Search

Complex systems in the real world often can be modeled as network structures, and community discovery algorithms for complex networks enable researchers to understand the internal structure and implicit information of networks. Existing community discovery algorithms are usually designed for single-layer networks or single-interaction relationships and do not consider the attribute information of nodes. However, many real-world networks consist of multiple types of nodes and edges, and there may be rich semantic information on nodes and edges. The methods for single-layer networks cannot effectively tackle multi-layer information, multi-relationship information, and attribute information. This paper proposes a community discovery algorithm based on multi-relationship embedding. The proposed algorithm first models the nodes in the network to obtain the embedding matrix for each node relationship type and generates the node embedding matrix for each specific relationship type in the network by node encoder. The node embedding matrix is provided as input for aggregating the node embedding matrix of each specific relationship type using a Graph Convolutional Network (GCN) to obtain the final node embedding matrix. This strategy allows capturing of rich structural and attributes information in multi-relational networks. Experiments were conducted on different datasets with baselines, and the results show that the proposed algorithm obtains significant performance improvement in community discovery, node clustering, and similarity search tasks, and compared to the baseline with the best performance, the proposed algorithm achieves an average improvement of 3.1% on Macro-F1 and 4.7% on Micro-F1, which proves the effectiveness of the proposed algorithm. [ABSTRACT FROM AUTHOR]

This paper aims to propose a table tennis target detection (TD) method based on deep learning (DL) and multi-scale feature fusion (MFF) to improve the detection accuracy of the ball in table tennis competition, optimize the training process of athletes, and improve the technical level. In this paper, DL technology is used to improve the accuracy of table tennis TD through MFF guidance. Initially, based on the FAST Region-based Convolutional Neural Network (FAST R-CNN), the TD is carried out in the table tennis match. Then, through the method of MFF guidance, different levels of feature information are fused, which improves the accuracy of TD. Through the experimental verification on the test set, it is found that the mean Average Precision (mAP) value of the target detection algorithm (TDA) proposed here reaches 87.3%, which is obviously superior to other TDAs and has higher robustness. The DL TDA combined with the proposed MFF can be applied to various detection fields and can help the application of TD in real life. [ABSTRACT FROM AUTHOR]

In this paper, we propose a novel anomaly detection method for data centers based on a combination of graph structure and abnormal attention mechanism. The method leverages the sensor monitoring data from target power substations to construct multidimensional time series. These time series are subsequently transformed into graph structures, and corresponding adjacency matrices are obtained. By incorporating the adjacency matrices and additional weights associated with the graph structure, an aggregation matrix is derived. The aggregation matrix is then fed into a pre-trained graph convolutional neural network (GCN) to extract graph structure features. Moreover, both the multidimensional time series segments and the graph structure features are inputted into a pretrained anomaly detectionmodel, resulting in corresponding anomaly detection results that help identify abnormal data. The anomaly detection model consists of a multi-level encoder-decoder module, wherein each level includes a transformer encoder and decoder based on correlation differences. The attention module in the encoding layer adopts an abnormal attention module with a dual-branch structure. Experimental results demonstrate that our proposed method significantly improves the accuracy and stability of anomaly detection. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Computer Engineering & Applications is the property of Beijing Journal of Computer Engineering & Applications Journal Co Ltd. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

This paper proposes a weakly-supervised structural surface crack detection algorithm that can detect the crack area in an image with low data labeling cost. The algorithm consists of a convolutional neural networks Vgg16-Crack for classification, an improved and optimized class activation map (CAM) algorithm for accurately reflecting the position and distribution of cracks in the image, and a method that combines superpixel segmentation algorithm simple linear iterative clustering (SLIC) with CAM for more accurate semantic segmentation of cracks. In addition, this paper uses Bayesian optimization algorithm to obtain the optimal parameter combination that maximizes the performance of the model. The test results show that the algorithm only requires image-level labeling, which can effectively reduce the labor and material consumption brought by pixel-level labeling while ensuring accuracy. [ABSTRACT FROM AUTHOR]

