In this study,an inverse design framework was established to find lightweight honeycomb structures(HCSs)with high impact resistance.The hybrid HCS,composed of re-entrant(RE)and elliptical annular re-entrant(EARE)honey...In this study,an inverse design framework was established to find lightweight honeycomb structures(HCSs)with high impact resistance.The hybrid HCS,composed of re-entrant(RE)and elliptical annular re-entrant(EARE)honeycomb cells,was created by constructing arrangement matrices to achieve structural lightweight.The machine learning(ML)framework consisted of a neural network(NN)forward regression model for predicting impact resistance and a multi-objective optimization algorithm for generating high-performance designs.The surrogate of the local design space was initially realized by establishing the NN in the small sample dataset,and the active learning strategy was used to continuously extended the local optimal design until the model converged in the global space.The results indicated that the active learning strategy significantly improved the inference capability of the NN model in unknown design domains.By guiding the iteration direction of the optimization algorithm,lightweight designs with high impact resistance were identified.The energy absorption capacity of the optimal design reached 94.98%of the EARE honeycomb,while the initial peak stress and mass decreased by 28.85%and 19.91%,respectively.Furthermore,Shapley Additive Explanations(SHAP)for global explanation of the NN indicated a strong correlation between the arrangement mode of HCS and its impact resistance.By reducing the stiffness of the cells at the top boundary of the structure,the initial impact damage sustained by the structure can be significantly improved.Overall,this study proposed a general lightweight design method for array structures under impact loads,which is beneficial for the widespread application of honeycomb-based protective structures.展开更多
Attribute-Based Encryption(ABE)has emerged as a fundamental access control mechanism in data sharing,enabling data owners to define flexible access policies.A critical aspect of ABE is key revocation,which plays a piv...Attribute-Based Encryption(ABE)has emerged as a fundamental access control mechanism in data sharing,enabling data owners to define flexible access policies.A critical aspect of ABE is key revocation,which plays a pivotal role in maintaining security.However,existing key revocation mechanisms face two major challenges:(1)High overhead due to ciphertext and key updates,primarily stemming from the reliance on revocation lists during attribute revocation,which increases computation and communication costs.(2)Limited universality,as many attribute revocation mechanisms are tailored to specific ABE constructions,restricting their broader applicability.To address these challenges,we propose LUAR(Lightweight and Universal Attribute Revocation),a novel revocation mechanism that leverages Intel Software Guard Extensions(SGX)while minimizing its inherent limitations.Given SGX’s constrained memory(≈90 MB in a personal computer)and susceptibility to side-channel attacks,we carefully manage its usage to reduce reliance while mitigating potential collusion risks between cloud service providers and users.To evaluate LUAR’s lightweight and universality,we integrate it with the classic BSW07 scheme,which can be seamlessly replaced with other ABE constructions.Experimental results demonstrate that LUAR enables secure attribute revocation with low computation and communication overhead.The processing time within the SGX environment remains stable at approximately 55 ms,regardless of the complexity of access policies,ensuring no additional storage or computational burden on SGX.Compared to the Hardware-based Revocable Attribute-Based Encryption(HR-ABE)scheme(IEEE S&P 2024),LUAR incurs a slightly higher computational cost within SGX;however,the overall time from initiating a data request to obtaining plaintext is shorter.As access policies grow more complex,LUAR’s advantages become increasingly evident,showcasing its superior efficiency and broader applicability.展开更多
Conventional lightweight refractory materials with low bulk density and more pores suffer from harsh corrosion and erosion in actual applications.A type of lightweight Al_(2)O_(3)-MgAl_(2)O_(4)aggregates with a core-s...Conventional lightweight refractory materials with low bulk density and more pores suffer from harsh corrosion and erosion in actual applications.A type of lightweight Al_(2)O_(3)-MgAl_(2)O_(4)aggregates with a core-shell structure was synthesized at 1750℃using a rolling granulation method.Microstructural evolution and properties of the spherical aggregates were systematically studied.Scanning electron microscope and X-ray computed tomography results confirmed that a continuous and dense MgAl_(2)O_(4)spinel shell structure with a thickness of 200-300μm was formed on the surface.The corrosion results indicated that the corrosion index of the core-shell aggregates exhibited a 60%enhancement when compared to Al_(2)O_(3)spherical.Moreover,Al_(2)O_(3)-MgAl_(2)O_(4)refractory materials,which are based on the lightweight Al_(2)O_(3)-MgAl_(2)O_(4)spherical aggregates,possessed a higher temperature modulus of rupture of 9.19 MPa,and the retention rate of residual flexural strength reached 70%after thermal shock testing.The above results showed an improvement of 129.75 and 44.28%compared with pure Al_(2)O_(3)aggregate samples,respectively.In addition,the MgAl_(2)O_(4)spinel shell could trap the Mn,Fe elements from infiltrated slag and transfer into(Mg,Fe,Mn)Al_(2)O_(4)spinel,infiltrated CaO reacts with Sample Al_(2)O_(3)matrix to form a calcium hexaluminate(CA6)isolation layer,and the above two reasons enhance the corrosion resistance of the material.The corrosion mechanism was elaborated in detail.展开更多
The strength-ductility trade-off in low-Mn lightweight steels is a significant challenge due to the low thermal stability of austenite and the presence ofδ-ferrite.Two types of low-Mn lightweight steels containing V ...The strength-ductility trade-off in low-Mn lightweight steels is a significant challenge due to the low thermal stability of austenite and the presence ofδ-ferrite.Two types of low-Mn lightweight steels containing V and NbVMo microalloying elements were developed by warm rolling.Among these,NbVMo steel demonstrated superior properties,achieving a tensile strength of~1.2 GPa and a product of strength and elongation exceeding 45 GPa%.In-depth mechanism analysis by atom probe tomography and quasi-in-situ electron backscatter diffraction revealed that different microalloying compositions influence the mechanical properties by strengtheningδ-ferrite,refining retained austenite and homogenizing matrix strain.In NbVMo steel,δ-ferrite strengthening is attributed to the synergistic effects of(V,Mo,Cr,Nb)C composite precipitation,fine NbC and MoC precipitates,and the solid solution strengthening of Mo.These mechanisms collectively contribute to a higher yield strength andδ-ferrite microhardness compared to V steel.Consequently,δ-ferrite and the surrounding matrix in NbVMo steel exhibit coordinated elongation during deformation,enhancing the ductility.The improved microstructural and strain uniformity in NbVMo steel mitigates stress concentration effects onδ-ferrite deformation and serves as a barrier that delays the transformation of retained austenite.In contrast,the retained austenite in V steel exhibits a blocky morphology with larger grain sizes,resulting in lower stability.Combined with localized stress concentrations due to non-uniform strain distribution,this leads to premature transformation of retained austenite to alleviate stress,ultimately impairing elongation and the continuity of strain hardening.Furthermore,the precipitation mechanisms of(V,Mo,Cr,Nb)C composite precipitates are elucidated.展开更多
As a fundamental component in computer vision,edges can be categorized into four types based on discontinuities in reflectance,illumination,surface normal,or depth.While deep CNNs have significantly advanced generic e...As a fundamental component in computer vision,edges can be categorized into four types based on discontinuities in reflectance,illumination,surface normal,or depth.While deep CNNs have significantly advanced generic edge detection,real-time multi-class semantic edge detection under resource constraints remains challenging.To address this,we propose a lightweight framework based on PiDiNet that enables fine-grained semantic edge detection.Our model simultaneously predicts background and four edge categories from full-resolution inputs,balancing accuracy and efficiency.Key contributions include:a multi-channel output structure expanding binary edge prediction to five classes,supported by a deep supervision mechanism;a dynamic class-balancing strategy combining adaptive weighting with physical priors to handle extreme class imbalance;and maintained architectural efficiency enabling real-time inference.Extensive evaluations on BSDS-RIND show our approach achieves accuracy competitive with state-of-the-art methods while operating in real time.展开更多
