Original Network Research Articles

Research on traffic sign detection algorithm based on improved SSD in complex environments

Abstract In complex traffic sign environments, detection challenges include uneven size distribution, insufficient valid information, and difficulties in identifying targets under resource constraints, leading to missed and false detections. This study proposes an enhanced lightweight traffic sign detection algorithm based on SSD (Single Shot MultiBox Detector). By replacing the original backbone network with MobileNetV2, the model is streamlined to have fewer parameters, which improves generalization in complex environments. This modification significantly boosts the recall rate, achieving a better balance between precision and recall. By introducing the FPN(Feature Pyramid Network) combined with the CBAM(Convolutional Block Attention Module) attention mechanism, the detailed and semantic information between deep and shallow layers is fully integrated, reducing the loss of feature information, thus enhancing the strengthening of key information of traffic signs and the adaptability to different scales of traffic signs. Finally, by integrating the cross-attention mechanism, the algorithm's anti-interference ability in complex environments is improved, and the positioning accuracy of traffic signs is enhanced by capturing the dependency between different positions. Through ablation experiments and comparative experiments on a public traffic sign dataset, our improved SSD algorithm achieved an mAP of 89.97%. Compared with the original algorithm, the mAP increased by 12.41%, the recall rate increased by 18.38%, and the sum of precision and recall F1 increased by 14.6%. These improvements significantly enhance the performance of traffic sign detection in complex environments, thereby meeting the performance requirements of traffic sign detection.

Continuous Dictionary of Nodes Model and Bilinear-Diffusion Representation Learning for Brain Disease Analysis.

Brain networks based on functional magnetic resonance imaging (fMRI) provide a crucial perspective for diagnosing brain diseases. Representation learning has recently attracted tremendous attention due to its strong representation capability, which can be naturally applied to brain disease analysis. However, traditional representation learning only considers direct and local node interactions in original brain networks, posing challenges in constructing higher-order brain networks to represent indirect and extensive node interactions. To address this problem, we propose the Continuous Dictionary of Nodes model and Bilinear-Diffusion (CDON-BD) network for brain disease analysis. The CDON model is innovatively used to learn the original brain network, with its encoder weights directly regarded as latent features. To fully integrate latent features, we further utilize Bilinear Pooling to construct higher-order brain networks. The Diffusion Module is designed to capture extensive node interactions in higher-order brain networks. Compared to state-of-the-art methods, CDON-BD demonstrates competitive classification performance on two real datasets. Moreover, the higher-order representations learned by our method reveal brain regions relevant to the diseases, contributing to a better understanding of the pathology of brain diseases.

A feature detection network based on self-attention mechanism for underwater image processing

Underwater image processing is seriously challenged by the absorption and scattering of light in water. This paper introduces a self supervised learning network that utilizes self attention mechanism for underwater visual applications. The goal is to improve the effectiveness and stability of underwater image feature detection. The combination of self attention mechanism enhances the sensitivity of the original network architecture to degraded features in blurred underwater images. A modified underwater image synthesis method is proposed for network training and evaluation to address the difficulties and high costs of obtaining underwater images. And a step-by-step training method was proposed, proving the effectiveness of this training method. The proposed network is trained using a step-by-step method and evaluated on various underwater image datasets. The experimental results show that the algorithm outperforms other methods in distribution, quantitative, and qualitative measurements, and exhibits enhanced feature detection ability in blurred images as water turbidity increases.

Feature Enhanced Spatial–Temporal Trajectory Similarity Computation

AbstractTrajectory similarity computation is a fundamental function in many applications of urban data analysis, such as trajectory clustering, trajectory compression, and route planning. In this paper, we study trajectory similarity computation on the road network. However, existing methods have been designed primarily for road network trajectories with spatial information, while ignoring the important temporal information in the real world. To solve this problem, we propose a Feature Enhanced Spatial–Temporal trajectory similarity computation framework FEST, which is a graph neural network (GNN) and sequence model pipeline. We first use the GNN model to capture global information on the road network. In particular, we enhance the process with multi-graph to learn multiple signals from the road network on different aspects. In addition to the original road network topology signal, we also take into account the content signal to learn spatial–temporal features from trajectory traffic, as well as the adaptive similarity signal of the road network to learn hidden features. From these three signals, we construct a multi-graph and use GCN to learn road intersection embedding jointly. Next, we propose a feature-enhanced Transformer with spatial–temporal information to learn correlation within the trajectory, and we further use mean-pooling to get the final trajectory embedding. We compare FEST with six trajectory similarity computation methods on two real-world datasets. The results show that FEST consistently outperforms all baselines and can improve the accuracy of the best-performing baseline.

