Search Results (204)

Red–green–blue and depth (RGB-D) semantic segmentation combines appearance cues from RGB images with geometric information from depth maps, but sensor noise, missing measurements, and boundary-inconsistent depth responses can introduce conflicting evidence during cross-modal fusion. We propose the Structure-Prior-Guided Network (SPGNet), a dual-branch, multi-stage framework that follows a correction-before-fusion strategy. At each feature scale, SPGNet estimates a learned structure prior from cross-modal agreement and discrepancy. The Cross-Modal Correction Module (CCM) uses this prior to regulate bidirectional information transfer, suppressing unreliable responses while retaining complementary cues. The Dual-branch Enhancement Fusion Module (DEF) then enhances the corrected RGB and depth features and integrates them through shared-representation-guided interaction, after which a lightweight multi-scale decoder produces the segmentation output. Under a unified training and evaluation protocol, SPGNet achieved three-run mean Intersection over Union (mIoU) scores of 50.845% on NYU Depth V2 and 48.457% on SUN RGB-D. Compared with the best reproduced baseline on each dataset, SPGNet improved mean mIoU by 2.111 and 0.899 percentage points, respectively. These results suggest that separating reliability-oriented correction from multimodal fusion can limit the propagation of unreliable cross-modal responses and improve indoor RGB-D semantic segmentation performance. Full article

Aviation ground safety and protective devices are critical for flight safety; however, their unintentional retention on aircraft after maintenance remains a persistent risk. Existing deep learning-based approaches for aviation safety have predominantly followed a reactive paradigm, detecting FOD on runways or inspecting the aircraft for inadvertently retained tools post-maintenance. In contrast, this paper advocates a proactive philosophy: using a neural network to recognize and inventory all ground safety and protective devices immediately after maintenance closure, thereby preventing retention incidents at their source. However, realizing this proactive verification is technically challenging—object detection for these devices often suffers from loss of fine-grained detail due to downsampling and inherently sparse semantic information of the targets. To this end, we propose an Information Retention and Feature Screening Synergistic Network (RS-Net) grounded in information bottleneck theory. The network comprises a main branch that enhances discriminative features through attention-guided screening, and an auxiliary branch, used only during training, that preserves fine-grained spatial details via information-retentive convolutions. A Dual-State Region Refinement Module (DRM) provides configurable support for both branches, decoupling the conflicting objectives of background compression and detail preservation. Experiments on a self-constructed dataset collected from real airline maintenance operations demonstrate that RS-Net substantially outperforms the strong YOLOv9 baseline, achieving gains of 4.531% in F1-score, 2.533% in mAP_0.5, and 1.429% in mAP_0.5:0.95. Cross-dataset experiments further validate its strong generalization capability. Full article

Cilium is among the most widely deployed Container Network Interfaces (CNIs), serving as the default CNI in the Google Kubernetes Engine. It extends standard Kubernetes NetworkPolicy (KNP) with two additional types—CiliumNetworkPolicy (CNP) and CiliumClusterwideNetworkPolicy (CCNP)—each with distinct semantics. When all three coexist in a cluster, the resulting composition is difficult to reason about formally, leading to misconfiguration and security incidents. Existing verification tools, KANO and VeriKube, address subsets of the problem but share two critical limitations: neither provides a formal denotational semantics that precisely characterizes the three-layer composition, nor a canonical representation enabling policy-equivalence checking with completeness guarantees. We close this gap with three contributions. First, we develop the first formal denotational semantics for Cilium’s three-layer composed policy—KNP (additive), CNP (deny-wins), CCNP (cluster-override)—and prove that the composite function is Hyper-Rectangular Piecewise-Constant (HRPC)-like. Second, we construct a Reduced Ordered Interval Decision Diagram (ROIDD) for the composite policy space and prove a canonicity theorem—canonical for a fixed field order—enabling policy-equivalence checking as structural isomorphism in O(|ROIDD|) time. Third, we develop certified conflict-detection algorithms for shadow, redundancy, and cross-layer conflict anomalies across all three layers with formal proofs of soundness and completeness. Experimental evaluation on synthetic policies confirms zero mismatches between ROIDD evaluation and ground-truth brute force; detection of shadow, redundancy and cross-layer anomalies at precision and recall of 1.000, scored against exhaustive enumeration of the entire packet space; agreement with a live Cilium v1.19.5 data plane on every probe of a scenario built to exercise each clause of the composite semantics; ROIDD compression ratios of 5–15× over the unshared decision tree on the compressed evaluation domain; and low-microsecond (0.74–1.95 µs) per-packet lookup latency that is independent of policy size. A native C++ implementation, evaluated on the same policy dataset, reconstructs the identical decision-diagram structure and classifies each packet in under 60 ns—roughly 30× faster than the Python reference—confirming that sub-microsecond classification is inherent to the algorithm rather than an artifact of the implementation language. Full article