With the large-scale deployment of the Internet ofThings(IoT)devices,their weak securitymechanisms make them prime targets for malware attacks.Attackers often use Domain Generation Algorithm(DGA)to generate random dom...With the large-scale deployment of the Internet ofThings(IoT)devices,their weak securitymechanisms make them prime targets for malware attacks.Attackers often use Domain Generation Algorithm(DGA)to generate random domain names,hiding the real IP of Command and Control(C&C)servers to build botnets.Due to the randomness and dynamics of DGA,traditional methods struggle to detect them accurately,increasing the difficulty of network defense.This paper proposes a lightweight DGA detection model based on knowledge distillation for resource-constrained IoT environments.Specifically,a teacher model combining CharacterBERT,a bidirectional long short-term memory(BiLSTM)network,and attention mechanism(ATT)is constructed:it extracts character-level semantic features viaCharacterBERT,captures sequence dependencieswith the BiLSTM,and integrates theATT for key feature weighting,formingmulti-granularity feature fusion.An improved knowledge distillation approach transfers the teacher model’s learned knowledge to the simplified DistilBERT student model.Experimental results show the teacher model achieves 98.68%detection accuracy.The student modelmaintains slightly improved accuracy while significantly compressing parameters to approximately 38.4%of the teacher model’s scale,greatly reducing computational overhead for IoT deployment.展开更多
The Internet of Things(IoT)ecosystem is inherently heterogeneous,comprising diverse devices that must interoperate seamlessly to enable federated message and data exchange.However,as the number of service requests gro...The Internet of Things(IoT)ecosystem is inherently heterogeneous,comprising diverse devices that must interoperate seamlessly to enable federated message and data exchange.However,as the number of service requests grows,existing approaches suffer from increased discovery time and degraded Quality of Service(QoS).Moreover,the massive data generated by heterogeneous IoT devices often contain redundancy and noise,posing challenges to efficient data management.To address these issues,this paper proposes a lightweight ontology-based architecture that enhances service discovery and QoS-aware semantic data management.The architecture employs Modified-Ordered Points to Identify theClustering Structure(M-OPTICS)to cluster and eliminate redundant IoT data.The clustered data are then modelled into a lightweight ontology,enabling semantic relationship inference and rule generation through an embedded inference engine.User requests,transmitted via theConstrainedApplication Protocol(CoAP),are semantically enriched and matched to QoS parameters using Dynamic Shannon Entropy optimized with the Salp Swarm Algorithm.Semantic matching is further refined using a bidirectional recurrent neural network(Bi-RNN),while a State–Action–Reward–State–Action(SARSA)reinforcement learning model dynamically defines and updates semantic rules to retrieve themost recent and relevant data across heterogeneous devices.Experimental results demonstrate that the proposed architecture outperforms existing methods in terms of response time,service delay,execution time,precision,recall,and F-score under varying CoAP request loads and communication overheads.The results confirm the effectiveness of the proposed lightweight ontology architecture for service discovery and data management in heterogeneous IoT environments.展开更多
To address critical challenges in nighttime ship detection—high small-target missed detection(over 20%),insufficient lightweighting,and limited generalization due to scarce,low-quality datasets—this study proposes a...To address critical challenges in nighttime ship detection—high small-target missed detection(over 20%),insufficient lightweighting,and limited generalization due to scarce,low-quality datasets—this study proposes a systematic solution.First,a high-quality Night-Ships dataset is constructed via CycleGAN-based day-night transfer,combined with a dual-threshold cleaning strategy(Laplacian variance sharpness filtering and brightness-color deviation screening).Second,a Cross-stage Lightweight Fusion-You Only Look Once version 8(CLF-YOLOv8)is proposed with key improvements:the Neck network is reconstructed by replacing Cross Stage Partial(CSP)structure with the Cross Stage Partial Multi-Scale Convolutional Block(CSP-MSCB)and integrating Bidirectional Feature Pyramid Network(BiFPN)for weighted multi-scale fusion to enhance small-target detection;a Lightweight Shared Convolutional and Separated Batch Normalization Detection-Head(LSCSBD-Head)with shared convolutions and layer-wise Batch Normalization(BN)reduces parameters to 1.8M(42% fewer than YOLOv8n);and the FocalMinimum Point Distance Intersection over Union(Focal-MPDIoU)loss combines Minimum Point Distance Intersection over Union(MPDIoU)geometric constraints and Focal weighting to optimize low-overlap targets.Experiments show CLFYOLOv8 achieves 97.6%mAP@0.5(0.7% higher than YOLOv8n)with 1.8 M parameters,outperforming mainstream models in small-target detection,overlapping target discrimination,and adaptability to complex lighting.展开更多
Fabric defect detection plays a vital role in ensuring textile quality.However,traditional manual inspection methods are often inefficient and inaccurate.To overcome these limitations,we propose FD-YOLO,an enhanced li...Fabric defect detection plays a vital role in ensuring textile quality.However,traditional manual inspection methods are often inefficient and inaccurate.To overcome these limitations,we propose FD-YOLO,an enhanced lightweight detection model based on the YOLOv11n framework.The proposed model introduces the Bi-level Routing Attention(BRAttention)mechanism to enhance defect feature extraction,enabling more detailed feature representation.It proposes Deep Progressive Cross-Scale Fusion Neck(DPCSFNeck)to better capture smallscale defects and incorporates a Multi-Scale Dilated Residual(MSDR)module to strengthen multi-scale feature representation.Furthermore,a Shared Detail-Enhanced Lightweight Head(SDELHead)is employed to reduce the risk of gradient explosion during training.Experimental results demonstrate that FD-YOLO achieves superior detection accuracy and Lightweight performance compared to the baseline YOLOv11n.展开更多
The demand for extended electric vehicle(EV)range necessitates advanced lightweighting strategies.This study introduces a materials genome approach,augmented by machine learning(ML),for optimizing lightweight composit...The demand for extended electric vehicle(EV)range necessitates advanced lightweighting strategies.This study introduces a materials genome approach,augmented by machine learning(ML),for optimizing lightweight composite designs for EVs.A comprehensive materials genome database was developed,encompassing composites based on carbon,glass,and natural fibers.This database systematically records critical parameters such as mechanical properties,density,cost,and environmental impact.Machine learning models,including Random Forest,Support Vector Machines,and Artificial Neural Networks,were employed to construct a predictive system for material performance.Subsequent material composition optimization was performed using amulti-objective genetic algorithm.Experimental validation demonstrated that an optimized carbon fiber/bio-based resin composite achieved a 45%weight reduction compared to conventional steel,while maintaining equivalent structural strength.The predictive accuracy of the models reached 94.2%.A cost-benefit analysis indicated that despite a 15%increase in material cost,the overall vehicle energy consumption decreased by 12%,leading to an 18%total cost saving over a five-year operational lifecycle,under a representative mid-size battery electric vehicle(BEV)operational scenario.展开更多
Tomato is a major economic crop worldwide,and diseases on tomato leaves can significantly reduce both yield and quality.Traditional manual inspection is inefficient and highly subjective,making it difficult to meet th...Tomato is a major economic crop worldwide,and diseases on tomato leaves can significantly reduce both yield and quality.Traditional manual inspection is inefficient and highly subjective,making it difficult to meet the requirements of early disease identification in complex natural environments.To address this issue,this study proposes an improved YOLO11-based model,YOLO-SPDNet(Scale Sequence Fusion,Position-Channel Attention,and Dual Enhancement Network).The model integrates the SEAM(Self-Ensembling Attention Mechanism)semantic enhancement module,the MLCA(Mixed Local Channel Attention)lightweight attention mechanism,and the SPA(Scale-Position-Detail Awareness)module composed of SSFF(Scale Sequence Feature Fusion),TFE(Triple Feature Encoding),and CPAM(Channel and Position Attention Mechanism).These enhancements strengthen fine-grained lesion detection while maintaining model lightweightness.Experimental results show that YOLO-SPDNet achieves an accuracy of 91.8%,a recall of 86.5%,and an mAP@0.5 of 90.6%on the test set,with a computational complexity of 12.5 GFLOPs.Furthermore,the model reaches a real-time inference speed of 987 FPS,making it suitable for deployment on mobile agricultural terminals and online monitoring systems.Comparative analysis and ablation studies further validate the reliability and practical applicability of the proposed model in complex natural scenes.展开更多