Magistrae of the Beguines of Valenciennes and Their Social Networks

ABSTRACT This article uses sources from the thirteenth and fourteenth centuries to provide a systematic examination of the magistrae at the beguinage of Saint Elizabeth of Valenciennes, in which Marguerite Porete resided for a while, with a particular focus on the post-inquiry era under the governance of Béatrice de Chastel. This study shows that over the course of her more than forty-year (1322-64) career as mistress and grand mistress, Béatrice procured substantial material resources and extensive social support for the beguinage. Benefiting from her outstanding leadership, rare longevity, aristocratic origin, and vast social network, the beguinage of Saint Elizabeth did not decline after the investigation of 1323 but experienced a rapid recovery over the next half-century.

Prediction of Bonding Strength of Heat-Treated Wood Based on an Improved Harris Hawk Algorithm Optimized BP Neural Network Model (IHHO-BP)

In this study, we proposed an improved Harris Hawks Optimization (IHHO) algorithm based on the Sobol sequence, Whale Optimization Algorithm (WOA), and t-distribution perturbation. The improved IHHO algorithm was then used to optimize the BP neural network, resulting in the IHHO-BP model. This model was employed to predict the bonding strength of heat-treated wood under varying conditions of temperature, time, feed rate, cutting speed, and grit size. To validate the effectiveness and accuracy of the proposed model, it was compared with the original BP neural network model, WOA-BP, and HHO-BP benchmark models. The results showed that the IHHO-BP model reduced the Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and Mean Absolute Percentage Error (MAPE) by at least 51.16%, 40.38%, and 51.93%, respectively, while increasing the coefficient of determination (R2) by at least 10.85%. This indicates significant model optimization, enhanced generalization capability, and higher prediction accuracy, better meeting practical engineering needs. Predicting the bonding strength of heat-treated wood using this model can reduce production costs and consumption, thereby significantly improving production efficiency.

Lewis-base ligand-reshaped interfacial hydrogen-bond network boosts CO2 electrolysis.

Both the catalyst and electrolyte strongly impact the performance of CO2 electrolysis. Despite substantial progress in catalysts, it remains highly challenging to tailor electrolyte compositions and understand their functions at the catalyst interface. Here, we report that the ethylenediaminetetraacetic acid (EDTA) and its analogs, featuring strong Lewis acid-base interaction with metal cations, are selected as electrolyte additives to reshape the catalyst-electrolyte interface for promoting CO2 electrolysis. Mechanistic studies reveal that EDTA molecules are dynamically assembled toward interface regions in response to bias potential due to strong Lewis acid-base interaction of EDTA4--K+. As a result, the original hydrogen-bond network among interfacial H2O is disrupted, and a hydrogen-bond gap layer at the electrified interface is established. The EDTA-reshaped K+ solvation structure promotes the protonation of *CO2 to *COOH and suppressing *H2O dissociation to *H, thereby boosting the co-electrolysis of CO2 and H2O toward carbon-based products. In particular, when 5mM of EDTA is added into the electrolytes, the Faradaic efficiency of CO on the commercial Ag nanoparticle catalyst is increased from 57.0% to 90.0% at an industry-relevant current density of 500mA cm-2. More importantly, the Lewis-base ligand-reshaped interface allows a range of catalysts (Ag, Zn, Pd, Bi, Sn, and Cu) to deliver substantially increased selectivity of carbon-based products in both H-type and flow-type electrolysis cells.