UAV-based insulator defect detection faces persistent challenges of multi-scale defect variation, background clutter, small-target missed detections and redundant detection-head computation. This paper proposes BEAD-Net, a real-time insulator defect detection network built upon YOLOv11n with four targeted improvements. First, an Hourglass Symmetric Residual Attention (HSRA) module replaces the standard bottleneck components within C3k2, expanding the multi-scale receptive field and suppressing background interference via a symmetric hourglass dilation schedule and channel attention recalibration. Second, a Bidirectional Diffusion Feature Pyramid Network (BDFPN) built upon the Group-wise Selective Feature Integrator (GSFI) employs group-wise adaptive gating and two-level bidirectional propagation to mitigate semantic dilution during cross-scale fusion. Third, a Task-Decoupled Asymmetric Detection Head (TDAH) concentrates spatial modeling in the regression branch while simplifying classification to lightweight $1\times 1$ operations, reducing parameter redundancy and alleviating inter-task gradient conflict. Finally, Focaler-PIoU2 integrates linear interval mapping with a normalized corner distance penalty to improve boundary regression for slender insulator structures. On the IDID dataset, BEAD-Net achieves mAP@0.5 of 83.98% and mAP@0.5:0.95 of 63.82% at 275.04 FPS with 2.49 M parameters, outperforming the baseline YOLOv11n by 2.83 and 1.55 percentage points and surpassing all compared state-of-the-art methods. Additional validation on the CPLID benchmark shows that BEAD-Net remains the top performer, confirming that the proposed improvements are not specific to a single dataset. Full article

Multi-Level Governance (MLG) frameworks effectively diagnose the complexity of regional renewable energy transitions but lack operational mechanisms for institutional implementation. This study develops a hybrid evidence-based architecture that integrates the Viable System Model (VSM) with computational intelligence and objective multi-criteria decision analysis. Methodologically, a systematic review following the PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) protocol of 339 peer-reviewed articles (2015–2025) feeds a Latent Dirichlet Allocation (LDA) model that extracts K = 30 strategic topics (semantic coherence $C_v$ optimized over K = 5–40), operationalizing System 4 environmental sensing. In parallel, a 1 km² pixel-based spatial model integrates a National Conflict Index (INC, 2019–2024) with technical feasibility layers (Global Wind Atlas v4.0, Solargis, Servicio Geológico Colombiano) and applies CRITIC (CRiteria Importance Through Intercriteria Correlation) objective weighting and TOPSIS (Technique for Order of Preference by Similarity to Ideal Solution) prioritization (System 3 control). Robustness is confirmed through $\pm 10$–$20\%$ weight perturbation (Spearman > 0.92). Empirically, the framework reveals a localization paradox: approximately 68% of optimal wind zones (>9 m/s at 100 m hub height) in the Colombian Caribbean overlap with the highest national conflict quartile, narrowing a theoretical capacity exceeding 100 GW (50 GW offshore wind, 30 GW onshore wind, 42 GW solar PV, 1.17 GW geothermal) to roughly 24 GW of governance-viable capacity. Scenario calibration (Accelerated 80/20, Balanced 50/50, Justice-Oriented 30/70 technical/conflict weighting) demonstrates that System 5 normative orientation materially reshapes territorial prioritization. The framework advances VSM from a qualitative diagnostic metaphor to a reproducible governance architecture for high-variety regional contexts. Full article