Aiming at the problem of potential information noise introduced during the generation of ghost feature maps in GhostNet,this paper proposes a novel lightweight neural network model called ResghostNet.This model constr...Aiming at the problem of potential information noise introduced during the generation of ghost feature maps in GhostNet,this paper proposes a novel lightweight neural network model called ResghostNet.This model constructs the Resghost Module by combining residual connections and Adaptive-SE Blocks,which enhances the quality of generated feature maps through direct propagation of original input information and selection of important channels before cheap operations.Specifically,ResghostNet introduces residual connections on the basis of the Ghost Module to optimize the information flow,and designs a weight self-attention mechanism combined with SE blocks to enhance feature expression capabilities in cheap operations.Experimental results on the ImageNet dataset show that,compared to GhostNet,ResghostNet achieves higher accuracy while reducing the number of parameters by 52%.Although the computational complexity increases,by optimizing the usage strategy of GPU cachememory,themodel’s inference speed becomes faster.The ResghostNet is optimized in terms of classification accuracy and the number of model parameters,and shows great potential in edge computing devices.展开更多
Single Image Super-Resolution(SISR)seeks to reconstruct high-resolution(HR)images from lowresolution(LR)inputs,thereby enhancing visual fidelity and the perception of fine details.While Transformer-based models—such ...Single Image Super-Resolution(SISR)seeks to reconstruct high-resolution(HR)images from lowresolution(LR)inputs,thereby enhancing visual fidelity and the perception of fine details.While Transformer-based models—such as SwinIR,Restormer,and HAT—have recently achieved impressive results in super-resolution tasks by capturing global contextual information,these methods often suffer from substantial computational and memory overhead,which limits their deployment on resource-constrained edge devices.To address these challenges,we propose a novel lightweight super-resolution network,termed Binary Attention-Guided Information Distillation(BAID),which integrates frequency-aware modeling with a binary attention mechanism to significantly reduce computational complexity and parameter count whilemaintaining strong reconstruction performance.The network combines a high–low frequency decoupling strategy with a local–global attention sharing mechanism,enabling efficient compression of redundant computations through binary attention guidance.At the core of the architecture lies the Attention-Guided Distillation Block(AGDB),which retains the strengths of the information distillation framework while introducing a sparse binary attention module to enhance both inference efficiency and feature representation.Extensive×4 superresolution experiments on four standard benchmarks—Set5,Set14,BSD100,and Urban100—demonstrate that BAID achieves Peak Signal-to-Noise Ratio(PSNR)values of 32.13,28.51,27.47,and 26.15,respectively,with only 1.22 million parameters and 26.1 G Floating-Point Operations(FLOPs),outperforming other state-of-the-art lightweight methods such as Information Multi-Distillation Network(IMDN)and Residual Feature Distillation Network(RFDN).These results highlight the proposed model’s ability to deliver high-quality image reconstruction while offering strong deployment efficiency,making it well-suited for image restoration tasks in resource-limited environments.展开更多
Deep learning has been recognized as an effective method for indoor positioning.However,most existing real-valued neural networks(RVNNs)treat the two constituent components of complex-valued channel state information(...Deep learning has been recognized as an effective method for indoor positioning.However,most existing real-valued neural networks(RVNNs)treat the two constituent components of complex-valued channel state information(CSI)as real-valued inputs,potentially discarding useful information embedded in the original CSI.In addition,existing positioning models generally face the contradiction between computational complexity and positioning accuracy.To address these issues,we combine graph neural network(GNN)with complex-valued neural network(CVNN)to construct a lightweight indoor positioning model named CGNet.CGNet employs complexvalued convolution operation to directly process the original CSI data,fully exploiting the correlation between real and imaginary parts of CSI while extracting local features.Subsequently,the feature values are treated as nodes,and conditional position encoding(CPE)module is applied to add positional information.To reduce the number of connections in the graph structure and lower themodel complexity,feature information is mapped to an efficient graph structure through a dynamic axial graph construction(DAGC)method,with global features extracted usingmaximum relative graph convolution(MRConv).Experimental results show that,on the CTW dataset,CGNet achieves a 10%improvement in positioning accuracy compared to existing methods,while the number of model parameters is only 0.8 M.CGNet achieves excellent positioning accuracy with very few parameters.展开更多
The ubiquity of mobile devices has driven advancements in mobile object detection.However,challenges in multi-scale object detection in open,complex environments persist due to limited computational resources.Traditio...The ubiquity of mobile devices has driven advancements in mobile object detection.However,challenges in multi-scale object detection in open,complex environments persist due to limited computational resources.Traditional approaches like network compression,quantization,and lightweight design often sacrifice accuracy or feature representation robustness.This article introduces the Fast Multi-scale Channel Shuffling Network(FMCSNet),a novel lightweight detection model optimized for mobile devices.FMCSNet integrates a fully convolutional Multilayer Perceptron(MLP)module,offering global perception without significantly increasing parameters,effectively bridging the gap between CNNs and Vision Transformers.FMCSNet achieves a delicate balance between computation and accuracy mainly by two key modules:the ShiftMLP module,including a shift operation and an MLP module,and a Partial group Convolutional(PGConv)module,reducing computation while enhancing information exchange between channels.With a computational complexity of 1.4G FLOPs and 1.3M parameters,FMCSNet outperforms CNN-based and DWConv-based ShuffleNetv2 by 1%and 4.5%mAP on the Pascal VOC 2007 dataset,respectively.Additionally,FMCSNet achieves a mAP of 30.0(0.5:0.95 IoU threshold)with only 2.5G FLOPs and 2.0M parameters.It achieves 32 FPS on low-performance i5-series CPUs,meeting real-time detection requirements.The versatility of the PGConv module’s adaptability across scenarios further highlights FMCSNet as a promising solution for real-time mobile object detection.展开更多
Detecting small forest fire targets in unmanned aerial vehicle(UAV)images is difficult,as flames typically cover only a very limited portion of the visual scene.This study proposes Context-guided Compact Lightweight N...Detecting small forest fire targets in unmanned aerial vehicle(UAV)images is difficult,as flames typically cover only a very limited portion of the visual scene.This study proposes Context-guided Compact Lightweight Network(CCLNet),an end-to-end lightweight model designed to detect small forest fire targets while ensuring efficient inference on devices with constrained computational resources.CCLNet employs a three-stage network architecture.Its key components include three modules.C3F-Convolutional Gated Linear Unit(C3F-CGLU)performs selective local feature extraction while preserving fine-grained high-frequency flame details.Context-Guided Feature Fusion Module(CGFM)replaces plain concatenation with triplet-attention interactions to emphasize subtle flame patterns.Lightweight Shared Convolution with Separated Batch Normalization Detection(LSCSBD)reduces parameters through separated batch normalization while maintaining scale-specific statistics.We build TF-11K,an 11,139-image dataset combining 9139 self-collected UAV images from subtropical forests and 2000 re-annotated frames from the FLAME dataset.On TF-11K,CCLNet attains 85.8%mAP@0.5,45.5%mean Average Precision(mAP)@[0.5:0.95],87.4%precision,and 79.1%recall with 2.21 M parameters and 5.7 Giga Floating-point Operations Per Second(GFLOPs).The ablation study confirms that each module contributes to both accuracy and efficiency.Cross-dataset evaluation on DFS yields 77.5%mAP@0.5 and 42.3%mAP@[0.5:0.95],indicating good generalization to unseen scenes.These results suggest that CCLNet offers a practical balance between accuracy and speed for small-target forest fire monitoring with UAVs.展开更多