Application of the bicharacteristic attention residual pyramid for the treatment of brain tumors

Currently, surgery remains the primary treatment for craniocerebral tumors. Before doctors perform surgeries, they need to determine the surgical plan according to the shape, location, and size of the tumor; however, various conditions of different patients make the tumor segmentation task challenging. To improve the accuracy of determining tumor shape and realizing edge segmentation, a U-shaped network combining a residual pyramid module and a dual feature attention module is proposed. The residual pyramid module can enlarge the receptive field, extract multiscale features, and fuse original information, which solves the problem caused by the feature pyramid pooling where the local information is not related to the remote information. In addition, the dual feature attention module is proposed to replace the skip connection in the original U-Net network, enrich the features, and improve the attention of the model to space and channel features with large amounts of information to be used for more accurate brain tumor segmentation. To evaluate the performance of the proposed model, experiments were conducted on the public datasets Kaggle_3M and BraTS2021. Because the model proposed in this study is applicable to two-dimensional image segmentation, it is necessary to obtain the crosscutting images of fair class in the BraTS2021 dataset in advance. Results show that the model accuracy, Jaccard similarity coefficient, Dice similarity coefficient, and false negative rate (FNR) on the Kaggle_3M dataset are 0.9395, 0.8812, 0.8958, and 0.007, respectively. The model accuracy, Jaccard similarity coefficient, Dice similarity coefficient, and FNR on the BraTS2021 dataset were 0.9375, 0.9072, 0.8981, and 0.0087, respectively. Compared with existing algorithms, all the indicators of the proposed algorithm have been improved, but the proposed model still has certain limitations and has not been applied to actual clinical trials. For specific datasets, the generalization ability of the model needs to be further improved. In the future work, the model will be further improved to address the aforementioned limitations.

Pineapple Detection with YOLOv7-Tiny Network Model Improved via Pruning and a Lightweight Backbone Sub-Network

High-complexity network models are challenging to execute on agricultural robots with limited computing capabilities in a large-scale pineapple planting environment in real time. Traditional module replacement often struggles to reduce model complexity while maintaining stable network accuracy effectively. This paper investigates a pineapple detection framework with a YOLOv7-tiny model improved via pruning and a lightweight backbone sub-network (the RGDP-YOLOv7-tiny model). The ReXNet network is designed to significantly reduce the number of parameters in the YOLOv7-tiny backbone network layer during the group-level pruning process. Meanwhile, to enhance the efficacy of the lightweight network, a GSConv network has been developed and integrated into the neck network, to further diminish the number of parameters. In addition, the detection network incorporates a decoupled head network aimed at separating the tasks of classification and localization, which can enhance the model’s convergence speed. The experimental results indicate that the network before pruning optimization achieved an improvement of 3.0% and 2.2%, in terms of mean average precision and F1 score, respectively. After pruning optimization, the RGDP-YOLOv7-tiny network was compressed to just 2.27 M in parameter count, 4.5 × 109 in computational complexity, and 5.0MB in model size, which were 37.8%, 34.1%, and 40.7% of the original YOLOv7-tiny network, respectively. Concurrently, the mean average precision and F1 score reached 87.9% and 87.4%, respectively, with increases of 0.8% and 1.3%. Ultimately, the model’s generalization performance was validated through heatmap visualization experiments. Overall, the proposed pineapple object detection framework can effectively enhance detection accuracy. In a large-scale fruit cultivation environment, especially under the constraints of hardware limitations and limited computational power in the real-time detection processes of agricultural robots, it facilitates the practical application of artificial intelligence algorithms in agricultural engineering.