With the rapid advancement of wearable devices, Human Activities Recognition (HAR) based on these devices has emerged as a prominent research field. The objective of this study is to enhance the recognition performance of HAR by proposing an LSTM-1DCNN recognition algorithm that utilizes a single triaxial accelerometer. This algorithm comprises two branches: one branch consists of a Long and Short-Term Memory Network (LSTM), while the other parallel branch incorporates a one-dimensional Convolutional Neural Network (1DCNN). The parallel architecture of LSTM-1DCNN initially extracts spatial and temporal features from the accelerometer data separately, which are then concatenated and fed into a fully connected neural network for information fusion. In the LSTM-1DCNN architecture, the 1DCNN branch primarily focuses on extracting spatial features during convolution operations, whereas the LSTM branch mainly captures temporal features. Nine sets of accelerometer data from five publicly availableHAR datasets are employed for training and evaluation purposes. The performance of the proposed LSTM-1DCNN model is compared with five other HAR algorithms including Decision Tree, Random Forest, Support Vector Machine, 1DCNN, and LSTM on these five public datasets. Experimental results demonstrate that the F1-score achieved by the proposed LSTM-1DCNN ranges from 90.36% to 99.68%, with a mean value of 96.22% and standard deviation of 0.03 across all evaluated metrics on these five public datasetsoutperforming other existingHAR algorithms significantly in terms of evaluationmetrics used in this study. Finally the proposed LSTM-1DCNN is validated in real-world applications by collecting acceleration data of seven human activities for training and testing purposes. Subsequently, the trained HAR algorithm is deployed on Android phones to evaluate its performance. Experimental results demonstrate that the proposed LSTM-1DCNNalgorithm achieves an impressive F1-score of 97.67% on our self-built dataset. In conclusion, the fusion of temporal and spatial information in the measured data contributes to the excellent HAR performance and robustness exhibited by the proposed 1DCNN-LSTM architecture. [ABSTRACT FROM AUTHOR]

Emotion recognition commonly relies on single-modal recognition methods, such as voice and video signals, which demonstrate a good practicability and universality in some scenarios. Nevertheless, as emotion-recognition application scenarios continue to expand and the data volume surges, single-modal emotion recognition proves insufficient to meet people's needs for accuracy and comprehensiveness when the amount of data reaches a certain scale. Thus, this paper proposes the application of multimodal thought to enhance emotion-recognition accuracy and conducts corresponding data preprocessing on the selected dataset. Appropriate models are constructed for both audio and video modalities: for the audio-modality emotion-recognition task, this paper adopts the "time-distributed CNNs + LSTMs" model construction scheme; for the video-modality emotion-recognition task, the "DeepID V3 + Xception architecture" model construction scheme is selected. Furthermore, each model construction scheme undergoes experimental verification and comparison with existing emotion-recognition algorithms. Finally, this paper attempts late fusion and proposes and implements a late-fusion method based on the idea of weight adaptation. The experimental results demonstrate the superiority of the multimodal fusion algorithm proposed in this paper. When compared to the single-modal emotion-recognition algorithm, the accuracy of recognition is increased by almost 4%, reaching 84.33%. [ABSTRACT FROM AUTHOR]

An autonomous place recognition system is essential for scenarios where GPS is useless, such as underground tunnels. However, it is difficult to use existing algorithms to fully utilize the small number of effective features in underground tunnel data, and recognition accuracy is difficult to guarantee. In order to solve this challenge, an efficient point cloud position recognition algorithm, named Dual-Attention Transformer Network (DAT-Net), is proposed in this paper. The algorithm firstly adopts the farthest point downsampling module to eliminate the invalid redundant points in the point cloud data and retain the basic shape of the point cloud, which reduces the size of the point cloud and, at the same time, reduces the influence of the invalid point cloud on the data analysis. After that, this paper proposes the dual-attention Transformer module to facilitate local information exchange by utilizing the multi-head self-attention mechanism. It extracts local descriptors and integrates highly discriminative global descriptors based on global context with the help of a feature fusion layer to obtain a more accurate and robust global feature representation. Experimental results show that the method proposed in this paper achieves an average F 1 score of 0.841 on the SubT-Tunnel dataset and outperforms many existing state-of-the-art algorithms in recognition accuracy and robustness tests. [ABSTRACT FROM AUTHOR]