Background:Medical imaging advancements are constrained by fundamental trade-offs between acquisition speed,radiation dose,and image quality,forcing clinicians to work with noisy,incomplete data.Existing reconstructio...Background:Medical imaging advancements are constrained by fundamental trade-offs between acquisition speed,radiation dose,and image quality,forcing clinicians to work with noisy,incomplete data.Existing reconstruction methods either compromise on accuracy with iterative algorithms or suffer from limited generalizability with task-specific deep learning approaches.Methods:We present LDM-PIR,a lightweight physics-conditioned diffusion multi-model for medical image reconstruction that addresses key challenges in magnetic resonance imaging(MRI),CT,and low-photon imaging.Unlike traditional iterative methods,which are computationally expensive,or task-specific deep learning approaches lacking generalizability,integrates three innovations.A physics-conditioned diffusion framework that embeds acquisition operators(Fourier/Radon transforms)and noise models directly into the reconstruction process.A multi-model architecture that unifies denoising,inpainting,and super-resolution via shared weight conditioning.A lightweight design(2.1M parameters)enabling rapid inference(0.8s/image on GPU).Through self-supervised fine-tuning with measurement consistency losses adapts to new imaging modalities using fewer annotated samples.Results:Achieves state-of-the-art performance on fastMRI(peak signal-to-noise ratio(PSNR):34.04 for single-coil/31.50 for multi-coil)and Lung Image Database Consortium and Image Database Resource Initiative(28.83 PSNR under Poisson noise).Clinical evaluations demonstrate superior preservation of anatomical structures,with SSIM improvements of 8.8%for single-coil and 4.36%for multi-coil MRI over uDPIR.Conclusion:It offers a flexible,efficient,and scalable solution for medical image reconstruction,addressing the challenges of noise,undersampling,and modality generalization.The model’s lightweight design allows for rapid inference,while its self-supervised fine-tuning capability minimizes reliance on large annotated datasets,making it suitable for real-world clinical applications.展开更多
Bird's-eye-view(BEV)perception is a core technology for autonomous driving systems.However,existing solutions face the dilemma of high costs associated with multimodal methods and limited performance of vision-onl...Bird's-eye-view(BEV)perception is a core technology for autonomous driving systems.However,existing solutions face the dilemma of high costs associated with multimodal methods and limited performance of vision-only approaches.To address this issue,this paper proposes a framework named“a lightweight pure visual BEV perception method based on dual distillation of spatial-temporal knowledge”.This framework innovatively designs a lightweight vision-only student model based on Res Net,which leverages a dual distillation mechanism to learn from a powerful teacher model that integrates temporal information from both image and light detection and ranging(LiDAR)modalities.Specifically,we distill efficient multi-modal feature extraction and spatial fusion capabilities from the BEVFusion model,and distill advanced temporal information fusion and spatiotemporal attention mechanisms from the BEVFormer model.This dual distillation strategy enables the student model to achieve perception performance close to that of multi-modal models without relying on Li DAR.Experimental results on the nu Scenes dataset demonstrate that the proposed model significantly outperforms classical vision-only algorithms,achieves comparable performance to current state-of-the-art vision-only methods on the nu Scenes detection leaderboard in terms of both mean average precision(mAP)and the nu Scenes detection score(NDS)metrics,and exhibits notable advantages in inference computational efficiency.Although the proposed dual-teacher paradigm incurs higher offline training costs compared to single-model approaches,it yields a streamlined and highly efficient student model suitable for resource-constrained real-time deployment.This provides an effective pathway toward low-cost,high-performance autonomous driving perception systems.展开更多
In modern construction,Lightweight Aggregate Concrete(LWAC)has been recognized as a vital material of concern because of its unique properties,such as reduced density and improved thermal insulation.Despite the extens...In modern construction,Lightweight Aggregate Concrete(LWAC)has been recognized as a vital material of concern because of its unique properties,such as reduced density and improved thermal insulation.Despite the extensive knowledge regarding its macroscopic properties,there is a wide knowledge gap in understanding the influence of microscale parameters like aggregate porosity and volume ratio on the mechanical response of LWAC.This study aims to bridge this knowledge gap,spurred by the need to enhance the predictability and applicability of LWAC in various construction environments.With the help of advanced numerical methods,including the finite element method and a random circular aggregate model,this study critically evaluates the role played by these microscale factors.We found that an increase in the aggregate porosity from 23.5%to 48.5%leads to a drastic change of weakness from the bonding interface to the aggregate,reducing compressive strength by up to 24.2%and tensile strength by 27.8%.Similarly,the increase in the volume ratio of lightweight aggregate from 25%to 40%leads to a reduction in compressive strength by 13.0%and tensile strength by 9.23%.These results highlight the imperative role of microscale properties on the mechanical properties of LWAC.By supplying precise quantitative details on the effect of porosity and aggregate volume ratio,this research makes significant contributions to construction materials science by providing useful recommendations for the creation and optimization of LWAC with improved performance and sustainability in construction.展开更多
The application of deep learning in fabric defect detection has become increasingly widespread.To address false positives and false negatives in fabric roll seam detection,and to improve automation efficiency and prod...The application of deep learning in fabric defect detection has become increasingly widespread.To address false positives and false negatives in fabric roll seam detection,and to improve automation efficiency and product quality,we propose the Multi-scale Context DeepLabV3+(MSC-DeepLabV3+),a semantic segmentation network designed for fabric roll seam detection,based on DeepLabV3+.The model improvements include enhancing the backbone performance through optimization of the UIB-MobileNetV2 network;designing the Dynamic Atrous and Sliding-window Fusion(DASF)module to improve adaptability to multi-scale seam structures with dynamic dilation rates and a sliding-window mechanism;and utilizing the Progressive Low-level Feature Fusion(PLFF)module to progressively restore seam boundary details via shallow feature fusion.Additionally,an enhanced 3-SE attention mechanism is employed,replacing the direct concatenation operation.Experimental results show thatMSCDeepLabV3+outperforms classical and recent segmentation models.Compared to DeepLabV3+with an Xception backbone,MSC-DeepLabV3+achieves a mean intersection over union(mIoU)of 92.30%and the boundary Fscore(BF)of 92.54%,representing improvements of 3.04%and 3.14%,respectively.Moreover,the model complexity is significantly reduced,with the model parameters(params)decreasing to 3.44M and Frames Per Second(FPS)increasing from 101 to 273,demonstrating its potential for deployment in resource-constrained industrial scenarios.展开更多
基金the financial supports from National Key R&D Program for Young Scientists of China(Grant No.2022YFC3080900)National Natural Science Foundation of China(Grant No.52374181)+1 种基金BIT Research and Innovation Promoting Project(Grant No.2024YCXZ017)supported by Science and Technology Innovation Program of Beijing institute of technology under Grant No.2022CX01025。
文摘In this study,an inverse design framework was established to find lightweight honeycomb structures(HCSs)with high impact resistance.The hybrid HCS,composed of re-entrant(RE)and elliptical annular re-entrant(EARE)honeycomb cells,was created by constructing arrangement matrices to achieve structural lightweight.The machine learning(ML)framework consisted of a neural network(NN)forward regression model for predicting impact resistance and a multi-objective optimization algorithm for generating high-performance designs.The surrogate of the local design space was initially realized by establishing the NN in the small sample dataset,and the active learning strategy was used to continuously extended the local optimal design until the model converged in the global space.The results indicated that the active learning strategy significantly improved the inference capability of the NN model in unknown design domains.By guiding the iteration direction of the optimization algorithm,lightweight designs with high impact resistance were identified.The energy absorption capacity of the optimal design reached 94.98%of the EARE honeycomb,while the initial peak stress and mass decreased by 28.85%and 19.91%,respectively.Furthermore,Shapley Additive Explanations(SHAP)for global explanation of the NN indicated a strong correlation between the arrangement mode of HCS and its impact resistance.By reducing the stiffness of the cells at the top boundary of the structure,the initial impact damage sustained by the structure can be significantly improved.Overall,this study proposed a general lightweight design method for array structures under impact loads,which is beneficial for the widespread application of honeycomb-based protective structures.