YOLO-Granada: a lightweight attentioned Yolo for pomegranates fruit detection

Pomegranate is an important fruit crop that is usually managed manually through experience. Intelligent management systems for pomegranate orchards can improve yields and address labor shortages. Fast and accurate detection of pomegranates is one of the key technologies of this management system, crucial for yield and scientific management. Currently, most solutions use deep learning to achieve pomegranate detection, but deep learning is not effective in detecting small targets and large parameters, and the computation speed is slow; therefore, there is room for improving the pomegranate detection task. Based on the improved You Only Look Once version 5 (YOLOv5) algorithm, a lightweight pomegranate growth period detection algorithm YOLO-Granada is proposed. A lightweight ShuffleNetv2 network is used as the backbone to extract pomegranate features. Using grouped convolution reduces the computational effort of ordinary convolution, and using channel shuffle increases the interaction between different channels. In addition, the attention mechanism can help the neural network suppress less significant features in the channels or space, and the Convolutional Block Attention Module attention mechanism can improve the effect of attention and optimize the object detection accuracy by using the contribution factor of weights. The average accuracy of the improved network reaches 0.922. It is only less than 1% lower than the original YOLOv5s model (0.929) but brings a speed increase and a compression of the model size. and the detection speed is 17.3% faster than the original network. The parameters, floating-point operations, and model size of this network are compressed to 54.7%, 51.3%, and 56.3% of the original network, respectively. In addition, the algorithm detects 8.66 images per second, achieving real-time results. In this study, the Nihui convolutional neural network framework was further utilized to develop an Android-based application for real-time pomegranate detection. The method provides a more accurate and lightweight solution for intelligent management devices in pomegranate orchards, which can provide a reference for the design of neural networks in agricultural applications.

ICS-ResNet: A Lightweight Network for Maize Leaf Disease Classification

The accurate identification of corn leaf diseases is crucial for preventing disease spread and improving corn yield. Plant leaf images are often affected by factors such as complex backgrounds, climate, light, and sample data imbalance. To address these issues, we propose a lightweight convolutional neural network, ICS-ResNet, based on ResNet50. This network incorporates improved spatial and channel attention modules as well as a deep separable residual structure to enhance recognition accuracy. (1) The residual connections in the ResNet network prevent gradient loss during deep network training. (2) The improved channel attention (ICA) and spatial attention (ISA) modules fully utilize semantic information from different feature layers to accurately localize key features of the network. (3) To reduce the number of parameters and lower computational costs, we replace traditional convolutional computation with a depth-separable residual structure. (4) We also employ cosine annealing to dynamically adjust the learning rate, enhancing the network’s training stability, improving model convergence, and preventing local optima. Experiments on the corn dataset in Plant Village compare the proposed ICS-ResNet with eight popular networks: CSPNet, InceptionNet_v3, EfficientNet, ShuffleNet, MobileNet, ResNet50, ResNet101 and ResNet152. The results show that the ICS-ResNet achieves an accuracy of 98.87%, which is 5.03%, 3.18%, 1.13%, 1.81%, 1.13%, 0.68%, 0.44% and 0.60% higher than the other networks, respectively. Furthermore, the number of parameters and computations are reduced by 69.21% and 54.88%, respectively, compared to the original ResNet50 network, significantly improving the efficiency of corn leaf disease classification. The study provides strong technical support for sustainable agriculture and the promotion of agricultural science and technology innovation.

An Algorithm for Ship Detection in Complex Observation Scenarios Based on Mooring Buoys

Marine anchor buoys, as fixed-point profile observation platforms, are highly susceptible to the threat of ship collisions. Installing cameras on buoys can effectively monitor and collect evidence from ships. However, when using a camera to capture images, it is often affected by the continuous shaking of buoys and rainy and foggy weather, resulting in problems such as blurred images and rain and fog occlusion. To address these problems, this paper proposes an improved YOLOv8 algorithm. Firstly, the polarized self-attention (PSA) mechanism is introduced to preserve the high-resolution features of the original deep convolutional neural network and solve the problem of image spatial resolution degradation caused by shaking. Secondly, by introducing the multi-head self-attention (MHSA) mechanism in the neck network, the interference of rain and fog background is weakened, and the feature fusion ability of the network is improved. Finally, in the head network, this model combines additional small object detection heads to improve the accuracy of small object detection. Additionally, to enhance the algorithm’s adaptability to camera detection scenarios, this paper simulates scenarios, including shaking blur, rain, and foggy conditions. In the end, numerous comparative experiments on a self-made dataset show that the algorithm proposed in this study achieved 94.2% mAP50 and 73.2% mAP50:95 in various complex environments, which is superior to other advanced object detection algorithms.

