proposed method has remarkable real-time performance and stability on the premise of ensuring the quality of the global solution. Ablation experiments verify that each improvement strategy can comprehensively improve the comprehensive performance of the algorithm. In experiments verifying algorithm solution quality and O&M indicators, compared with the baseline algorithm, the proposed algorithm increases the recovered power generation loss by 6.63%, reduces the total O&M cost by 2.55%, and shortens the average operation time by 13.29%. In performance comparison experiments under scale changes, all indicators of the proposed algorithm are optimal under different numbers of robots and tasks. In real-time and stability experiments, the average running time of the proposed algorithm does not exceed 6.09s. Its standard deviation and Wilcoxon rank-sum test results have statistical significance. It can formulate economical and efficient task allocation schemes. It provides effective technical support for the refined O&M of PV power stations.

During the feature extraction process of point cloud retrieval, mainstream methods typically form global shape descriptors through hierarchical aggregation of local neighborhood features. Their perception of object structural information mainly relies on indirect inference from the spatial distribution of surface points, lacking direct exploitation of explicit structural priors such as point cloud skeletons. To address the problem of insufficient structural information utilization in existing point cloud retrieval methods when extracting global shape features, this paper proposes a skeleton‑prior enhanced structure‑aware point cloud retrieval network. The method first introduces point cloud skeletons as structural priors, enhances the structural expressiveness of point cloud features through a dual mechanism of explicit fusion and implicit guidance, and designs an adaptive feature aggregation module to aggregate multi‑scale features, thereby forming the final global descriptor. Specifically, the proposed method consists of two modules. First, a dual‑branch feature fusion module. This module extracts skeleton point clouds from the input point cloud, and then employs two independent PointNet++ branches to extract multi‑scale local features from the original point cloud and the skeleton, respectively. At each scale, using skeleton features as keys and values and point cloud features as queries, a multi‑scale cross‑attention mechanism is adopted to weightedly integrate the skeleton structural information into the point cloud features. Meanwhile, a contrastive learning task is constructed: the combination of a cropped point cloud and the complete skeleton serves as the anchor, the combination of the complete point cloud and the skeleton serves as the positive sample, and samples from other categories within the batch serve as negative samples. The contrastive loss implicitly guides the model to learn structural consistency, forming an“explicit fusion + implicit guidance” dual structural enhancement mechanism. Second, a multi‑scale local adaptive aggregation descriptor (MVLAAD) module. This module consists of the Vector Local Adaptive Aggregation Descriptor (VLAAD) and multi‑scale aggregation enhancement. Based on a lightweight Transformer decoder, VLAAD takes the local feature sequence of the input point cloud as keys and values, and the initial generic cluster centers as queries. Through iterative updates via multi‑layer cross‑attention, it dynamically generates personalized cluster centers that adapt to each input sample. A momentum update strategy further combines the generic centers with the personalized centers, balancing adaptability and stability. Subsequently, using the updated cluster centers, soft assignment weights are computed for the refined multi‑scale features and the residuals are aggregated to produce global descriptors at three scales. Finally, a gating mechanism enhances the features and outputs the compact global descriptor. In addition, during training a dynamic weight adjustment strategy is adopted, combining classification loss, triplet loss, and contrastive loss. The training emphasizes contrastive learning in the early stage and shifts focus to triplet loss in the later stage, thereby reinforcing both structure perception and discriminative learning. Experimental results show that the proposed method achieves 82.6% mAP on the ModelNet40 dataset, outperforming the state‑of‑the‑art method CF3D by 1.3%, and attains 84.6% mAP on ShapeNet, exceeding existing methods. Ablation studies verify the effectiveness of each module: the baseline PointNet++ achieves 62.0% mAP, while the full model increases the performance to 82.6%. A lightweight version reduces the number of parameters to 20.34M and further improves mAP to 84.7%. Robustness experiments demonstrate that the method performs robustly under moderate sparsity, low‑level noise, and mild occlusion, but its performance degrades significantly under extreme degradation conditions. In summary, the proposed skeleton‑prior enhanced structure‑aware point cloud retrieval network addresses the deficiency of structural information utilization in existing methods through explicit skeleton prior fusion and implicit contrastive learning guidance. The designed MVLAAD module dynamically generates personalized cluster centers, thereby improving the discriminability of the global descriptor.