Self-supervised knowledge distillation is widely used to compress reusable encoders, but an untrusted distillation implementation can itself become an attack surface. We study an algorithm-level threat in which the teacher encoder and user-visible distillation dataset remain unchanged, while malicious code internally generates trigger-bearing views and optimizes an additional backdoor objective. To instantiate this threat, we propose soft gradient collaboration (SGC), which combines distribution-alignment-based distillation, target-representation-based backdoor design, and conflict-avoidance gradient collaboration to reduce interference with benign representation transfer while embedding a trigger-to-target association in the student encoder. Experiments on CIFAR-10 and STL-10 show that SGC maintains competitive downstream accuracy and effective non-target attack success. Quantitative CKA, feature-distribution, and class-structure analyses further indicate that SGC retains clean representations closer to benign distillation than fixed scalarization or removal of distribution alignment. Its no-defense attack success is not the highest among the compared attacks; instead, its main empirical advantage is stronger residual attack persistence after MIMIC, MKD, and SSLDefender. Under SSLDefender, SGC retains 9.12% non-target ASR on CIFAR-10 and 9.06% on STL-10, the highest residual values among the compared attacks. Additional experiments with a compact ResNet-18 student, multiple target classes and trigger configurations, and a supplemental CIFAR-100 setting broaden the empirical evaluation across student capacity, target semantics, trigger configurations, and label-space complexity. These results show that security assessment of self-supervised distillation should include executable training logic in addition to model weights and visible data. The concealment considered here is limited to dataset-only inspection and clean-output validation; SGC is not claimed to evade source-code auditing, runtime data-flow monitoring, or training-log inspection. Full article

Intelligent transportation and aerial remote sensing scenes suffer from complex scene variations, abundant miniature targets and unpredictable out-of-distribution obstacles, which brings tough mathematical challenges to open-world detection tasks. Conventional detection algorithms lack rigorous frequency-domain separation and spatial constraint mathematical formulations, resulting in severe tiny-object feature attenuation, inefficient multimodal feature matching and catastrophic forgetting during incremental category iteration. To solve these mathematical bottlenecks, this paper constructs the TPCA-Net model built upon frequency decomposition and spatial dependency mathematical modelling. The entire framework consists of four fixed core modules: High-Frequency-Aware Multi-Scale Feature Enhancement (HSE), Reparameterized Adaptive Text–Visual Alignment (RTA), Double Wildcard Spatial Dependency Fusion (WSF), and Incremental Forgetting-Free Dual-Path Detection (DPD). From the mathematical perspective, the HSE module adopts discrete cosine transform-based filtering equations to split high-frequency object details from low-frequency background signals and establishes cross-attention spatial constraint formulas to make up for missing contextual information of small targets. The RTA module introduces low-rank decomposition mathematical optimization and reparameterized tensor fusion rules to realize domain-adaptive text embedding calibration and zero-cost cross-modal mapping at the inference stage. The WSF module constructs dual-wildcard self-supervised mathematical loss to finish unsupervised unknown-object identification and builds decoupled semantic–spatial fusion equations to improve the positioning precision of novel targets. The DPD module designs two sets of independent optimization objective functions and category-freezing incremental mathematical constraints to avoid conflicting parameter updates and eliminate forgetting defects in new-class expansion. Validated on COCO, DOTA and AI-TOD datasets, TPCA-Net achieves 56.0% AP on COCO, 79.30% mAP on DOTA, and 40.5% overall AP with 28.7% small-object AP on AI-TOD while delivering an inference throughput of 101.2 FPS on the Tesla T4 edge GPU. The proposed method outperforms existing mainstream open-world detection algorithms in tiny-object and rare-category recognition while maintaining efficient inference speed. Full article