基金support from the National Key Research and Development Program of China(Grant No.2021YFF0704102)the Chongqing Education Commission Key Project of Science and Technology Research(Grant No.KJZD-K202400610)the Chongqing Natural Science Foundation General Project(Grant No.CSTB2025NSCQ-GPX1263).
文摘Attribute-Based Encryption(ABE)has emerged as a fundamental access control mechanism in data sharing,enabling data owners to define flexible access policies.A critical aspect of ABE is key revocation,which plays a pivotal role in maintaining security.However,existing key revocation mechanisms face two major challenges:(1)High overhead due to ciphertext and key updates,primarily stemming from the reliance on revocation lists during attribute revocation,which increases computation and communication costs.(2)Limited universality,as many attribute revocation mechanisms are tailored to specific ABE constructions,restricting their broader applicability.To address these challenges,we propose LUAR(Lightweight and Universal Attribute Revocation),a novel revocation mechanism that leverages Intel Software Guard Extensions(SGX)while minimizing its inherent limitations.Given SGX’s constrained memory(≈90 MB in a personal computer)and susceptibility to side-channel attacks,we carefully manage its usage to reduce reliance while mitigating potential collusion risks between cloud service providers and users.To evaluate LUAR’s lightweight and universality,we integrate it with the classic BSW07 scheme,which can be seamlessly replaced with other ABE constructions.Experimental results demonstrate that LUAR enables secure attribute revocation with low computation and communication overhead.The processing time within the SGX environment remains stable at approximately 55 ms,regardless of the complexity of access policies,ensuring no additional storage or computational burden on SGX.Compared to the Hardware-based Revocable Attribute-Based Encryption(HR-ABE)scheme(IEEE S&P 2024),LUAR incurs a slightly higher computational cost within SGX;however,the overall time from initiating a data request to obtaining plaintext is shorter.As access policies grow more complex,LUAR’s advantages become increasingly evident,showcasing its superior efficiency and broader applicability.
基金funded by the Key Project of the National Natural Science Foundation of China(Grant No.U21A2058)Research Project of Hubei Provincial Department of Science and Technology(2024CSA075)support from the Taizhou Fengcheng Talent Program(2024).
文摘Conventional lightweight refractory materials with low bulk density and more pores suffer from harsh corrosion and erosion in actual applications.A type of lightweight Al_(2)O_(3)-MgAl_(2)O_(4)aggregates with a core-shell structure was synthesized at 1750℃using a rolling granulation method.Microstructural evolution and properties of the spherical aggregates were systematically studied.Scanning electron microscope and X-ray computed tomography results confirmed that a continuous and dense MgAl_(2)O_(4)spinel shell structure with a thickness of 200-300μm was formed on the surface.The corrosion results indicated that the corrosion index of the core-shell aggregates exhibited a 60%enhancement when compared to Al_(2)O_(3)spherical.Moreover,Al_(2)O_(3)-MgAl_(2)O_(4)refractory materials,which are based on the lightweight Al_(2)O_(3)-MgAl_(2)O_(4)spherical aggregates,possessed a higher temperature modulus of rupture of 9.19 MPa,and the retention rate of residual flexural strength reached 70%after thermal shock testing.The above results showed an improvement of 129.75 and 44.28%compared with pure Al_(2)O_(3)aggregate samples,respectively.In addition,the MgAl_(2)O_(4)spinel shell could trap the Mn,Fe elements from infiltrated slag and transfer into(Mg,Fe,Mn)Al_(2)O_(4)spinel,infiltrated CaO reacts with Sample Al_(2)O_(3)matrix to form a calcium hexaluminate(CA6)isolation layer,and the above two reasons enhance the corrosion resistance of the material.The corrosion mechanism was elaborated in detail.
基金supported by the Fundamental Research Funds for the Central Universities(Grant No.FRF-BD-25-001)Development and Application of Ultra-High Strength Hot Stamping Steel Strip for Automobiles(Grant No.20232BCJ22030)Manufacturing and Application Innovation and Integration of High-Safety Automotive Steel(Grant No.24431002D).
文摘The strength-ductility trade-off in low-Mn lightweight steels is a significant challenge due to the low thermal stability of austenite and the presence ofδ-ferrite.Two types of low-Mn lightweight steels containing V and NbVMo microalloying elements were developed by warm rolling.Among these,NbVMo steel demonstrated superior properties,achieving a tensile strength of~1.2 GPa and a product of strength and elongation exceeding 45 GPa%.In-depth mechanism analysis by atom probe tomography and quasi-in-situ electron backscatter diffraction revealed that different microalloying compositions influence the mechanical properties by strengtheningδ-ferrite,refining retained austenite and homogenizing matrix strain.In NbVMo steel,δ-ferrite strengthening is attributed to the synergistic effects of(V,Mo,Cr,Nb)C composite precipitation,fine NbC and MoC precipitates,and the solid solution strengthening of Mo.These mechanisms collectively contribute to a higher yield strength andδ-ferrite microhardness compared to V steel.Consequently,δ-ferrite and the surrounding matrix in NbVMo steel exhibit coordinated elongation during deformation,enhancing the ductility.The improved microstructural and strain uniformity in NbVMo steel mitigates stress concentration effects onδ-ferrite deformation and serves as a barrier that delays the transformation of retained austenite.In contrast,the retained austenite in V steel exhibits a blocky morphology with larger grain sizes,resulting in lower stability.Combined with localized stress concentrations due to non-uniform strain distribution,this leads to premature transformation of retained austenite to alleviate stress,ultimately impairing elongation and the continuity of strain hardening.Furthermore,the precipitation mechanisms of(V,Mo,Cr,Nb)C composite precipitates are elucidated.