Multi-Resolution Learning and Semantic Edge Enhancement for Super-Resolution Semantic Segmentation of Urban Scene Images.

Super-resolution semantic segmentation (SRSS) is a technique that aims to obtain high-resolution semantic segmentation results based on resolution-reduced input images. SRSS can significantly reduce computational cost and enable efficient, high-resolution semantic segmentation on mobile devices with limited resources. Some of the existing methods require modifications of the original semantic segmentation network structure or add additional and complicated processing modules, which limits the flexibility of actual deployment. Furthermore, the lack of detailed information in the low-resolution input image renders existing methods susceptible to misdetection at the semantic edges. To address the above problems, we propose a simple but effective framework called multi-resolution learning and semantic edge enhancement-based super-resolution semantic segmentation (MS-SRSS) which can be applied to any existing encoder-decoder based semantic segmentation network. Specifically, a multi-resolution learning mechanism (MRL) is proposed that enables the feature encoder of the semantic segmentation network to improve its feature extraction ability. Furthermore, we introduce a semantic edge enhancement loss (SEE) to alleviate the false detection at the semantic edges. We conduct extensive experiments on the three challenging benchmarks, Cityscapes, Pascal Context, and Pascal VOC 2012, to verify the effectiveness of our proposed MS-SRSS method. The experimental results show that, compared with the existing methods, our method can obtain the new state-of-the-art semantic segmentation performance.

Lightweight-detection: The strip steel surface defect identification based on improved YOLOv5d

The surface defect of strip steel seriously affects the product quality, the current identification methods have the inherent defects of low detection efficiency and poor detection accuracy, and it is important an improved an algorithm based on the modified YOLOv5 framework to enhance the defect detection effectiveness and accuracy. To create a more lightweight network model, the C3 module and part of the convolutional structure in the original YOLOv5 network were replaced with the GhostBottleneck structure. Additionally, the Coordinate Attention (CA) mechanism was incorporated into the Backbone section to compensate for the original model's lack of attention mechanisms. To address the angular mismatch between the actual frame and the predicted bounding box, a limitation overlooked by the CIOU loss function was further refined. Experimental results demonstrate that the enhanced YOLOv5s-CSG model outperformed the original YOLOv5s by 7.3 %, achieving a significant 37.9 % reduction in model size and an mAP value of 77.5 %. Furthermore, the detection precision and speed were notably higher compared to several widely-used algorithms. The proposed YOLOv5s-CSG model is well-suited for deployment on mobile devices and shows great potential for rapidly and accurately detecting both the type and location of strip surface defects, surpassing the performance of existing methods.

Fast identification of tomatoes in natural environments by improved YOLOv5s

Real time recognition and detection of tomato fruit maturity is a key function of tomato picking robots. Existing recognition and detection algorithms have slow speed and low recognition accuracy for small tomatoes. Here, a tomato fruit maturity detection model YOLOv5s3 based on improved YOLOv5s was proposed and its accuracy was verified through comparative experiments. On the basis of YOLOv5s, an SC module was proposed based on channel shuffle packet convolution. Then, A C3S module is constructed, which replaced the original C3 module with this C3S module to reduce the number of parameters while maintaining the feature expression ability of the original network. And a 3-feature fusion FF module was put forward, which accepted inputs from three feature layers. The FF module fused two feature maps from the backbone network. The C2 layer of the backbone was integrated, and the large target detection head was removed to use dual head detection to enhance the detection ability of small targets. The experimental results showed that the improved model has a detection accuracy of 94.8%, a recall rate of 96%, a parameter quantity of 3.02M, and an average accuracy (mAP0.5) of 93.3% for an intersection over union (IoU) of 0.5. The detection speed reaches 9.4ms. It can quickly and accurately identify the maturity of tomato fruits, and the detection speed is 22.95%, 33.33%, 48.91%, 68.35%, 15%, and 25.98% higher than the original YOLOv5s, YOLOv5m, YOLOv5l, YOLOv5x, YOLOv5n, and YOLOv4, respectively. The real-time testing visualization results of different models indicated that the improved model can effectively improve detection speed and solve the problem of low recognition rate for small tomatoes, which can provide reference for the development of picking robots.