On the transmission line, the invasion of foreign objects such as kites, plastic bags, and balloons and the damage to electronic components are common transmission line faults. Detecting these faults is of great significance for the safe operation of power systems. Therefore, a YOLOv5 target detection method based on a deep convolution neural network is proposed. In this paper, Mobilenetv2 is used to replace Cross Stage Partial (CSP)-Darknet53 as the backbone. The structure uses depth-wise separable convolution to reduce the amount of calculation and parameters; improve the detection rate. At the same time, to compensate for the detection accuracy, the Squeeze-and-Excitation Networks (SENet) attention model is fused into the algorithm framework and a new detection scale suitable for small targets is added to improve the significance of the fault target area in the image. Collect pictures of foreign matters such as kites, plastic bags, balloons, and insulator defects of transmission lines, and sort theminto a data set. The experimental results on datasets show that themean Accuracy Precision (mAP) and recall rate of the algorithm can reach 92.1% and 92.4%, respectively. At the same time, by comparison, the detection accuracy of the proposed algorithm is higher than that of other methods. [ABSTRACT FROM AUTHOR]

Deepfakes are digital audio, video, or images manipulated using machine learning algorithms. These manipulated media files can convincingly depict individuals doing or saying things they never actually did. Deepfakes pose significant risks to our lives, including national security, financial markets, and personal privacy. The ability to create convincing deep fakes can also harm individuals' reputations and can be used to spread disinformation and fake news. As such, there is a growing need for reliable and accurate methods to detect deep fakes and prevent their harmful effects. In this paper, a hybrid convolutional neural network (CNN) and recurrent neural network (RNN) with a particle swarm optimization (PSO) algorithm is utilized to demonstrate a deep learning strategy for detecting deepfake videos. High accuracy, sensitivity, specificity, and F1 score were attained by the proposed approach when tested on two publicly available datasets: Celeb-DF and the Deepfake Detection Challenge Dataset (DFDC). Specifically, the proposed method achieved an average accuracy of 97.26% on Celeb-DF and an average accuracy of 94.2% on DFDC. The results were compared to other state-of-the-art methods and showed that the proposed method outperformed many. The proposed method can effectively detect deepfake videos, which is essential for identifying and preventing the spread of manipulated content online. [ABSTRACT FROM AUTHOR]

The No Free Lunch Theorem tells us that no algorithm can beat other algorithms on all types of problems. The algorithm selection structure is proposed to select the most suitable algorithm from a set of algorithms for an unknown optimization problem. This paper introduces an innovative algorithm selection approach called the CNN-HT, which is a two-stage algorithm selection framework. In the first stage, a Convolutional Neural Network (CNN) is employed to classify problems. In the second stage, the Hypothesis Testing (HT) technique is used to suggest the best-performing algorithm based on the statistical analysis of the performance metric of algorithms that address various problem categories. The two-stage approach can adapt to different algorithm combinations without the need to retrain the entire model, and modifications can be made in the second stage only, which is an improvement of one-stage approaches. To provide a more general structure for the classification model, we adopt Exploratory Landscape Analysis (ELA) features of the problem as input and utilize feature selection techniques to reduce the redundant ones. In problem classification, the average accuracy of classifying problems using CNN is 96%, which demonstrates the advantages of CNN compared to Random Forest and Support Vector Machines. After feature selection, the accuracy increases to 98.8%, further improving the classification performance while reducing the computational cost. This demonstrates the effectiveness of the first stage of the CNN-HT method, which provides a basis for algorithm selection. In the experiments, CNN-HT shows the advantages of the second stage algorithm as well as good performance with better average rankings in different algorithm combinations compared to the individual algorithms and another algorithm combination approach. [ABSTRACT FROM AUTHOR]