基金supported by the National Natural Science Foundation of China 62402171.
文摘As a fundamental component in computer vision,edges can be categorized into four types based on discontinuities in reflectance,illumination,surface normal,or depth.While deep CNNs have significantly advanced generic edge detection,real-time multi-class semantic edge detection under resource constraints remains challenging.To address this,we propose a lightweight framework based on PiDiNet that enables fine-grained semantic edge detection.Our model simultaneously predicts background and four edge categories from full-resolution inputs,balancing accuracy and efficiency.Key contributions include:a multi-channel output structure expanding binary edge prediction to five classes,supported by a deep supervision mechanism;a dynamic class-balancing strategy combining adaptive weighting with physical priors to handle extreme class imbalance;and maintained architectural efficiency enabling real-time inference.Extensive evaluations on BSDS-RIND show our approach achieves accuracy competitive with state-of-the-art methods while operating in real time.
基金supported by the following projects:National Natural Science Foundation of China(62461041)Natural Science Foundation of Jiangxi Province China(20242BAB25068).
文摘With the large-scale deployment of the Internet ofThings(IoT)devices,their weak securitymechanisms make them prime targets for malware attacks.Attackers often use Domain Generation Algorithm(DGA)to generate random domain names,hiding the real IP of Command and Control(C&C)servers to build botnets.Due to the randomness and dynamics of DGA,traditional methods struggle to detect them accurately,increasing the difficulty of network defense.This paper proposes a lightweight DGA detection model based on knowledge distillation for resource-constrained IoT environments.Specifically,a teacher model combining CharacterBERT,a bidirectional long short-term memory(BiLSTM)network,and attention mechanism(ATT)is constructed:it extracts character-level semantic features viaCharacterBERT,captures sequence dependencieswith the BiLSTM,and integrates theATT for key feature weighting,formingmulti-granularity feature fusion.An improved knowledge distillation approach transfers the teacher model’s learned knowledge to the simplified DistilBERT student model.Experimental results show the teacher model achieves 98.68%detection accuracy.The student modelmaintains slightly improved accuracy while significantly compressing parameters to approximately 38.4%of the teacher model’s scale,greatly reducing computational overhead for IoT deployment.
文摘The Internet of Things(IoT)ecosystem is inherently heterogeneous,comprising diverse devices that must interoperate seamlessly to enable federated message and data exchange.However,as the number of service requests grows,existing approaches suffer from increased discovery time and degraded Quality of Service(QoS).Moreover,the massive data generated by heterogeneous IoT devices often contain redundancy and noise,posing challenges to efficient data management.To address these issues,this paper proposes a lightweight ontology-based architecture that enhances service discovery and QoS-aware semantic data management.The architecture employs Modified-Ordered Points to Identify theClustering Structure(M-OPTICS)to cluster and eliminate redundant IoT data.The clustered data are then modelled into a lightweight ontology,enabling semantic relationship inference and rule generation through an embedded inference engine.User requests,transmitted via theConstrainedApplication Protocol(CoAP),are semantically enriched and matched to QoS parameters using Dynamic Shannon Entropy optimized with the Salp Swarm Algorithm.Semantic matching is further refined using a bidirectional recurrent neural network(Bi-RNN),while a State–Action–Reward–State–Action(SARSA)reinforcement learning model dynamically defines and updates semantic rules to retrieve themost recent and relevant data across heterogeneous devices.Experimental results demonstrate that the proposed architecture outperforms existing methods in terms of response time,service delay,execution time,precision,recall,and F-score under varying CoAP request loads and communication overheads.The results confirm the effectiveness of the proposed lightweight ontology architecture for service discovery and data management in heterogeneous IoT environments.
基金the Shandong Provincial Key Research and Development Program(Grant No.2024SFGC0201).
文摘To address critical challenges in nighttime ship detection—high small-target missed detection(over 20%),insufficient lightweighting,and limited generalization due to scarce,low-quality datasets—this study proposes a systematic solution.First,a high-quality Night-Ships dataset is constructed via CycleGAN-based day-night transfer,combined with a dual-threshold cleaning strategy(Laplacian variance sharpness filtering and brightness-color deviation screening).Second,a Cross-stage Lightweight Fusion-You Only Look Once version 8(CLF-YOLOv8)is proposed with key improvements:the Neck network is reconstructed by replacing Cross Stage Partial(CSP)structure with the Cross Stage Partial Multi-Scale Convolutional Block(CSP-MSCB)and integrating Bidirectional Feature Pyramid Network(BiFPN)for weighted multi-scale fusion to enhance small-target detection;a Lightweight Shared Convolutional and Separated Batch Normalization Detection-Head(LSCSBD-Head)with shared convolutions and layer-wise Batch Normalization(BN)reduces parameters to 1.8M(42% fewer than YOLOv8n);and the FocalMinimum Point Distance Intersection over Union(Focal-MPDIoU)loss combines Minimum Point Distance Intersection over Union(MPDIoU)geometric constraints and Focal weighting to optimize low-overlap targets.Experiments show CLFYOLOv8 achieves 97.6%mAP@0.5(0.7% higher than YOLOv8n)with 1.8 M parameters,outperforming mainstream models in small-target detection,overlapping target discrimination,and adaptability to complex lighting.
基金financially supported by the Fujian Provincial Department of Science and Technology,the Collaborative Innovation Platform Project for Key Technologies of Smart Warehousing and Logistics Systems in the Fuzhou-Xiamen-Quanzhou National Independent Innovation Demonstration Zone(No.2025E3024).
文摘Fabric defect detection plays a vital role in ensuring textile quality.However,traditional manual inspection methods are often inefficient and inaccurate.To overcome these limitations,we propose FD-YOLO,an enhanced lightweight detection model based on the YOLOv11n framework.The proposed model introduces the Bi-level Routing Attention(BRAttention)mechanism to enhance defect feature extraction,enabling more detailed feature representation.It proposes Deep Progressive Cross-Scale Fusion Neck(DPCSFNeck)to better capture smallscale defects and incorporates a Multi-Scale Dilated Residual(MSDR)module to strengthen multi-scale feature representation.Furthermore,a Shared Detail-Enhanced Lightweight Head(SDELHead)is employed to reduce the risk of gradient explosion during training.Experimental results demonstrate that FD-YOLO achieves superior detection accuracy and Lightweight performance compared to the baseline YOLOv11n.
文摘The demand for extended electric vehicle(EV)range necessitates advanced lightweighting strategies.This study introduces a materials genome approach,augmented by machine learning(ML),for optimizing lightweight composite designs for EVs.A comprehensive materials genome database was developed,encompassing composites based on carbon,glass,and natural fibers.This database systematically records critical parameters such as mechanical properties,density,cost,and environmental impact.Machine learning models,including Random Forest,Support Vector Machines,and Artificial Neural Networks,were employed to construct a predictive system for material performance.Subsequent material composition optimization was performed using amulti-objective genetic algorithm.Experimental validation demonstrated that an optimized carbon fiber/bio-based resin composite achieved a 45%weight reduction compared to conventional steel,while maintaining equivalent structural strength.The predictive accuracy of the models reached 94.2%.A cost-benefit analysis indicated that despite a 15%increase in material cost,the overall vehicle energy consumption decreased by 12%,leading to an 18%total cost saving over a five-year operational lifecycle,under a representative mid-size battery electric vehicle(BEV)operational scenario.