This paper addresses foreign object detection in UAV-based transmission line inspection, where target scales vary greatly, small objects are easily missed, background interference is strong, occlusion is common, and airborne edge devices have limited computing resources. Based on YOLOv8n, a multi-branch lightweight algorithm named MBL-YOLO is proposed for detecting typical foreign objects such as bird nests, kites, waste objects, and balloons. In inspection images, these objects often overlap visually with conductors, insulators, towers, and natural backgrounds. Their boundaries may be unclear, which can lead to localization offsets and insufficient confidence in compact detectors. The goal of the proposed method is to improve representation for multi-scale foreign objects and complex backgrounds without significantly increasing model complexity, while also meeting the real-time, low-power, and lightweight deployment requirements of UAV edge platforms. In terms of network structure, MBL-YOLO improves YOLOv8n from three aspects: backbone feature extraction, cross-scale feature fusion, and detection-head lightweighting. First, a Mixed Dynamic Fusion Block is embedded into the C2f structure to build C2F-MDFB. Through dynamic kernel weighting and multi-scale depthwise separable convolution branches, this module adaptively adjusts the contributions of features with different receptive fields. It enables the network to simultaneously focus on large objects such as bird nests, slender objects such as kite strings, small objects such as balloons, and local features of irregular waste objects. Residual connections, channel mixing, and normalization further enhance the interaction between low-level details and high-level semantics, reducing missed detections caused by blurred object boundaries or partial occlusion. Second, a weighted bidirectional feature pyramid, BI-FPN, is introduced into the neck. Learnable weights are used to fuse hierarchical features such as P3, P4, and P5. Shallow features retain edge texture and location information, while deeper semantic constraints suppress background noise from conductors, towers, vegetation, and sky regions, improving small-object localization and class discrimination under complex backgrounds. Finally, to address repeated parameters and redundant inference in the original multi-scale independent detection branches, a shared convolutional detection head named Detect-LSCD is designed. It replaces repeated convolutions with two shared convolutional layers and uses GroupNorm to stabilize feature distributions under small-batch inference, reducing parameter scale and computational cost while maintaining multi-scale detection capability. Experiments are conducted on a self-built transmission line foreign object dataset. The dataset contains 4,200 images and 8,207 fully annotated targets, covering mountain and urban backgrounds. It includes 2,103 bird nests, 2,560 kites, 1,648 waste objects, and 1,896 balloons, with 2,940 training images, 840 validation images, and 420 test images. Under the same training and testing conditions, MBL-YOLO achieves 97.5% Precision, 97.1% Recall, 97.3% mAP50, and 70.4% mAP50-95, with 2.08M parameters, 5.8 GFLOPs, and an inference speed of 175.6 FPS. Compared with YOLOv8n, mAP50-95 increases by 1.9 percentage points, the parameter size and computational cost decrease by about 30.9% and 29.3%, respectively, and FPS increases from 168.6 to 175.6. Compared with YOLOv9t, YOLOv10n, YOLOv11n, YOLOv12n, Gold-YOLO, YOLO-world, D-Fine-N, and DEIM-D-Fine-N, MBL-YOLO obtains the highest mAP50-95 while maintaining the lowest computational cost and a relatively small parameter size, indicating that the improvement does not rely on model stacking. Ablation experiments show that C2F-MDFB, BI-FPN, and Detect-LSCD improve dynamic representation, weighted fusion, and lightweight prediction, respectively, and provide complementary benefits when combined. To verify cross-scene generalization, experiments are also conducted on the VisDrone2019 public dataset. MBL-YOLO achieves 41.7% Precision, 31.1% Recall, 31.2% mAP50, and 18.6% mAP50-95, improving YOLOv8n by 4.3, 5.5, 4.3, and 2.1 percentage points, respectively. Visual results show that the model reduces missed detections, false detections, and redundant boxes in small-object, multi-class, dense-background, and occlusion scenes, demonstrating good feature retention and scene transfer capability. For engineering feasibility, MBL-YOLO is deployed on a UAV platform equipped with NVIDIA Jetson TX2 and a ZED2 stereo camera. The model is exported to ONNX and then optimized with TensorRT 8.2 through FP16 half-precision quantization and operator fusion to generate a native TX2 inference engine. With a 640×640 input resolution, batch size of 1, and TX2 Max-P mode, the end-to-end latency of MBL-YOLO is 38.5 ms per frame, and the overall frame rate remains stable at 26 FPS. Under the same conditions, YOLOv8n reaches 55.1 ms and 18 FPS, so the practical deployment speed is improved by about 44.4%. The average power consumption of TX2 and ZED2 working together is about 9.2 W, less than 3% of the UAV flight power consumption. Overall, MBL-YOLO improves detection accuracy, recall, and real-time performance while reducing parameters and computation. It is suitable for UAV edge inspection platforms and can provide an engineering-practical solution for automatic foreign object recognition, abnormal-object warning, and intelligent operation and maintenance of transmission lines, while also laying a foundation for future multimodal perception and online inspection system integration.

Equipment document knowledge acquisition and utilization are critical to equipment analysis and decision support. However, multi-source heterogeneous equipment documents generally exhibit complex layout structures, significant cross-modal mixed content, dense formulas and tables, and fragmented semantic expression, which makes traditional methods insufficient for complex structure restoration, uni-fied parsing of heterogeneous content, and knowledge-driven question answering. To address these issues, a large language model-based method, EKADE, is proposed for equipment-document knowledge analysis and decision enhancement. The method constructs a dual-stage decoupled framework of global layout analysis and local content recognition. Specifically, it first identifies the spatial distribution, category attributes, and hierarchical relationships of text paragraphs, images, formulas, and tables from a global page perspective. It then performs fine-grained parsing and structural reconstruction of local text, formulas, and tables, and reorganizes page-level results into a unified standardized Markdown document. On this basis, a retrievable structured knowledge fragment repository is constructed, and a retrieval-augmented generation mechanism is introduced to support knowledge-driven question answering for equipment performance understanding, tactical application analysis, and specification interpretation. Experimental results show that the proposed method achieves an average F1 score of 86.0% on page layout parsing, a Character Detection Matching (CDM) score of 86.7% on formula recognition, and a Tree Edit Distance-based Similarity (TEDS) score of 90.7% on table recognition. In the decision question-answering task, it outperforms comparative methods in Recall, BLEU, ROUGE-L, and accuracy. The results demonstrate that EKADE effectively improves the structured parsing capability, knowledge organization capability, and question-answering reasoning accuracy for complex equipment documents, and provides an effective so-lution for intelligent equipment knowledge modeling and decision support.

Existing voice-face cross-modal matching methods often suffer from limited channel-wise feature discrimina¬tion and indistinct distinction of hard identities. To address these issues, this paper proposes an improved voice-face matching framework that incorporates a multi-scale channel attention mechanism and enhanced contrastive learning. Building upon an adaptive identity-weighted center baseline, we design a parallel main-fine-coarse branch channel attention module with explicit-implicit statistical fusion, greatly en¬hancing the activation of informative channels in both mel-spectrograms and facial feature maps. Furthermore, a bidirec¬tional InfoNCE contrastive loss is introduced and jointly optimized with the original cross-entropy and cross-modal N-pair losses under the guidance of adaptive identity weighting, which further widens the separation of challenging identities. Extensive experiments on the VoxCeleb and VGGFace overlapping dataset demonstrate that the proposed method con¬sistently outperforms state-of-the-art approaches such as SVHF and DIMNet in cross-modal verification, matching, and retrieval tasks. Compared with the baseline, it achieves 2.1% and 2.4% AUC gains in voice-to-face and face-to-voice verification, respectively.In addition, ablation studies confirm the effectiveness and complementarity of the multi-scale channel attention and contrastive learning components.