Diffusion models have become a central paradigm for image generation because they combine stable optimization, high-fidelity synthesis, and controllability through iterative denoising. However, precise image generation in practical settings requires more than text prompts. Here, precision means measurable satisfaction of semantic, spatial, structural, identity, interaction, and domain constraints rather than pixelwise reproduction alone. This survey examines collaborative control methods for diffusion-based image generation from a system-level perspective. We distinguish ordinary controllable diffusion from collaborative control, then review theoretical foundations, conditional generation, single-condition extensions, multi-condition fusion, conflict mediation, controller–evaluator loops, scalability, applications, and evaluation protocols. The discussion emphasizes how control signals are represented, injected, scheduled, and evaluated along the denoising trajectory. It also compares representative methods in terms of controllability, computational overhead, scalable inference, and task-oriented metrics. We identify three continuing challenges: robust coordination under conflicting heterogeneous conditions, fine-grained control under few-step sampling, and reliable benchmarks that jointly measure constraint satisfaction and efficiency. Overall, collaborative control reframes precise diffusion generation as a coordinated modeling and optimization problem involving models, conditions, schedulers, and evaluators. Full article

In the context of the large-scale integration of high-proportion renewable energy into power grids, Power Conversion Systems (PCS) and Static Var Generator (SVG), as the core power electronic devices for ensuring grid frequency stability and power quality, demand high-level safety and reliability. To tackle challenges such as the wide range of defect sizes in PCS and SVG, the low recognition accuracy for microscopic fuzzy defects, and complex background interference, this paper presents a lightweight and high-precision defect recognition model (RD-YOLO) based on the latest YOLOv11 benchmark. First, an improved Mosaic algorithm is introduced. This algorithm utilizes conflict relationship tables to preserve physical context semantics during traditional non-discriminative data augmentation. Second, to surmount the limitations of scale-aware feature extraction, the YOLOv11 is re-engineered within the backbone network by integrating a Res2Net multi-scale cascaded mechanism. This enhances the network’s capacity to capture both fine-grained defect features and large-scale defect boundaries. Third, Focal Loss is employed for difficult sample detection. Nonlinear gradient modulation is utilized to guide the model to focus on ambiguous defect edges. Finally, the Soft-NMS post-processing strategy significantly enhances the regression accuracy in densely corroded regions. Experimental validation on a self-developed dataset consisting of 8500 high-resolution PCS and SVG defect images reveals that the enhanced RD-YOLO attains an average precision of 89.6% and a frame inference rate of 98 FPS (in RTX 3090), offering robust technical support for intelligent visual maintenance in renewable energy facilities. Full article

Accurate land-use polygon mapping for forest-oriented natural resource monitoring requires both image-based class prediction and the reconciliation of heterogeneous geospatial semantics. In operational mapping, land survey, forestry survey, and natural resource monitoring datasets often differ in classification systems, management objectives, boundary rules, and mapping scales, causing semantic conflicts when overlaid or forced into one-to-one categories. To address this issue, this study proposes a topology-aware land-use polygon-mapping framework that integrates multi-source semantic representation learning, semantically guided relation learning, and topology-constrained polygon optimization. In the experiments, remote sensing imagery provides visual evidence, the Third National Land Survey (TNLS) and forestry survey (FS) datasets are encoded as source-specific auxiliary semantic priors, and the natural resource integrated monitoring (NRIM) data serve only as reference labels for training and evaluation. Rather than resolving cross-source conflicts using predefined rules, the framework learns a unified land-use representation, transforms semantic boundary cues into a vertex-edge topology graph, and reconstructs GIS-compatible polygons through topology-constrained polygon optimization. In a representative forest–agricultural landscape, the method achieved an mIoU of 79.86%, an APLS of 56.82%, and a TOPO-F1 of 54.56%. These results suggest that learnable semantic harmonization and topology-aware polygon generation can improve the semantic consistency and vector reliability of land-use products for forest and natural resource monitoring. Full article