Introduction: Navigation satellite systems can fail to work or work incorrectly in a number of conditions: signal shadowing, electromagnetic interference, atmospheric conditions, and technical problems. All of these factors can significantly affect the localization accuracy of autonomous driving systems. This emphasizes the need for other localization technologies, such as Lidar. Methods: The use of the Kalman filter in combination with Lidar can be very effective in various applications due to the synergy of their capabilities. The Kalman filter can improve the accuracy of lidar measurements by taking into account the noise and inaccuracies present in the measurements. Results: In this paper, we propose a parallel Kalman algorithm in three- dimensional space to speed up the computational speed of Lidar localization. At the same time, the initial localization accuracy of the latter is preserved. A distinctive feature of the proposed approach is that the Kalman localization algorithm itself is parallelized, rather than the process of building a map for navigation. The proposed algorithm allows us to obtain the result 3.8 times faster without compromising the localization accuracy, which was 3% for both cases, making it effective for real-time decision-making. Discussion: The reliability of this result is confirmed by a preliminary theoretical estimate of the acceleration rate based on Ambdahl's law. Accelerating the Kalman filter with CUDA for Lidar localization can be of significant practical value, especially in real-time and in conditions where large amounts of data from Lidar sensors need to be processed. [ABSTRACT FROM AUTHOR]

In this paper, we propose a one-dimensional convolutional neural network model based on the adaptive batch normalization (AdaBN) algorithm to improve the CNN model, which is difficult to extract features from multi-rotor unmanned aerial vehicle (UAV) rotor structural faults under variable conditions and has poor fault diagnosis performance. The method accomplishes fault diagnosis and classification by feature extraction from lower dimensional multi-rotor UAV data. The AdaBN algorithm adjusts the parameters of the BN layer in the model during the testing phase to improve the domain adaptive capability of the model in scenarios with variable operating conditions. Also, to improve the robustness of the model under noisy conditions, the first layer of convolutional kernel dropout operation is introduced to improve the noise immunity of the model. To reduce the complexity of manual tuning and to find the optimal combination of hyperparameters for the network model more effectively, the grey wolf optimizer algorithm is used to optimize the hyperparameters and further improve the model performance. Finally, the effectiveness of the proposed method is verified through comparison tests, and it shows good diagnostic effects in noise and variable conditions compared with several commonly used fault diagnosis methods. [ABSTRACT FROM AUTHOR]

Copyright of Automotive Engineer (1674-6546) is the property of Auto Engineering Editorial Office and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

This paper presents a methodology for predicting the remaining usability of rolling bearings. The method combines a fully adaptive ensemble empirical modal decomposition of noise (CEEMDAN), convolutional neural network (CNN), and attention bidirectional long short-term memory network (ABiLSTM). Firstly, a finite number of intrinsic mode functions (IMFs) are obtained from breaking down the initial vibration signals using CEEMDAN. The IMFs are further screened by combining the correlation criterion and the craggy criterion. Then, time-frequency domain features, which are extracted from the screened IMFs, are reconstructed into a feature set. The SPT is recognized through some features, like the root mean square (RMS), variance, and kurtosis. Secondly, the deterioration character of rolling bearings was extracted using CNN and used to train the ABiLSTM network. Based on the output of the ABiLSTM network, it forecasts how long rolling bearings will last during use. Finally, the XJTU-SY rolling bearing dataset validated the validity of the suggested rolling bearing remaining life prediction method. We compare our algorithm with other algorithms, such as GRU, LSTM, and CNN–BiLSTM, in which the accuracy of MAE, MSE, RMSE, MAPE, and R2_score is significantly improved. Thus, the results of the validation experiments demonstrate that our proposed algorithm has excellent prediction accuracy. [ABSTRACT FROM AUTHOR]