In real-world clinical consultations, patient chief complaints are typically expressed verbally and subsequently recorded as text by physicians. Physicians need to comprehensively use both patients’ spoken descriptions and corresponding textual records to judge and classify symptoms, thereby providing a basis for subsequent clinical decision-making. However, this task still faces several challenges. Speech information is susceptible to environmental noise and individual pronunciation differences, while textual records are unable to fully reflect speech-related expressive features such as speaking rate, pauses, and intonation. Meanwhile, patient chief complaints are usually colloquial, subjective, and unstructured, and semantic boundaries among different symptom categories may be ambiguous. These factors make it difficult for single-modality methods to achieve satisfactory classification performance. To address these issues, a dynamic weight decision fusion-based multimodal symptom classification method, named DWDF-MSC, is proposed to fully exploit the complementarity between textual and speech information and improve the accuracy and robustness of symptom classification. The proposed method mainly consists of three stages: multimodal feature extraction, preliminary classification, and adaptive gated decision fusion. In the multimodal feature extraction stage, a text branch and a speech branch are constructed to model patient chief complaint text and speech data in parallel. In the text branch, global semantic features and local lexical features are simultaneously extracted based on the clinical pre-trained language model Bio_ClinicalBERT, and the two are fused through a heterogeneous textual feature fusion module, thereby enhancing the model’s representation capability for the overall semantics of chief complaints and local symptom-related keywords. In the speech branch, an audio spectrogram Transformer is used to extract temporal acoustic representations from speech, thereby supplementing speech expressive information that is difficult to capture from textual records. In the preliminary classification stage, the text branch and the speech branch output initial classification results through their respective classification modules, allowing the two modalities to independently perform symptom judgment. In the final classification stage, an adaptive gated decision fusion strategy is designed to dynamically generate fusion weights according to the features of different samples. The initial classification results from the text and speech branches are then weighted and fused to obtain the final symptom classification result. Unlike simple feature concatenation or fixed-weight fusion, this strategy can adaptively adjust the contribution of the two modalities in the final decision according to sample differences, thereby enhancing the influence of discriminative information on the classification result and improving the classification stability of the model in complex chief complaint scenarios. Experimental results on a public medical dataset show that DWDF-MSC achieves 82.43%, 87.44%, and 81.52% in Accuracy, Precision, and F1-score, respectively, outperforming most mainstream baseline models across all metrics. The comparison of multimodal fusion schemes further demonstrates that the proposed dynamic weight decision fusion achieves better classification performance than feature-level fusion. In the ablation study, the complete DWDF-MSC model achieved relative improvements of 4.25% and 7.60% in Accuracy and F1-Score, respectively, compared with the variant that only employed heterogeneous text feature fusion, thereby demonstrating the effectiveness of the speech branch and the adaptive gated decision fusion mechanism. The McNemar test results show that the p-values between DWDF-MSC and multiple comparison methods are less than 0.0001, indicating that the differences in classification results between DWDF-MSC and these comparison methods are statistically significant. The anti-noise performance experiments demonstrate that DWDF-MSC can still maintain relatively stable classification performance under different signal-to-noise ratio conditions. In summary, DWDF-MSC can effectively fuse textual and speech information from patient chief complaints, improve model classification performance, and provide a feasible multimodal method for intelligent symptom classification based on patient chief complaints.

Cultural relic 3D reconstruction is an important technical support for the digital preservation, virtual exhibition, and digital restoration of cultural heritage. Compared with modeling approaches such as structured-light scanning and laser scanning, which rely on specialized equipment and controlled acquisition environments, multi-view image-based 3D reconstruction methods have the advantages of low acquisition cost, flexible operation, and low deployment requirements, making them more suitable for cultural relic digitization in museum exhibition spaces. However, images captured in real museum collection scenes are often affected by complex backgrounds, glass reflections, uneven illumination, local occlusions, and limited shooting viewpoints. As a result, the target relic is highly intertwined with display platforms, walls, and other background regions in image space. Although the original 3D Gaussian Splatting (3DGS) method can achieve efficient training and real-time rendering through explicit Gaussian primitives, it is mainly designed for complete scene modeling and lacks a semantic focusing mechanism for cultural relic subjects. Consequently, redundant background point clouds and non-target Gaussians are likely to participate in optimization, increasing GPU memory consumption, training time, and model size. In addition, abnormal elongation and artifacts may occur around object boundaries, affecting the stable representation of the geometric shape and texture details of cultural relics. To improve the accuracy and efficiency of cultural relic subject reconstruction in complex museum collection environments, a high-fidelity 3D reconstruction method based on segmentation priors and 3DGS is proposed, in which two-dimensional subject segmentation results are introduced into the 3D Gaussian modeling process. The Segment Anything Model is used to generate subject masks of cultural relics from multi-view images. Combined with camera poses and sparse point clouds estimated by SfM, 3D points are projected onto the corresponding mask planes. Points that consistently fall into background regions are removed according to multi-view semantic consistency, thereby obtaining cleaner and more compact subject point clouds from the initialization stage. During Gaussian optimization, a mask-guided constraint is introduced to restrict the color reconstruction loss to the cultural relic target region, enabling parameter updates to focus on the subject geometry, surface texture, and local details while reducing the interference of background regions in the optimization process. To address abnormal elongation of Gaussian ellipsoids caused by insufficient sampling and depth discontinuities near cultural relic contours, an edge pruning strategy based on geometric morphological constraints is designed. Morphologically abnormal Gaussian primitives near object boundaries are identified and removed according to the major-to-minor axis ratio, suppressing “black spike” artifacts and edge noise diffusion while enhancing the continuity, compactness, and visual stability of subject boundaries. Experimental results on public datasets, including Tanks&Temples, Mip-NeRF 360, LERF, and LLFF, as well as a self-built cultural relic dataset, demonstrate that the proposed method achieves favorable overall performance in reconstruction accuracy, structural consistency, and perceptual quality. On the public datasets, the average PSNR, SSIM, and LPIPS reach 32.99 dB, 0.977, and 0.026, respectively. On the self-built cultural relic dataset, the average PSNR, SSIM, and LPIPS reach 35.48 dB, 0.983, and 0.027, respectively. Compared with the original 3DGS and related methods, including LightGaussian, 3DGSR, 2DGS, Perceptual-GS, and FCGS, the proposed method produces clearer subject contours and more stable texture representations under complex background conditions. Resource consumption comparisons and ablation experiments show that segmentation prior-guided point cloud filtering and edge pruning can jointly reduce redundant background Gaussians and alleviate contour artifacts, while significantly lowering training costs without compromising reconstruction quality. Compared with the original 3DGS, the training time is reduced by approximately 60%, GPU memory consumption by approximately 40%, and model size by approximately 50%, providing a feasible solution for low-cost, efficient, and high-fidelity 3D reconstruction of museum cultural relics under uncontrolled acquisition conditions.