基金Tianmin Tianyuan Boutique Vegetable Industry Technology Service Station(Grant No.2024120011003081)Development of Environmental Monitoring and Traceability System for Wuqing Agricultural Production Areas(Grant No.2024120011001866)。
文摘Tomato is a major economic crop worldwide,and diseases on tomato leaves can significantly reduce both yield and quality.Traditional manual inspection is inefficient and highly subjective,making it difficult to meet the requirements of early disease identification in complex natural environments.To address this issue,this study proposes an improved YOLO11-based model,YOLO-SPDNet(Scale Sequence Fusion,Position-Channel Attention,and Dual Enhancement Network).The model integrates the SEAM(Self-Ensembling Attention Mechanism)semantic enhancement module,the MLCA(Mixed Local Channel Attention)lightweight attention mechanism,and the SPA(Scale-Position-Detail Awareness)module composed of SSFF(Scale Sequence Feature Fusion),TFE(Triple Feature Encoding),and CPAM(Channel and Position Attention Mechanism).These enhancements strengthen fine-grained lesion detection while maintaining model lightweightness.Experimental results show that YOLO-SPDNet achieves an accuracy of 91.8%,a recall of 86.5%,and an mAP@0.5 of 90.6%on the test set,with a computational complexity of 12.5 GFLOPs.Furthermore,the model reaches a real-time inference speed of 987 FPS,making it suitable for deployment on mobile agricultural terminals and online monitoring systems.Comparative analysis and ablation studies further validate the reliability and practical applicability of the proposed model in complex natural scenes.
基金funded by Science and Technology Innovation Project grant No.ZZKY20222304.
文摘Aiming at the problem of potential information noise introduced during the generation of ghost feature maps in GhostNet,this paper proposes a novel lightweight neural network model called ResghostNet.This model constructs the Resghost Module by combining residual connections and Adaptive-SE Blocks,which enhances the quality of generated feature maps through direct propagation of original input information and selection of important channels before cheap operations.Specifically,ResghostNet introduces residual connections on the basis of the Ghost Module to optimize the information flow,and designs a weight self-attention mechanism combined with SE blocks to enhance feature expression capabilities in cheap operations.Experimental results on the ImageNet dataset show that,compared to GhostNet,ResghostNet achieves higher accuracy while reducing the number of parameters by 52%.Although the computational complexity increases,by optimizing the usage strategy of GPU cachememory,themodel’s inference speed becomes faster.The ResghostNet is optimized in terms of classification accuracy and the number of model parameters,and shows great potential in edge computing devices.
基金funded by Project of Sichuan Provincial Department of Science and Technology under 2025JDKP0150the Fundamental Research Funds for the Central Universities under 25CAFUC03093.
文摘Single Image Super-Resolution(SISR)seeks to reconstruct high-resolution(HR)images from lowresolution(LR)inputs,thereby enhancing visual fidelity and the perception of fine details.While Transformer-based models—such as SwinIR,Restormer,and HAT—have recently achieved impressive results in super-resolution tasks by capturing global contextual information,these methods often suffer from substantial computational and memory overhead,which limits their deployment on resource-constrained edge devices.To address these challenges,we propose a novel lightweight super-resolution network,termed Binary Attention-Guided Information Distillation(BAID),which integrates frequency-aware modeling with a binary attention mechanism to significantly reduce computational complexity and parameter count whilemaintaining strong reconstruction performance.The network combines a high–low frequency decoupling strategy with a local–global attention sharing mechanism,enabling efficient compression of redundant computations through binary attention guidance.At the core of the architecture lies the Attention-Guided Distillation Block(AGDB),which retains the strengths of the information distillation framework while introducing a sparse binary attention module to enhance both inference efficiency and feature representation.Extensive×4 superresolution experiments on four standard benchmarks—Set5,Set14,BSD100,and Urban100—demonstrate that BAID achieves Peak Signal-to-Noise Ratio(PSNR)values of 32.13,28.51,27.47,and 26.15,respectively,with only 1.22 million parameters and 26.1 G Floating-Point Operations(FLOPs),outperforming other state-of-the-art lightweight methods such as Information Multi-Distillation Network(IMDN)and Residual Feature Distillation Network(RFDN).These results highlight the proposed model’s ability to deliver high-quality image reconstruction while offering strong deployment efficiency,making it well-suited for image restoration tasks in resource-limited environments.
文摘Deep learning has been recognized as an effective method for indoor positioning.However,most existing real-valued neural networks(RVNNs)treat the two constituent components of complex-valued channel state information(CSI)as real-valued inputs,potentially discarding useful information embedded in the original CSI.In addition,existing positioning models generally face the contradiction between computational complexity and positioning accuracy.To address these issues,we combine graph neural network(GNN)with complex-valued neural network(CVNN)to construct a lightweight indoor positioning model named CGNet.CGNet employs complexvalued convolution operation to directly process the original CSI data,fully exploiting the correlation between real and imaginary parts of CSI while extracting local features.Subsequently,the feature values are treated as nodes,and conditional position encoding(CPE)module is applied to add positional information.To reduce the number of connections in the graph structure and lower themodel complexity,feature information is mapped to an efficient graph structure through a dynamic axial graph construction(DAGC)method,with global features extracted usingmaximum relative graph convolution(MRConv).Experimental results show that,on the CTW dataset,CGNet achieves a 10%improvement in positioning accuracy compared to existing methods,while the number of model parameters is only 0.8 M.CGNet achieves excellent positioning accuracy with very few parameters.
基金funded by the National Natural Science Foundation of China under Grant No.62371187the Open Program of Hunan Intelligent Rehabilitation Robot and Auxiliary Equipment Engineering Technology Research Center under Grant No.2024JS101.
文摘The ubiquity of mobile devices has driven advancements in mobile object detection.However,challenges in multi-scale object detection in open,complex environments persist due to limited computational resources.Traditional approaches like network compression,quantization,and lightweight design often sacrifice accuracy or feature representation robustness.This article introduces the Fast Multi-scale Channel Shuffling Network(FMCSNet),a novel lightweight detection model optimized for mobile devices.FMCSNet integrates a fully convolutional Multilayer Perceptron(MLP)module,offering global perception without significantly increasing parameters,effectively bridging the gap between CNNs and Vision Transformers.FMCSNet achieves a delicate balance between computation and accuracy mainly by two key modules:the ShiftMLP module,including a shift operation and an MLP module,and a Partial group Convolutional(PGConv)module,reducing computation while enhancing information exchange between channels.With a computational complexity of 1.4G FLOPs and 1.3M parameters,FMCSNet outperforms CNN-based and DWConv-based ShuffleNetv2 by 1%and 4.5%mAP on the Pascal VOC 2007 dataset,respectively.Additionally,FMCSNet achieves a mAP of 30.0(0.5:0.95 IoU threshold)with only 2.5G FLOPs and 2.0M parameters.It achieves 32 FPS on low-performance i5-series CPUs,meeting real-time detection requirements.The versatility of the PGConv module’s adaptability across scenarios further highlights FMCSNet as a promising solution for real-time mobile object detection.
基金funded by the Natural Science Foundation of Hunan Province(Grant No.2025JJ80352)the National Natural Science Foundation Project of China(Grant No.32271879).