Recently, synthetic aperture radar (SAR) target detection algorithms based on Convolutional Neural Networks (CNN) have received increasing attention. However, the large amount of computation required burdens the real-time detection of SAR ship targets on resource-limited and power-constrained satellite-based platforms. In this paper, we propose a hardware-aware model speed-up method for single-stage SAR ship targets detection tasks, oriented towards the most widely used hardware for neural network computing—Graphic Processing Unit (GPU). We first analyze the process by which the task of detection is executed on GPUs and propose two strategies according to this process. Firstly, in order to speed up the execution of the model on a GPU, we propose SAR-aware model quantification to allow the original model to be stored and computed in a low-precision format. Next, to ensure the loss of accuracy is negligible after the acceleration and compression process, precision-aware scheduling is used to filter out layers that are not suitable for quantification and store and execute them in a high-precision mode. Trained on the dataset HRSID, the effectiveness of this model speed-up algorithm was demonstrated by compressing four different sizes of models (yolov5n, yolov5s, yolov5m, yolov5l). The experimental results show that the detection speeds of yolov5n, yolov5s, yolov5m, and yolov5l can reach 234.7785 fps, 212.8341 fps, 165.6523 fps, and 139.8758 fps on the NVIDIA AGX Xavier development board with negligible loss of accuracy, which is 1.230 times, 1.469 times, 1.955 times, and 2.448 times faster than the original before the use of this method, respectively. [ABSTRACT FROM AUTHOR]

Channel pruning has recently become a widely used model compression method. However, most existing channel pruning methods only prune to decrease the model size, such as the number of parameters or FLOPs, and hence the decrease in model size does not effectively lead to an improvement in inference performance. To address this problem, this paper proposes a latency-optimized channel pruning method for CNN inference acceleration on GPU platforms by latency stair-step discrimination, two-stage benefit assessment and latency-sharing channel pruning. Compared with recent state-of-the-art model compression methods, it can achieve significant improvements in inference performance with comparable compression rates and model accuracy. The contributions of this paper include the following: first, a three-point latency stair-step discrimination method is proposed for determining the candidate prunable coordinates with the best latency performance adapted to the current hardware. Then, a two-stage benefit assessment method based on interlayer dependencies is proposed for determining the optimal channel pruning rate of each layer in the network. Finally, a latency-sharing channel pruning framework is proposed to accelerate the model pruning adaptation process. The method proposed in this paper can significantly reduce the model inference latency on multiple types of GPU platforms. To verify the effectiveness, we use three general-purpose GPU platforms and two embedded GPU platforms to evaluate the algorithm performance. The experimental results show that for recent state-of-the-art CNNs, the proposed method can achieve a 22.0–6.6% latency reduction and a 1.3 –3.0 inference performance improvement as well as a 1.2–4.3 pruning adaptation speedup with high model accuracy. [ABSTRACT FROM AUTHOR]

Collecting and analyzing data from all devices to improve the efficiency of business processes is an important task of Industrial Internet of Things (IIoT). In the age of data explosion, extensive text data generated by the IIoT have given birth to a variety of text representation methods. The task of text representation is to convert the natural language to a form that computer can understand with retaining the original semantics. However, these methods are difficult to effectively extract the semantic features among words and distinguish polysemy in natural language. Combining the advantages of convolutional neural network (CNN) and variational autoencoder (VAE), this paper proposes an intelligent CNN-VAE text representation algorithm as an advanced learning method for social big data within next-generation IIoT, which help users identify the information collected by sensors and perform further processing. This method employs the convolution layer to capture the local features of the context and uses the variational technique to reconstruct feature space to make it conform to the normal distribution. In addition, the improved word2vec model based on topical word embedding (TWE) is utilized to add topical information to word vectors to distinguish polysemy. This paper takes the social big data as an example to illustrate the way of the proposed algorithm applied in the next-generation IIoT and utilizes Cnews dataset to verify the performance of proposed method with four evaluating metrics (i.e., recall, accuracy, precision, and F1-score). Experimental results indicate that the proposed method outperforms word2vec-avg and CNN-AE in K-nearest neighbor (KNN), random forest (RF), and support vector machine (SVM) classifiers and distinguishes polysemy effectively. [ABSTRACT FROM AUTHOR]