As a distributed learning architecture, federated learning allows clients to perform global model training without sharing local data, which can effectively balance the contradiction between privacy and efficiency. However, its distributed characteristics also make it vulnerable to data poisoning attacks. Malicious clients tamper with local training data to inject biased or wrong updates into the global model, so as to achieve the purpose of reducing the accuracy of the model or manipulating the behavior of the model under specific inputs. Label inversion attack, as a classical method in data poisoning attack, is simple to implement and has low computational cost, and only requires modifying local data labels without modifying features. It is difficult to be found by conventional statistical analysis, but it can effectively reduce the accuracy of the global model or complete the backdoor implantation. In order to improve the accuracy of the global model and the overall security of the system in federated learning, the model update parameters uploaded by each local client are usually screened and filtered from the server before the global model aggregation, so as to accurately identify the malicious client behavior and perform robust aggregation to resist data poisoning attacks. To solve the above problems, this paper proposes a Label Flipping Attack Defense Algorithm (LFADA) oriented to federated learning, which aims to improve the accuracy and security of the model in the face of data poisoning. LFADA uses the log-likelihood score mechanism, and first flattens and reduces the dimension of the updated parameters of each client model to construct the sample set. Secondly, the Gaussian Mixture Model (GMM) is used to model the processed updated parameter sample set. Then, the update probability of each client is quantified by the Log-Likelihood Score (LLS), and the "normality" score of each client is obtained. Then, based on the current parameter set, the filtering threshold score is set according to the required quantile, and the clients below the score are considered as malicious clients. The update parameters of all malicious clients are eliminated, and only the update parameters of the clients that pass the filtering are aggregated, so as to realize the unsupervised anomaly detection and filtering of client updates and the secure aggregation of the global model. This paper conducts experiments on the MNIST dataset, Fashion-MNIST dataset and CIFAR-10 dataset, and uniformly sets a Convolutional Neural Network (CNN) containing three convolutional blocks as the basic model to carry out label flipping attacks respectively. The experiments of model accuracy and attack success rate show that LFADA can effectively resist label flipping attacks when the proportion of malicious clients is 0.1, 0.2, 0.3 and 0.5, and LFADA still performs well when the proportion of malicious clients is 0.5. Compared with nine mainstream algorithms such as Multi-Krum, Median, Foolsgold and Lfighter, the accuracy of the model using LFADA is increased by 3.28%, 3.38% and 2.62% on average, while the attack success rate is kept low as a whole. Among them, it is lower than 3% on MNIST and Fashion-MNIST datasets, and significantly lower than most methods on CIFAR-10 dataset, which can maintain similar performance to the federal average FedAvg scheme model in the environment without poisoning attack. In terms of algorithm stability, the overall process of federated learning using LFADA can maintain overall stability throughout the training phase, especially on the more complex Fashion-MNIST and CIFAR-10 datasets, there is no large fluctuation up and down, and the overall amplitude is controllable, which is obviously stable compared with other algorithms. Time overhead experiments show that compared with the comparison algorithms, LFADA's time overhead is significantly reduced under the premise of ensuring the same accuracy and attack success rate.

Weakly supervised semantic segmentation commonly utilizes Class Activation Maps (CAMs) to generate pseudo-labels for training segmentation networks. However, since CAMs originate from image-level classification tasks, their responses tend to concentrate on salient object regions, resulting in incomplete foreground activation. Meanwhile, CAM responses in object boundary regions and complex structural areas are unstable, which easily introduces pseudo-label noise and limits segmentation performance improvement. To address the above problems, this paper proposes a single-stage weakly supervised semantic segmentation method with semantic uncertainty region enhancement. First, this paper designs a contrastive learning module based on semantic uncertainty regions. The module fuses multiple uncertainty cues to perform fine-grained modeling of semantic uncertainty regions in CAMs, thereby enhancing foreground activation completeness. Second, this paper introduces a dynamic adaptive Gaussian denoising module. The module applies dynamic threshold adjustment and Gaussian mixture denoising strategies to adaptively identify and progressively remove pseudo-label noise, thereby suppressing pseudo-label noise. Experimental results demonstrate that, under the supervision of image-level labels only, the proposed method achieves mIoU scores of 72.2% and 72.8% on the PASCAL VOC 2012 validation set and test set respectively, and achieves 42.5% on the MS COCO 2014 dataset. Ablation experiments further demonstrate that the separate introduction of the semantic uncertainty region contrastive learning module and the dynamic adaptive Gaussian denoising module improves mIoU by 1.6% and 2.5% respectively. The results verify the effectiveness of both modules in enhancing foreground completeness and suppressing pseudo-label noise, thus improving the overall segmentation performance of the model.

ew-Shot Object Detection (FSOD) aims to detect novel objects using only a few annotated samples. Although existing meta-learning-based FSOD methods have achieved performance improvement through the collaboration of query and support branches, they still encounter three primary bottlenecks. First, fixed multi-scale feature fusion strategies overlook the relative importance of features across different resolutions, making it difficult to handle multi-scale objects; second, class-level prototypes generated via simple average pooling fail to capture the complex intra-class structures and are susceptible to noise interference; third, the semantic scarcity of the support set leads to semantic bias during query-prototype interactions, resulting in false positives or missed detection. To address these challenges, this paper proposes a Feature Fusion and Semantic Enhancement (FFSE) model for few-shot object detection. Built upon the Meta R-CNN framework, FFSE enhances detection performance through three synergistic core modules from three dimensions, i.e., feature fusion, prototype representation, and feature modulation. First, the Dynamic Weight-based Feature Fusion (DWFF) module adaptively assigns weights to features of different scales, effectively integrating local textures with global semantics to strengthen the model's perception of multi-scale objects. Second, to improve class-level prototype quality, the Prototype Graph Network (PGN) mechanism is introduced. By leveraging the message-passing mechanism of graph neural networks, PGN achieves higher-order semantic enhancement, producing refined prototypes with stronger discriminative power and robustness. Finally, inspired by feature linear modulation, the Feature Modulation Driven by Support set (FMDS) module decomposes the fused query features across multiple receptive fields. It then utilizes refined prototypes to generate dynamic scaling and shifting factors for channel-wise affine transformations. The scaling factors amplify target-related features, while the shifting factors guide the query feature distribution toward the support semantic space, effectively correcting semantic biases and enhancing object saliency. Quantitative evaluations have been conducted on PASCAL VOC and MS COCO benchmarks. On PASCAL VOC, FFSE outperforms the baseline method across all three novel-class splits, specifically, for 5-shot and 10-shot settings, the nAP50 increases by at least 2.2%. On the challenging MS COCO dataset, FFSE achieves at least a 5% improvement in nAP over the baseline. Results from multiple experimental runs (mean and standard deviation) demonstrate that FFSE maintains low performance fluctuations and superior robustness while improving accuracy compared to some methods. Qualitative analysis compared to some methods on PASCAL VOC dataset further indicates that FFSE can effectively handle heavy occlusion, diverse tiny objects, and high-similarity background interference, significantly reducing cross-category misidentification. In conclusion, the extensive experimental results validate the effectiveness of the proposed FFSE model. In the future, we will explore the advanced attention mechanisms at the pixel level to effectively suppress background noise for improving the performance of FSOD.