Optimal path calculation method of optical network under complex constraints

AbstractTo address the optimal solution problem of loop‐free paths in software‐defined optical networks with multiple complex logical relationships, a unified constraint expression is utilized to describe the constraints. The logical relationships of complex constraints are dissected and simplified, differentiating between “AND” and “OR” constraints. An optimal path calculation method is proposed, involving the transformation of the network topology based on various constraints. This transformation includes layering the topology, removing specific links, and adding necessary links to portray the different complex constraints onto the original network structure. Following the topology transformation, an enhanced K‐shortest path algorithm is employed to compute the route satisfying the combination of multiple complex constraints, resulting in the global optimal solution. Experimental results demonstrate that this method can determine the optimal path under intricate constraints in a single computational iteration without requiring prior knowledge of the optimal constraint sequence. Therefore, it offers significant practical value compared to existing algorithms.

An improved 3D-UNet-based brain hippocampus segmentation model based on MR images

ObjectiveAccurate delineation of the hippocampal region via magnetic resonance imaging (MRI) is crucial for the prevention and early diagnosis of neurosystemic diseases. Determining how to accurately and quickly delineate the hippocampus from MRI results has become a serious issue. In this study, a pixel-level semantic segmentation method using 3D-UNet is proposed to realize the automatic segmentation of the brain hippocampus from MRI results. Methods: Two hundred three-dimensional T1-weighted (3D-T1) nongadolinium contrast-enhanced magnetic resonance (MR) images were acquired at Hangzhou Cancer Hospital from June 2020 to December 2022. These samples were divided into two groups, containing 175 and 25 samples. In the first group, 145 cases were used to train the hippocampus segmentation model, and the remaining 30 cases were used to fine-tune the hyperparameters of the model. Images for twenty-five patients in the second group were used as the test set to evaluate the performance of the model. The training set of images was processed via rotation, scaling, grey value augmentation and transformation with a smooth dense deformation field for both image data and ground truth labels. A filling technique was introduced into the segmentation network to establish the hippocampus segmentation model. In addition, the performance of models established with the original network, such as VNet, SegResNet, UNetR and 3D-UNet, was compared with that of models constructed by combining the filling technique with the original segmentation network. Results: The results showed that the performance of the segmentation model improved after the filling technique was introduced. Specifically, when the filling technique was introduced into VNet, SegResNet, 3D-UNet and UNetR, the segmentation performance of the models trained with an input image size of 48 × 48 × 48 improved. Among them, the 3D-UNet-based model with the filling technique achieved the best performance, with a Dice score (Dice score) of 0.7989 ± 0.0398 and a mean intersection over union (mIoU) of 0.6669 ± 0.0540, which were greater than those of the original 3D-UNet-based model. In addition, the oversegmentation ratio (OSR), average surface distance (ASD) and Hausdorff distance (HD) were 0.0666 ± 0.0351, 0.5733 ± 0.1018 and 5.1235 ± 1.4397, respectively, which were better than those of the other models. In addition, when the size of the input image was set to 48 × 48 × 48, 64 × 64 × 64 and 96 × 96 × 96, the model performance gradually improved, and the Dice scores of the proposed model reached 0.7989 ± 0.0398, 0.8371 ± 0.0254 and 0.8674 ± 0.0257, respectively. In addition, the mIoUs reached 0.6669 ± 0.0540, 0.7207 ± 0.0370 and 0.7668 ± 0.0392, respectively. Conclusion: The proposed hippocampus segmentation model constructed by introducing the filling technique into a segmentation network performed better than models built solely on the original network and can improve the efficiency of diagnostic analysis.