The key step in the intelligence of tongue diagnosis is the segmentation of the tongue image, and the accuracy of the segmented edges has a significant impact on the subsequent medical judgment. Deep learning can predict the class of pixel points to achieve pixel-level segmentation of images, so it can be used to handle tongue segmentation tasks. However, different models have different segmentation effects, and they did not learn the connection between space and channels, resulting in inaccurate tongue segmentation. This paper first discussed the choice of model and loss function and then compared the results of different options to find the better model. Associating the red feature of the tongue is very conducive to segmentation as a feature, this paper tested many methods to try to get the color features of the original image to be paid attention to. Finally, this paper proposed an improved Encoder-Decoder network model to solve the problem based on the results. Start with Resnet as the backbone network, then introduce the U-Net model, and then we fused the attention layer, obtained from the source image through convolution and CBAM attention mechanism, and the feature layer obtained from the last upsampling in U-Net. Experimental results show that: The new, improved algorithm results are 2-3 percentage points higher than the popular algorithm, making it more suitable for tongue segmentation tasks. [ABSTRACT FROM AUTHOR]

The recognition of abnormal behavior in surveillance video is the focus of current research, which has high research value and broad application possibilities. Its main applications are in the fields of intelligent surveillance, intelligent security, and smart cities, and it is of great significance to study the recognition of abnormal behaviors. Because of the complexity of human movement and the variability of the external environment, the recognition and detection of abnormal behaviors have some challenges. The recognition and detection of abnormal human behaviors in surveillance video still needs further research and development. This paper uses the multi-branch convolutional neural network to extract the spatial features of video frames for the first time, and as an encoder to pass the condensed features to the Gated Recurrent Unit (GRU), which extracts Temporal features from multiple video frames. And then the Gated Recurrent Unit output the result as the decoder. We did a series of comparative experiments on UCF-Crime dataset. And finally, we achieved an accuracy of 86.78% in the test set. The experimental results show that our multi-branch convolutional fusion neural network is better than previous surveillance video abnormal behavior recognition algorithms. At the same time, in order to verify the generalization performance and efficiency of the algorithm, we also conducted an experimental validation on the UCF-101 dataset in this paper, and the results show that the algorithm in this paper can also show a high accuracy rate on the UCF-101 dataset, and the speed of the algorithm is almost close to that of the C3D method with improved accuracy rate, making it possible to develop simple recognition applications based on the algorithm studied in this paper subsequently. [ABSTRACT FROM AUTHOR]

The real-time transmission of ship status data from vessels to shore is crucial for live status monitoring and guidance. Traditional reliance on expensive maritime satellite systems for this purpose is being reconsidered with the emergence of the global short message communication service offered by the BeiDou-3 navigation satellite system. While this system presents a more cost-effective solution, its bandwidth is notably insufficient for handling real-time ship status data. This inadequacy necessitates the compression of such data. Therefore, this paper introduces an algorithm tailored for real-time compression of sequential ship status data. The algorithm is engineered to ensure both accuracy and the preservation of valid data range integrity. Our methodology integrates quantization, predictive coding employing an attention-averaging-based predictor, and arithmetic coding. This combined approach facilitates the transmission of succinct messages through the BeiDou Navigation System, enabling the live monitoring of ocean-going vessels. Experimental trials conducted with authentic data obtained from ship monitoring systems validate the efficiency of our approach. The achieved compression rates closely approximate theoretical minimum values. Consequently, this method exhibits substantial promise for the real-time transmission of parameters across various systems. [ABSTRACT FROM AUTHOR]