Oral diseases seriously affect public health, and timely and effective diagnosis and treatment are of great significance for reducing the risk of disease progression. Conventional diagnosis of oral diseases mainly relies on manual interpretation of imaging data by experienced clinicians, which is often time-consuming and may overlook lesions with blurred boundaries. Therefore, image segmentation techniques are needed to assist the clinical diagnosis of dental diseases. Dental panoramic radiographs can present the overall morphology of teeth and jawbone structures in a single image and are commonly used in clinical dental diagnosis. However, due to low gray-level contrast, blurred lesion boundaries, noise, and artifact interference commonly present in dental panoramic radiographs, multi-class dental disease segmentation, including dental caries, periapical periodontitis, furcation involvement, and impacted teeth, remains highly challenging. To address these issues, this paper proposes Teeth-Net, a network for multi-class dental disease segmentation in dental panoramic radiographs. Based on the TransUNet architecture, Teeth-Net introduces targeted improvements in three key stages: feature extraction, feature reconstruction, and skip connections. In the feature extraction stage, a Cross-Scale Pyramid Fusion Module (CPFM) is introduced to optimize the original encoder. Multi-scale features are extracted through parallel group convolutions with different receptive fields, and the correlations among features at different scales are modeled using a cross-scale attention mechanism, thereby enhancing the model’s ability to capture small lesions and alleviating the loss of detailed features. In the feature reconstruction stage, a Parallel Multi-Kernel Pooling Module (PMKP) is designed to extract local details and global contextual information in parallel through multi-scale max pooling and average pooling. Channel compression and feature fusion are then performed to provide richer semantic information for the decoder. At each skip connection, a Spatial-Channel Collaborative Attention module (SCCA) is embedded to adaptively filter shallow features transmitted from the encoder through spatial and channel attention mechanisms, suppress background noise interference, and improve the quality of cross-layer feature fusion between the encoder and decoder. Comparative and ablation experiments are conducted on a self-built dental panoramic radiograph dataset. The experimental results show that Teeth-Net achieves a mean Dice coefficient, Hausdorff Distance (HD), precision, and recall of 84.22%, 18.546 mm, 94.13%, and 95.96%, respectively. Compared with the baseline TransUNet model, the mean Dice coefficient, precision, and recall are improved by 3.34, 2.89, and 4.21 percentage points, respectively, while the HD value is reduced by 6.869 mm. These results indicate that the proposed method achieves significant improvements in overall segmentation accuracy, boundary consistency, and lesion detection capability. To further evaluate the generalization ability and cross-dataset adaptability of the model, external tests are conducted on two public-source datasets. On the re-annotated MICCAI 2023 STS external test set, Teeth-Net achieves a mean Dice coefficient, HD value, precision, and recall of 80.26%, 19.520 mm, 92.58%, and 93.41%, respectively. Compared with the baseline TransUNet model, the mean Dice coefficient, precision, and recall are improved by 3.32, 4.33, and 3.89 percentage points, respectively, while the HD value is reduced by 6.705 mm. On the public Multi-Center Dental Panoramic Radiography Image (MCDP) dataset, Teeth-Net achieves a mean Dice coefficient, HD value, precision, and recall of 88.99%, 12.126 mm, 90.61%, and 92.45%, respectively. Compared with the baseline TransUNet model, the mean Dice coefficient, precision, and recall are improved by 3.83, 4.03, and 3.33 percentage points, respectively, while the HD value is reduced by 7.222 mm. The results on the self-built dataset and the two external test datasets demonstrate that Teeth-Net achieves better segmentation accuracy, boundary delineation ability, and cross-domain adaptability than the baseline TransUNet model under different data sources and imaging conditions. The proposed method can provide effective technical support for the assisted diagnosis of multi-class dental diseases in dental panoramic radiographs.

Image steganography embeds secret data into cover images for covert communication and is an important topic in information and multimedia security. With social-media compression, format conversion, image resampling, and active steganalysis, traditional methods face more complex scenarios. Existing deep steganography methods mainly focus on visual imperceptibility and embedding capacity, but pay insufficient attention to message confidentiality, integrity authentication, and error tolerance after extraction. Thus, covert transmission, content protection, and robust recovery are still difficult to unify. To address these problems, this paper proposes an information-encryption-driven high-security image steganography model. It jointly designs authenticated encryption, error-correction coding, key-controlled scrambling, and a deep steganographic network to achieve secure, covert, and reliable transmission over complex channels. At the payload generation stage, an “encryption-error correction-scrambling” defense scheme is built. HKDF-SHA256 is used to derive encryption and scrambling keys. AES-GCM provides authenticated encryption and generates ciphertext with confidentiality and integrity verification. Reed-Solomon coding is introduced to provide symbol-level error correction for the steganographic channel. If the number of erroneous symbols after inverse scrambling is within the RS correction radius, the correct data packet can be recovered. If the error exceeds the correction ability or AES-GCM authentication fails, decryption is stopped to avoid incorrect plaintext output. In addition, CSPRNG-based position and bit scrambling reduce payload correlation and statistical bias, while a sparse bitmap controls embedding positions and reduces structural clues exploitable by steganalyzers. At the embedding stage, a hybrid U-Net combining MS-DiSpAC and ViT is designed. MS-DiSpAC extracts texture, edge, and local structural features through multi-scale convolution, and uses dilated spatial attention to enlarge the receptive field while preserving resolution. It guides high-entropy payloads into complex texture regions. ViT supplements global context modeling and improves long-range dependency representation. The network generates stego images through a residual perturbation map and an intensity map, balancing image fidelity and recovery stability under high payloads. A WGAN discriminator with Wasserstein distance is further used for adversarial distribution alignment, making stego images statistically closer to cover images and reducing detection by SRNet, ZhuNet, and other steganalyzers. Experiments are conducted on ImageNet, COCO, and Visual Genome, including performance, generalization, payload whitening, robustness, and ablation tests. Metrics include PSNR, MS-SSIM, LPIPS, BER, ESR, ACC.1, ACC.2, and Dacc. At 0.4 bpp, the proposed method achieves 38.65 dB PSNR, 0.975 MS-SSIM, 0.036 LPIPS, and 99.14% bit recovery accuracy on ImageNet. Payload whitening results show that, after AES-GCM encryption, RS coding, and dual scrambling, single-bit entropy increases from 0.8932 to 0.9998, and average absolute autocorrelation decreases from 0.1285 to 0.0028. The final payload is close to random. Compared with representative methods, the proposed model achieves a better balance among visual fidelity, information recovery, and anti-steganalysis capability. It also maintains high recovery success under complex distortions within the RS correction range, providing a feasible solution for high-security image steganography in real network environments.