Combating temporal composition inference by high-order camouflaged network topology obfuscation

Topology inference driven by non-collaborative or incomplete prior knowledge is widely used in pivotal target network sieving and completion. However, perceivable topology also allows attackers to identify the fragile bottlenecks and perform efficacious attacks that are difficult to defend against by injecting indistinguishable low-volume attacks. Most existing countermeasures are proposed to obfuscate network data or set up honeypots with adversarial examples. However, there are two challenges when adding perturbations to live network links or nodes. Firstly, the perturbations imposed on the network cannot be conveniently projected to the original network with poor scalability. Secondly, applying significant changes to network information is laborious and impractical. In short, making a good trade-off between concealment and complexity is challenging. To address the above issues, we propose a fraudulent proactive defending tactic, namely HBB-TSP, to protect live network privacy by combating attacks of temporal network inference. Specifically, to penetrate the critical network structures, HBB-TSP first brings in the Statistical Validation of Hypergraph (SVH) method to identify the pivotal connection information of the network and extract the deep backbone structure. Then, the Temporal Simple Decomposition Weighting (TSDW) strategy is introduced, which can predict the backbone network with evolution rules and add highly obfuscated features at a minimized overhead. Finally, a discriminator with multiple centrality models is used to evaluate the deceptiveness and, in turn, affect the TSDW prediction. The entire process ensures the consistency and robustness of network changes while ensuring effective adversarial resistance. Experimental results on two scale real-world datasets demonstrate the effectiveness and generalization of adversarial perturbations. In particular, it is encouraging that our proposed defending scheme outperforms the advanced countermeasures. It ensures the realization of a deceptive obfuscated network at minimum overhead and is suitable for widespread deployment in scenarios of different scales.

A Lightweight Detection Algorithm for Surface Defects in Small-Sized Bearings

Background: To address issues in current deep learning models for detecting defects on industrial bearing surfaces, such as large parameter sizes and low precision in identifying small defects, we propose a lightweight detection algorithm for small-sized bearing appearance defects. Methods: First, we introduce a large separable convolution attention module on the spatial pyramid pooling fusion module. The deep convolutional layer with large convolutional kernels effectively captures more extensive context information of small-sized bearing defects while reducing the computation burden and learns attention weights to adaptively select the importance of input features. Secondly, we integrate the SimAM (simple attention mechanism) into the model without increasing the original network parameters, thereby augmenting the capacity to extract small-sized features and enhancing the model’s feature fusion capability. Finally, utilizing SIoU (Scylla IoU) as the regression loss and Soft-NMS (soft non-max suppression) for handling redundant boxes strengthens the model’s capacity to identify overlapping areas. Results: Experimental results demonstrate that our improved YOLOv8n model, sized at 6.5 MB, outperforms the baseline in terms of precision, recall, and mAP (mean average precision), with FPS (frames per second) of 146.7 (f/s), significantly enhancing bearing defect recognition for industrial applications.

Physics-guided deep learning based on modal sensitivity for structural damage identification with unseen damage patterns

Recently, wide attention has been paid to develop deep learning-based models for structural damage identification. However, in most of the studies, the pure deep learning-based structural damage identification methods lack physical interpretability and scientific consistency for generalization. Therefore, although such methods can achieve good damage identification results when data with similar damage patterns used for training and testing the network model, they usually give poor predictions in the unseen domain, i.e., the network’s extrapolation ability is limited. In this research, a physics-guided deep learning neural network (PGDLNN) is proposed by incorporating the physical loss constructed by structural modal parameters sensitivity analysis into the original Convolutional Neural Network (CNN) for structural damage identification. The model’s physical interpretability is improved, which can lead to more accurate damage identification results, especially in the domain with unseen damage patterns in training the network model. Numerical and experimental studies on simply supported beams are carried out to demonstrate the feasibility and effectiveness of the proposed method. The influences of measurement noise and incomplete measurements are also investigated. The results show that the incorporation of physical knowledge to deep learning model can enhance the generalization of the network, and hence improve the accuracy of damage severity quantification. This study not only performs damage localization and quantification simultaneously using the physics-guided deep learning neural network, but also demonstrates the superiority of incorporating physical knowledge into data-driven deep learning models for structural health monitoring.