[Objective] With the rapid popularization of new energy vehicles and the vigorous development of autonomous driving technology, 3D object detection algorithms play a pivotal role in real road scenes. The LiDAR point clouds contain precise position and geometric structure information of the object, which can accurately describe the 3D space position of the target. Moreover, LiDAR makes environmental perception and route planning of unmanned vehicles a reality. However, cars in real scenes often fall into complex and difficult situations, such as occlusion and truncation of objects, which contribute to highly sparse clouds and incomplete contours. Therefore, the effective use of disordered and unevenly distributed point clouds for accurate 3D object detection has important research significance and practical value for the safety of autonomous driving. [Methods] This paper uses LiDAR point clouds in an autonomous driving scene to conduct in-depth research on a high-performance 3D object detection algorithm based on deep learning. The 3D object detection algorithm based on multibranch feature fusion (PSANet) is designed to improve the capacity and viability of autonomous driving technology. After the disordered point clouds are divided into regular voxels, the voxel feature coding module and convolutional neural network are used to learn the voxel features, and the sparse 3D data are compressed into a dense 2D bird's-eye view. Furthermore, the multiscale bird's-eye view features are deeply fused through the coarse and fine branches of the 2D backbone network. The splitting and aggregation feature pyramid module in the fine branch splits and aggregates the bird's-eye view features at different levels and realizes the deep fusion of semantic information, texture information, and context information of multiscale features to obtain more expressive features. The multiscale features in the coarse branch are fused after transposed convolution, and the precise original spatial location information is preserved. After the feature extraction of coarse and fine branches, the element-wise addition method can obtain more accurate and complete features for object classification, position regression, and orientation prediction. [Results] The experimental results on the KITTI dataset show that the average precision of PSANet in 3D object detection and bird's-eye view object detection tasks reach 81.72% and 88.25%, respectively. The inference speed on a single GTX 1080Ti GPU can reach 24 frames per second, and it shows strong robustness in complex scenes. Compared with the two-stage target detection algorithms MV3D, AVOD-FPN, F-PointNet, and IPOD, the average accuracy of 3D target detection of this algorithm increased by 18.21%, 5.89%, 8.94%, and 3.12%, respectively. Compared with the one-stage target detection algorithms VoxelNet, SECOND, PointPillars, and VoTr-SSD, the average accuracy of 3D target detection of this algorithm increased by 11.63%, 4.05%, 3.19%, 0.98%, 1.16%, and 0.7%, respectively. The detection speed of this algorithm improved by 14 frames per second compared with the two-stage algorithm PointRCNN with similar accuracy. [Conclusions] In comparison with other advanced algorithms, this algorithm exhibits strong performance, in which it can better balance the accuracy and speed of target detection in autonomous driving scenarios. Higher accuracy is required in the algorithm, whether one- or two-stage, for novel applications. The method with the highest efficiency must be used in an autonomous car. [ABSTRACT FROM AUTHOR]

Hybrid approaches combining machine learning with traditional inverse problem solution methods represent a promising direction for the further development of inverse modeling algorithms. The paper proposes an approach to emission source identification from measurement data for advection–diffusion–reaction models. The approach combines general-type source identification and post-processing refinement: first, emission source identification by measurement data is carried out by a sensitivity operator-based algorithm, and then refinement is done by incorporating a priori information about unknown sources. A general-type distributed emission source identified at the first stage is transformed into a localized source consisting of multiple point-wise sources. The second, refinement stage consists of two steps: point-wise source localization and emission rate estimation. Emission source localization is carried out using deep learning with convolutional neural networks. Training samples are generated using a sensitivity operator obtained at the source identification stage. The algorithm was tested in regional remote sensing emission source identification scenarios for the Lake Baikal region and was able to refine the emission source reconstruction results. Hence, the aggregates used in traditional inverse problem solution algorithms can be successfully applied within machine learning frameworks to produce hybrid algorithms. [ABSTRACT FROM AUTHOR]

This paper presents an in-depth study and analysis of oil painting classification and simulation using an improved embedded learning fusion vision perception algorithm. This paper analyzes and models the image quality evaluation problem by simulating the human visual system and extracting quality perception features as the main entry point to improve the prediction accuracy of the overall algorithm. This paper proposes a multi-classification method of CCNN, which uses the similarity measure based on information first to achieve multi-classification of artwork styles and artists, and this part is the main part of this paper. This paper uses the wiki art repository to construct a dataset of oil paintings, including over 2000 works by 20 artists in 13 styles. CNN achieves an accuracy of 85.75% on the artist classification task, which is far more effective than traditional deep learning networks such as Resnet. Finally, we use the network model of this paper and other network models to train the classification of 3, 4, and 6 categories of art images. The accuracy of art image classification by this paper's algorithm is higher than that of the current mainstream convolutional neural network models, and the extracted features are more comprehensive and more accurate than traditional art image feature extraction methods, which do not rely on researchers to extract image features. Experiments show that the proposed method can achieve excellent prediction accuracy for both synthetic distorted images and distorted images. [ABSTRACT FROM AUTHOR]