Highway traffic flow during holidays exhibits significant spatiotemporal heterogeneity, making accurate short-term Origin-Destination (OD) flow prediction a key technology for enhancing the intelligent level of road network management. To address issues such as the high-dimensional sparsity of OD data, complex spatiotemporal dependencies, and holiday pattern shifts, this paper proposes a short-term highway OD flow prediction method based on spatiotemporal fusion and holiday adjustment, and constructs a Dual-stage Spatio-Temporal Fusion Network (DSTF) model. First, a spatiotemporal feature extraction architecture for multi-source data fusion is designed: a dual-branch Graph Attention Network (GAT) is used to extract and fuse spatial features representing macroscopic travel correlations from the OD perspective and microscopic node state dependencies from the entrance and exit flow perspectives. Then, a gated fusion module combining a Temporal Convolutional Network (TCN) and a Convolutional Long Short-Term Memory network (CNN-LSTM) collaboratively captures both the short-term fluctuations and long-term periodic trends of traffic flow. Simultaneously, a Cross-Attention mechanism is introduced to achieve multi-task collaborative prediction of entrance flow, exit flow, and baseline OD flow. To adapt to the special travel patterns during holidays, the model adopts a two-stage training strategy: the first stage trains the baseline prediction model using sufficient and stable non-holiday data; the second stage introduces a lightweight Sequence-to-Sequence (Seq2Seq) holiday adjustment module, focusing on learning the deviation of holiday patterns from the baseline, and performs adaptive fine-tuning on the baseline OD flow predictions. Experimental results based on real highway toll data show that the proposed DSTF model significantly outperforms various baseline models across multiple evaluation metrics in holiday short-term OD prediction tasks, achieving reductions of 11.7% in MAE and 12.2% in RMSE compared to the best baseline model STGCN in 1-step prediction, demonstrating higher prediction accuracy, stronger robustness, and superior scenario adaptability.

Hypergraph Neural Networks (HGNNs) have emerged as a prominent research direction in recommender systems due to their capability to model high-order interactions and integrate multi-source heterogeneous information. Unlike traditional Graph Neural Networks (GNNs), which are limited to pairwise relationships, HGNNs employ hyperedges to capture high-order associations among an arbitrary number of nodes, thereby preserving complex semantics in user–item interactions, such as many-to-many relationships, group structures, and multimodal information.This paper first outlines the general pipeline of HGNN-based recommendation from four aspects: data input, hypergraph construction, representation learning, and recommendation generation. Furthermore, recent advances in HGNN-based recommendation are systematically reviewed from two perspectives: hypergraph construction strategies and feature propagation mechanisms. These developments are analyzed across multiple application scenarios, including sequential recommendation, multi-behavior recommendation, social recommendation, multimodal recommendation, and group recommendation.In the context of sequential recommendation, existing studies have explored various hypergraph construction strategies, including local dependency modeling based on session interactions, global co-occurrence modeling, cross-session collaborative modeling, and multi-scale spatiotemporal dynamic modeling. Correspondingly, feature propagation mechanisms that integrate hypergraph attention-based denoising and self-supervised contrastive learning have been investigated to enhance temporal representation learning. These approaches help overcome the “neighborhood limitation” inherent in conventional GNNs and enable more accurate modeling of users’ evolving interests and long-range dependencies.For multi-behavior recommendation, hypergraph construction strategies are categorized into behavior-specific modeling, unified behavior modeling, and temporal behavior modeling. Feature propagation mechanisms, such as cascaded dependency propagation, behavior-aware attention, and cross-view contrastive learning-based denoising, have been developed to address data sparsity in target behaviors, facilitate semantic alignment across behaviors, and support knowledge transfer.In social recommendation, existing works focus on hypergraph construction methods based on homophily-driven dual views, heterogeneous semantic relationships, and privacy-preserving mechanisms. Feature propagation strategies incorporating trust-aware attention and dual-channel gated fusion have been proposed, which extend beyond traditional pairwise social modeling and contribute to capturing complex group influence and high-order social structures.For multimodal recommendation, hypergraph construction strategies include modality-specific separation, collaborative semantic association, and multimodal hypergraph optimization. Feature propagation mechanisms based on modality-specific convolutional aggregation and cross-modal contrastive alignment have demonstrated effectiveness in reducing modality noise and enabling high-order reasoning within a unified semantic space, thereby improving representation quality.In group recommendation, hypergraph construction approaches involve multi-view hierarchical alignment, group structure-aware optimization, and tripartite relationship modeling. Feature propagation mechanisms that incorporate cross-level feedback and attention-based aggregation better align with the inherent “inclusion” relationships within groups and provide an effective solution for alleviating cold-start issues in dynamic group scenarios. Despite these advancements, several challenges remain in HGNN-based recommendation. Dynamic hypergraph models often face difficulties in meeting the requirements of real-time recommendation. High-order aggregation may introduce information resolution loss, while noisy pseudo-hyperedges can adversely affect model robustness. In addition, the computational and storage complexity of hypergraphs poses scalability challenges in large-scale applications. Furthermore, issues related to interpretability and fairness in recommendation results remain insufficiently addressed.To address these challenges, this paper discusses several promising research directions for future HGNN-based recommendation systems, including representation learning based on generative self-supervised disentanglement, lightweight and efficient training and inference frameworks, robustness enhancement via causal inference, scenario-aware multimodal fusion, and collaborative integration with large language models. These directions are expected to provide valuable insights for advancing research in this field.