Foundation models such as the Segment Anything Model 2 (SAM2) have demonstrated strong performance in image segmentation; however, their application to industrial defect detection faces significant challenges due to the substantial domain gap between natural and industrial images, insufficient sensitivity to fine-grained high-frequency structures, and reliance on manual prompts. To address these issues, this study proposes a Hierarchical Cross-Branch Frequency-Aware Network (HCFNet) to adapt SAM2 for prompt-free industrial defect segmentation. First, a Gated Adapter is introduced into the frozen SAM2 encoder, enabling efficient cross-domain transfer without massive parameter retraining, thereby effectively preserving the pre-trained visual priors. Secondly, a Laplacian-enhanced Auxiliary Branch is designed to explicitly amplify high-frequency components, compensating for the inherent perception limitations of the Transformer backbone and significantly awakening the model’s sensitivity to subtle defects like micro-cracks. Finally, a Cross-branch Multi-scale Fusion Module is proposed to seamlessly align and integrate global semantic information with local structural details in a unified manner, resolving heterogeneous feature distribution conflicts. Extensive experiments on the MVTec AD and VisA datasets demonstrate that the proposed method consistently outperforms SAM2-based baselines in terms of mIoU and mDice. This study establishes an effective approach for leveraging foundation models in automated industrial inspection and is expected to drive advancements in precise defect perception technologies. Full article

Long-term LiDAR maps support urban remote sensing, infrastructure inventories, change analysis, digital twins, and map-based localization. Most map-cleaning pipelines, however, output a single retained map, forcing transient observed-dynamic artifacts and movable but currently stationary scene content into the same keep/remove decision. This paper formulates layer assignment within long-term LiDAR map maintenance as a layered representation problem. The resulting output comprises a long-term static (LTS) layer, a potentially dynamic (PD) layer, and a removed observed-dynamic (OD) set. For geometry-only accumulated maps with per-frame semantic predictions, the proposed layer-assignment framework combines map-to-scan geometric inconsistency, dual-timescale map-side semantic memory, PD admission, and PD-preserving recovery within the geometric rejection set. Geometric evidence first identifies OD candidates and rejection regions, while semantic memory converts framewise class predictions into map-side movability evidence for LTS/PD assignment. Experiments on SemanticKITTI under a controlled in-sequence LTS/PD/OD protocol show that this framework maintains high LTS-layer retention, improves PD retention, and reduces PD over-cutting while maintaining competitive OD suppression with predicted semantic input. Mechanism analyses further show that the dual-timescale semantic memory helps preserve PD evidence under conflicting or temporally sparse semantic predictions, while PD-preserving recovery provides boundary compensation with an explicit OD-suppression trade-off. Collapsing the output to a single retained map also improves the balance between static and PD retention relative to single-layer cleaning baselines. Qualitative Apollo SouthBay transfer observations further indicate that the same configuration can produce a three-layer map output without retuning, with binary annotations used only to support map-quality checks. Full article