文摘Detecting small forest fire targets in unmanned aerial vehicle(UAV)images is difficult,as flames typically cover only a very limited portion of the visual scene.This study proposes Context-guided Compact Lightweight Network(CCLNet),an end-to-end lightweight model designed to detect small forest fire targets while ensuring efficient inference on devices with constrained computational resources.CCLNet employs a three-stage network architecture.Its key components include three modules.C3F-Convolutional Gated Linear Unit(C3F-CGLU)performs selective local feature extraction while preserving fine-grained high-frequency flame details.Context-Guided Feature Fusion Module(CGFM)replaces plain concatenation with triplet-attention interactions to emphasize subtle flame patterns.Lightweight Shared Convolution with Separated Batch Normalization Detection(LSCSBD)reduces parameters through separated batch normalization while maintaining scale-specific statistics.We build TF-11K,an 11,139-image dataset combining 9139 self-collected UAV images from subtropical forests and 2000 re-annotated frames from the FLAME dataset.On TF-11K,CCLNet attains 85.8%mAP@0.5,45.5%mean Average Precision(mAP)@[0.5:0.95],87.4%precision,and 79.1%recall with 2.21 M parameters and 5.7 Giga Floating-point Operations Per Second(GFLOPs).The ablation study confirms that each module contributes to both accuracy and efficiency.Cross-dataset evaluation on DFS yields 77.5%mAP@0.5 and 42.3%mAP@[0.5:0.95],indicating good generalization to unseen scenes.These results suggest that CCLNet offers a practical balance between accuracy and speed for small-target forest fire monitoring with UAVs.
文摘Background:Medical imaging advancements are constrained by fundamental trade-offs between acquisition speed,radiation dose,and image quality,forcing clinicians to work with noisy,incomplete data.Existing reconstruction methods either compromise on accuracy with iterative algorithms or suffer from limited generalizability with task-specific deep learning approaches.Methods:We present LDM-PIR,a lightweight physics-conditioned diffusion multi-model for medical image reconstruction that addresses key challenges in magnetic resonance imaging(MRI),CT,and low-photon imaging.Unlike traditional iterative methods,which are computationally expensive,or task-specific deep learning approaches lacking generalizability,integrates three innovations.A physics-conditioned diffusion framework that embeds acquisition operators(Fourier/Radon transforms)and noise models directly into the reconstruction process.A multi-model architecture that unifies denoising,inpainting,and super-resolution via shared weight conditioning.A lightweight design(2.1M parameters)enabling rapid inference(0.8s/image on GPU).Through self-supervised fine-tuning with measurement consistency losses adapts to new imaging modalities using fewer annotated samples.Results:Achieves state-of-the-art performance on fastMRI(peak signal-to-noise ratio(PSNR):34.04 for single-coil/31.50 for multi-coil)and Lung Image Database Consortium and Image Database Resource Initiative(28.83 PSNR under Poisson noise).Clinical evaluations demonstrate superior preservation of anatomical structures,with SSIM improvements of 8.8%for single-coil and 4.36%for multi-coil MRI over uDPIR.Conclusion:It offers a flexible,efficient,and scalable solution for medical image reconstruction,addressing the challenges of noise,undersampling,and modality generalization.The model’s lightweight design allows for rapid inference,while its self-supervised fine-tuning capability minimizes reliance on large annotated datasets,making it suitable for real-world clinical applications.
基金supported by the National Natural Science Foundation of China(42476084,62203456,42276199)the Stable Support Project of National Key Laboratory(WDZC 20245250302)the National Key R&D Program of China(2024YFC2813502,2024YFC2813302)。
文摘Bird's-eye-view(BEV)perception is a core technology for autonomous driving systems.However,existing solutions face the dilemma of high costs associated with multimodal methods and limited performance of vision-only approaches.To address this issue,this paper proposes a framework named“a lightweight pure visual BEV perception method based on dual distillation of spatial-temporal knowledge”.This framework innovatively designs a lightweight vision-only student model based on Res Net,which leverages a dual distillation mechanism to learn from a powerful teacher model that integrates temporal information from both image and light detection and ranging(LiDAR)modalities.Specifically,we distill efficient multi-modal feature extraction and spatial fusion capabilities from the BEVFusion model,and distill advanced temporal information fusion and spatiotemporal attention mechanisms from the BEVFormer model.This dual distillation strategy enables the student model to achieve perception performance close to that of multi-modal models without relying on Li DAR.Experimental results on the nu Scenes dataset demonstrate that the proposed model significantly outperforms classical vision-only algorithms,achieves comparable performance to current state-of-the-art vision-only methods on the nu Scenes detection leaderboard in terms of both mean average precision(mAP)and the nu Scenes detection score(NDS)metrics,and exhibits notable advantages in inference computational efficiency.Although the proposed dual-teacher paradigm incurs higher offline training costs compared to single-model approaches,it yields a streamlined and highly efficient student model suitable for resource-constrained real-time deployment.This provides an effective pathway toward low-cost,high-performance autonomous driving perception systems.
基金supported by National Science Foundation of China(10972015,11172015)the Beijing Natural Science Foundation(8162008).
文摘In modern construction,Lightweight Aggregate Concrete(LWAC)has been recognized as a vital material of concern because of its unique properties,such as reduced density and improved thermal insulation.Despite the extensive knowledge regarding its macroscopic properties,there is a wide knowledge gap in understanding the influence of microscale parameters like aggregate porosity and volume ratio on the mechanical response of LWAC.This study aims to bridge this knowledge gap,spurred by the need to enhance the predictability and applicability of LWAC in various construction environments.With the help of advanced numerical methods,including the finite element method and a random circular aggregate model,this study critically evaluates the role played by these microscale factors.We found that an increase in the aggregate porosity from 23.5%to 48.5%leads to a drastic change of weakness from the bonding interface to the aggregate,reducing compressive strength by up to 24.2%and tensile strength by 27.8%.Similarly,the increase in the volume ratio of lightweight aggregate from 25%to 40%leads to a reduction in compressive strength by 13.0%and tensile strength by 9.23%.These results highlight the imperative role of microscale properties on the mechanical properties of LWAC.By supplying precise quantitative details on the effect of porosity and aggregate volume ratio,this research makes significant contributions to construction materials science by providing useful recommendations for the creation and optimization of LWAC with improved performance and sustainability in construction.
文摘The application of deep learning in fabric defect detection has become increasingly widespread.To address false positives and false negatives in fabric roll seam detection,and to improve automation efficiency and product quality,we propose the Multi-scale Context DeepLabV3+(MSC-DeepLabV3+),a semantic segmentation network designed for fabric roll seam detection,based on DeepLabV3+.The model improvements include enhancing the backbone performance through optimization of the UIB-MobileNetV2 network;designing the Dynamic Atrous and Sliding-window Fusion(DASF)module to improve adaptability to multi-scale seam structures with dynamic dilation rates and a sliding-window mechanism;and utilizing the Progressive Low-level Feature Fusion(PLFF)module to progressively restore seam boundary details via shallow feature fusion.Additionally,an enhanced 3-SE attention mechanism is employed,replacing the direct concatenation operation.Experimental results show thatMSCDeepLabV3+outperforms classical and recent segmentation models.Compared to DeepLabV3+with an Xception backbone,MSC-DeepLabV3+achieves a mean intersection over union(mIoU)of 92.30%and the boundary Fscore(BF)of 92.54%,representing improvements of 3.04%and 3.14%,respectively.Moreover,the model complexity is significantly reduced,with the model parameters(params)decreasing to 3.44M and Frames Per Second(FPS)increasing from 101 to 273,demonstrating its potential for deployment in resource-constrained industrial scenarios.