Video virtual try-on technology aims to accurately transfer target garments onto human subjects in videos while maintaining high consistency between body motion and garment appearance, serving as a core technology in fields such as e-commerce, virtual reality, and short-video creation. However, existing technical frameworks still face significant challenges in balancing generation quality and computational efficiency. Traditional Generative Adversarial Network (GAN)-based methods often rely on optical flow estimation for garment warping, which are highly prone to texture distortion and visual artifacts when handling complex motions. In recent years, U-Net-based diffusion models have achieved high-fidelity generation by introducing a garment reference branch. However, when such dual-branch architectures are migrated to larger and more expressive Diffusion Transformer (DiT) backbones, they introduce substantial parameter redundancy and VRAM overhead. Furthermore, existing methods typically inject static garment features repeatedly during the denoising process of each frame. This not only significantly exacerbates the computational burden but also, due to the lack of natural temporal correlation in static features, makes it difficult for models to maintain spatiotemporal coherence during non-rigid deformations, resulting in severe flickering artifacts. To address the aforementioned challenges regarding the adaptability, training efficiency, and resource consumption of DiT architectures in video virtual try-on tasks, this study proposes a lightweight framework named OIE (Once is Enough). The OIE framework adopts a novel single-branch strategy featuring first-frame guidance and one-time injection, effectively decoupling garment editing from temporal generation tasks. First, during the garment appearance injection stage, a pre-trained high-fidelity image virtual try-on model, FiT-DiT, is utilized to precisely edit the video's initial frame, yielding results integrated with fine-grained garment textures. Second, to maximally preserve the temporal priors of the DiT model, only the edited first frame is embedded as the starting token into the latent feature sequence of the backbone network. This avoids the dense cross-branch feature interaction modules typical of traditional dual-branch architectures, achieving zero structural modification to the backbone. Additionally, to address the loss of background layout information caused by human motion, this method designs a lightweight background encoder that smoothly accumulates background information into the backbone features via a mask guider. Finally, during the fine-tuning stage, Low-Rank Adaptation (LoRA) is applied to all self-attention, cross-attention, and feed-forward network (FFN) modules of the DiT, enabling dynamic regulation of the large-scale parameter model with an extremely low number of trainable parameters. Experiments conducted on the ViViD and VVT datasets yield quantitative evaluation results demonstrating that, in terms of efficiency, OIE introduces only a 0.50% additional parameter overhead, with FLOPs and FPS remaining virtually unchanged. Its performance significantly surpasses dual-branch methods such as MagicTryOn (15.11% parameter increase) and ViViD (157.10% parameter increase). Regarding quality metrics, OIE achieves competitive video quality scores under both paired and unpaired settings on the ViViD dataset, attaining a VFIDp of 9.3983 and a VFIDu of 17.0831, significantly leading existing mainstream methods. Ablation studies confirm that high-quality first-frame guidance effectively suppresses error generation in the early stages of synthesis, improving the SSIM metric to 0.8466. Through its decoupling strategy, the OIE framework effectively resolves the computational burden of DiT architectures in video generation, achieving an excellent balance among garment fidelity, temporal coherence, and computational efficiency. This method demonstrates that leveraging strong temporal priors within a single-branch architecture can replace high-frequency feature injection, offering a highly valuable lightweight pathway for high-resolution and real-time video editing tasks.

Industrial steel surface defects exhibit pronounced anisotropic texture characteristics with large intra-class variation, yet existing real-time detection methods lack effective perception mechanisms for such directional local patterns in feature pyramid networks. This paper proposes a direction-aware heterogeneous convolution feature enhancement method based on RT-DETR, incorporating three key technical contributions. First, a Direction-Aware Sparse Convolution (DASC) kernel is designed, which partitions input channels into five directional groups with fixed sparse spatial masks to achieve parallel perception of right, left, down, up, and center directional textures at approximately 11.5% of the computational cost of equivalent standard convolutions. Second, a Direction-aware Interaction and Refinement (DIR) bottleneck is constructed using an expand-activate-compress dual-layer DASC structure to realize hierarchical fusion of directional features across channels, forming the complete Lightweight Feature Enhancement module with Cross-stage 3 modules for RT-DETR (LFEC3-RT). Third, a Cross-scale FPN Consistent Deployment (CFPD) strategy is introduced, globally deploying LFEC3-RT across all four fusion positions in the feature pyramid to eliminate cross-scale feature style inconsistency caused by selective deployment. Experiments on the NEU and GC10-DET steel surface defect benchmarks demonstrate that the proposed method achieves 76.3% mAP@0.5 on NEU (+2.2% over RT-DETR-R18 baseline) and 64.4% mAP@0.5 on GC10-DET (+3.3% over baseline), achieving competitive or superior performance over YOLOv11m on both datasets while requiring only 56.0 GFLOPs and 19.8M parameters. Ablation studies confirm that increasing direction count from 1 to 5 raises mAP from 74.4% to 76.3%, expansion ratio λ=4 is optimal, and CFPD global deployment outperforms selective deployment by +0.9% mAP.

There is a profound structural contradiction between the surge in the parameter scale of large language models and the limited physical resources of edge terminals, which restricts their large - scale implementation. Traditional cloud-centralized inference highly depends on network transmission and faces high communication latency, making it difficult to meet the dual requirements of extremely low latency and strict data privacy in scenarios such as autonomous driving and intelligent healthcare. However, edge physical hardware, ranging from microcontrollers to edge gateways, has great heterogeneity, and the general cloud - side compression schemes are difficult to be directly applied. Therefore, based on the heterogeneous physical constraints of edge devices, this paper systematically reviews the technical system of efficient compression and software - hardware collaborative deployment of large models for the edge side. First, this paper analyzes the underlying mechanisms of three core compression technologies, namely model quantization, parameter pruning, and knowledge distillation, in edge scenarios. In terms of quantization, although post - training quantization has deployment agility, it faces the problem of representation collapse caused by the abnormal long - tail activation of large language models. Although quantization - aware training has a certain degree of robustness, it is limited by the lack of retraining computing power at the edge. In terms of pruning, this paper demonstrates the actual energy - efficiency advantages of structured pruning on hardware with limited memory access bandwidth and points out that the high theoretical compression rate of unstructured pruning is easily offset by the index addressing overhead of general - purpose edge chips. In terms of distillation, traditional shallow parameter alignment has the risk of feature loss and bias amplification when crossing the capacity gap between the teacher and edge student models. Overall, single compression technologies show an obvious diminishing marginal return effect under extreme constraints. Second, to alleviate the performance bottleneck of single technologies, this paper summarizes a multi - level hybrid compression paradigm driven by both model architecture and physical scenarios. Three core optimization links are systematically sorted out: the serial pipeline strategy aiming at a high physical compression rate, which is suitable for real - time inference at edge gateways; the deeply coupled joint optimization flow for a strict trade - off between energy efficiency and accuracy, which synchronously updates quantization, pruning, and low - rank decomposition within the same framework and is suitable for mobile terminals with limited power consumption; and the distillation - driven mechanism for the deployment of large - parameter models, which uses teacher priors to guide structure reshaping and quantization. This multi - level paradigm effectively expands the multi - dimensional trade - off space among model scale, computing power consumption, and fidelity. Furthermore, in the face of a wide range of computing power and energy consumption levels, this paper constructs a four - layer software - hardware collaborative design mechanism of "system - model - operator - instruction". It clearly points out that the focus of collaborative optimization needs to be dynamically shifted according to physical base constraints: at the system level, it focuses on resource - aware scheduling and task distribution in the cloud - edge environment; at the model level, it relies on hardware - aware architecture search to achieve structural adaptation; at the operator level, it promotes cross - layer fusion and memory access locality reconstruction; at the instruction level, it focuses on custom - extended instructions for specific micro - architectures (such as RISC - V) to precisely control the underlying energy consumption. Combined with the full - chain deployment process of model conversion, compilation reconstruction, and memory management (such as SwapNet), this mechanism effectively maps compression algorithms to underlying physical execution and improves the comprehensive utilization efficiency of heterogeneous computing power. Finally, this paper prospectively points out the future research challenges in the field of edge - intelligent lightweighting. It emphasizes that the robust compensation mechanism for ultra - low bit - width (4bit and below), hardware - adaptive dynamic semi - structured pruning, and effective knowledge transfer for the deep - level logical reasoning of large models are the core directions to overcome the current lightweighting bottlenecks. At the same time, it is urgent to build a hardware - agnostic unified toolchain based on deep - learning compilers to eliminate the deployment barriers of fragmented heterogeneous devices. Through a systematic review of technologies, this paper provides a solid theoretical support and reference guide for the development of an edge - intelligent ecosystem with low latency and strong privacy.