UAV-borne visual sensors provide high-resolution aerial observations for low-altitude scene understanding, urban monitoring, traffic observation, emergency inspection, and infrastructure assessment. However, semantic perception from UAV visual sensor data remains challenging because aerial images often contain dense small objects, elongated road structures, fragmented boundaries, scale variations caused by flight-altitude changes, oblique viewpoints, and strict onboard or edge computational constraints. To address these challenges, this paper proposes MEKD-UAVSeg, a lightweight semantic perception framework based on conflict-suppressed heterogeneous expert distillation. During training, a Transformer-based semantic expert provides global contextual understanding and region-level class consistency, while a Mamba-based spatial expert provides complementary structural guidance for roads, roofs, boundaries, and other continuous aerial structures. Both experts are used only during training, and the final inference model remains a compact CNN-based segmentation network. In addition, UAV-aware density and hard-region priors are designed to emphasize small-object-dense areas, boundary-sensitive regions, rare classes, and uncertain aerial categories. A conflict-suppressed reliability routing strategy is further developed to reduce inconsistent supervision between heterogeneous experts and selectively transfer reliable knowledge to the student model. Experiments on UAVid and UDD6 demonstrate that the proposed framework achieves a favorable accuracy–efficiency trade-off compared with representative CNN-, Transformer-, Mamba-, and hybrid-based UAV segmentation methods, without introducing expert-induced inference complexity. Full article

CustomNerd is a modular framework to build domain-specific expertise systems using a common set of core components. The system is designed to support multiple domains and has already been deployed to instantiate expertise-driven systems in nutrition (DietNerd), news (NewsNerd), cloud technology (CloudNerd), and space science (SpaceNerd). CustomNerd incorporates specialized knowledge from domain-specific data sources that a user provides and processes the results of searches on those sources. The system consists of five main components: (i) Query Conversion, which converts user input into structured queries for one or more application-specific databases of validated articles; (ii) Extraction, which uses the structured queries to extract relevant data from those external databases; (iii) Filtering and Classification, which identifies and prioritizes relevant documents among those retrieved; (iv) Reliability Assessment, which evaluates the quality of the documents (e.g., statistical validity, absence of conflict of interest); and (v) Response Generation, which produces text outputs based on data sources considered to be relevant and reliable. In addition, the system can (i) generate questionnaires based on a user query, (ii) allow users to upload their own documents, and (iii) work with a variety of Large Language Model (LLM) engines. This paper describes how the system can be used to build an expertise-driven application through a running example, explains the system’s architecture, discusses related work, and evaluates CloudNerd on a 120-question Stack Overflow benchmark against OpenDeepResearch and AutoGPT using semantic textual similarity and RAGAS metrics with paired permutation tests. The results show that the RAG frameworks of CustomNerd and OpenDeepResearch both achieve better semantic fidelity than the agentic AutoGPT. Further, CustomNerd and OpenDeepResearch show complementary semantic advantages. Further, simply concatenating the two answers yields semantic benefits compared to the outputs of either one alone. Full article

Large language models can emit fluent text while leaving intermediate semantic structure implicit. We study whether explicit event-role and logical-primitive side-state can improve a pretrained decoder without damaging its language behavior. We introduce PAT-ER, a decoder architecture with a normal token stream, an event-role register stream, and a primitive register stream. The primitive stream is motivated by the view that logical primitives answer characteristic semantic questions, such as what licenses a conclusion, what conflicts with it, or why evidence is insufficient. Across eight seeds on the same Qwen3-0.6B backbone, replacing token-pooled auxiliary heads with typed PAT-ER registers improves primitive macro-F1 by 0.209 (95% CI [0.182, 0.237]) and role-to-primitive macro-F1 by 0.091 (95% CI [0.074, 0.110]) with no language-model loss cost. A generic-register control shows that this is not merely the effect of adding latent registers: typed PAT-ER improves over generic registers by 0.116 primitive macro-F1 and 0.110 role-to-primitive macro-F1, with both confidence intervals excluding zero. A warm-started model then recovers pretrained language quality (LM loss 1.344 versus 2.555 for the frozen-backbone register model) while retaining most side-state behavior. Finally, a decoupled interface mode produces robust schema-grounded function calls on 242 held-out prompts (Hermes parse 0.952, exact arguments 0.981, JSON validity 1.000, IDK F1 1.000) while base-mode side-state metrics remain byte-identical to the warm-start baseline. The model is not a theorem prover and does not achieve perfect unseen tool-name copying; the contribution is a measured architecture signal and a usable, guarded interface. Full article