To address the difficulty of deploying backdoor defenses in Machine Learning as a Service (MLaaS) black-box scenarios, this paper proposes an adaptive image preprocessing defense framework that relies solely on natural image statistics priors. The framework performs multi-dimensional feature analysis on input images to construct a backdoor risk quantification metric. According to the risk level, it dynamically selects and combines multi-level processing operations—including compression–reconstruction, geometric transformations, color perturbations, and dynamic random sequences—to disrupt the activation conditions of potential backdoor triggers. A quality feedback mechanism is introduced to balance defense effectiveness and visual usability. Experiments on the GTSRB, CIFAR-10, and MINI-ImageNet datasets evaluate five representative attacks, namely BadNets, Blended, WaNet, reflection attacks, and WaveAttack, which cover explicit patches, global blending, geometric warping, physical reflection, and frequency-domain perturbations. The results show that the proposed method reduces the average attack success rate to below 10% while preserving the model’s normal classification performance (with an average accuracy drop of no more than 3.5%). Notably, the suppression effect on WaveAttack is significant, achieving a success rate as low as 2.38%. Ablation studies confirm the critical role of the adaptive strategy and the quality feedback mechanism in performance improvement, and the framework exhibits stable performance across three datasets of varying scales, demonstrating strong generalization. This research provides an efficient and practical adaptive backdoor defense solution for MLaaS black-box services.

Remote sensing image change detection aims to precisely localize land cover changes by comparatively analyzing the spatiotemporal evolution information contained in bi-temporal imagery, and has become a core task in fields such as dynamic monitoring of land resources, urban expansion assessment, and disaster emergency response. However, influenced by multiple factors including complex terrain interference, variations in illumination conditions, seasonal vegetation succession, and sensor imaging noise, change regions often exhibit characteristics such as substantial scale variations, discrete spatial distribution, and ambiguous boundary delineation. Existing change detection models suffer from insufficient exploitation of multi-scale information and inadequate extraction of deep global semantic correlations, rendering it challenging for these models to effectively discriminate genuine land surface changes from pseudo-changes, thereby constraining their discrimination accuracy in open-scene scenarios. To address the aforementioned limitations, a Multi-level Loss-assisted Siamese Network (MLLA_SiaNet) for remote sensing image change detection is proposed. The model adopts a weight-sharing Siamese architecture to extract multi-dimensional features from bi-temporal images separately, and generates hierarchical feature maps through a multi-level differential encoder. To overcome the linear limitations inherent in conventional differencing methods, we introduce a multi-angle difference representation strategy coupled with a channel-spatial hybrid attention mechanism, and design a Differential Fusion Module (DFM) to acquire high-quality difference features, thereby achieving adaptive suppression of background interference and precise focusing on genuine change characteristics. To compensate for the deficiency in global semantic representation, we integrate a spatial pooling pyramid with a Gaussian pyramid and propose a Deep Semantic Pyramid (DSP) module to construct multi-level semantic aggregation features, effectively expanding the receptive field and strengthening long-range contextual dependency modeling. During the decoding stage, the model employs a progressive upsampling strategy combined with a feature fusion mechanism to hierarchically restore spatial details, thereby enabling the reconstruction of high-resolution prediction maps. Furthermore, we introduce a deeply supervised Multi-level Loss-assisted (MLA) strategy to optimize the training process; by imposing auxiliary constraints on the outputs of each decoder layer, this strategy ensures consistency between local edge information and global contextual semantics, thereby constructing an end-to-end feature learning framework. To systematically validate the effectiveness of the proposed model, comparative experiments are conducted and results are comprehensively analyzed on two publicly available benchmark datasets, namely SYSU-CD and LEVIR-CD. On the SYSU-CD dataset, MLLA_SiaNet achieves an F1-score of 82.13%, outperforming seven other comparative methods and surpassing the second-best method, SFEARNet, by 1.3 percentage points; its precision and recall attain optimal values of 83.42% and 80.88%, respectively, achieving simultaneous improvement in both precision and recall metrics. On the LEVIR-CD dataset, MLLA_SiaNet achieves a precision of 89.48%, fully demonstrating the effectiveness of the proposed method in suppressing pseudo-change factors such as illumination variations, shadow effects, and seasonal vegetation changes; the F1-score of our model on the LEVIR-CD dataset reaches 85.87%, outperforming other state-of-the-art methods including SFEARNet (precision 84.89%), BIT (precision 82.80%), and IFN (precision 82.29%).Both quantitative and qualitative analyses of the experimental results demonstrate that the model exhibits superior robustness under varying spatial resolutions and complex land cover conditions. Ablation studies further corroborate the advantages of the DFM, DSP, and MLA modules in enhancing overall model performance, and the effectiveness of each architectural stage is empirically verified through analysis of the visualized response feature maps. In summary, this study mitigates the impacts of several critical challenges in remote sensing image change detection tasks, including insufficient multi-scale feature interaction, weak correlation modeling of global semantic information, and difficulties in suppressing pseudo-change interference. Future work will focus on lightweight model deployment, multi-temporal sequence modeling, and self-supervised pre-training techniques, as well as expanding systematic evaluations of model robustness across diverse application scenarios.