The environmental perception of intelligent vehicles is an important part of achieving autonomous driving, and object detection is one of the crucial functions of the environmental perception systems. Accurate and rapid detection of targets nearby is essentail for the smart vehicles. The faster and more accurately the detection algorithm can detect surrounding targets, the better. However, the detection speed and accuracy of existing perception algorithms still need improving. The distance information between the surrounding objects and the vehicle itself during its operation should also be obtained. The widely used YOLO series of algorithms, as representatives of one-stage object detection methods, have achieved a satisfactory balance between speed and detection accuracy, but there is stiil great potential for further improvement. This paper proposes an end-to-end object detection algorithm based on YOLO algorithm improvement-dual YOLO, which detects and measures the major participants in road traffic ( cars, pedestrians, cyclists) . The attention mechanism acts in the same way as humans observe things, focusing on more important information. The basic idea of combining attention mechanism with convolutional neural networks is to allow the network to concentrate on effective feature information in the image and ignore invalid information, thus improving the overall detection precision. Its mechanism of action is achieved by changing the weights of neurons in the network. This paper is based on the improvement of YOLOV5 algorithm, adding a channel attention mechanism to the backbone network to increase the learning weight of the neural network for effective features, thereby enhancing detection accuracy. In the meantime, a distance measurement module for binocular cameras is introduced to obtain target distance information. The basic principle of binocular camera ranging is triangulation, which employs the visual difference between two cameras to determine the distance of an object. Firstly, the camera is calibrated and distortion is corrected, and then stereo matching using the Semi Global Block Matching (SGBM) algorithm is performed to obtain disparity. Finally, the distance information of the target is calculated. The test results show the dual YOLO object detection algorithm achieves a detection speed of 60 fps and an average accuracy of 85. 99% in identifying the major participants in road traffic, which are far more superior to other algorithms in detecting cyclists and pedestrians. The proposed algorithm well meets the real-time detection and distance measurement requirements of intelligent vehicles. [ABSTRACT FROM AUTHOR]

In some fire classification task samples, it is especially important to learn and select limited features. Therefore, enhancing shallow characteristic learning and accurately reserving deep characteristics play a decisive role in the final fire classification task. In this paper, we propose an integrated algorithm based on bidirectional characteristics and feature selection for fire image classification called BCFS-Net. This algorithm is integrated from two modules, a bidirectional characteristics module and feature selection module; hence, it is called an integrated algorithm. The main process of this algorithm is as follows: First, we construct a bidirectional convolution module to obtain multiple sets of bidirectional traditional convolutions and dilated convolutions for the feature mining and learning shallow features. Then, we improve the Inception V3 module. By utilizing the bidirectional attention mechanism and Euclidean distance, feature points with greater correlation between the feature maps generated by convolutions in the Inception V3 module are selected. Next, we comprehensively consider and integrate feature points with richer semantic information from multiple dimensions. Finally, we use convolution to further learn the deep features and complete the final fire classification task. We validated the feasibility of our proposed algorithm in three sets of public fire datasets, and the overall accuracy value in the BoWFire dataset reached 88.9%. The overall accuracy in the outdoor fire dataset reached 96.96%. The overall accuracy value in the Fire Smoke dataset reached 81.66%. [ABSTRACT FROM AUTHOR]

Recently, there had been a massive group of people, who were being rapidly affected by melanoma. Melanoma is a form of skin cancer that develops on the skin's surface layer. This is primarily caused due to excessive skin exposure to UV radiation and severe sunburns. Thus, the early detection of melanoma can aid us to cure it completely. This paper intends to introduce a new melanoma detection framework with four main phases viz. segmentation, feature extraction, optimal feature selection, as well as detection. Initially, the segmentation process takes place to the input skin image via Fuzzy C-Means Clustering (FCM) approach. From the segmented image ( Im seg ) , some of the features such as Gray Level Run Length Matrix (GLRM), Local Vector Pattern (LVP), Local Binary Pattern (LBP), Local Directional Pattern (LDP) and Local Tetra Pattern (LTrP) are extracted. As the extracted features (F) suffered from the issue of "curse of dimensionality", this paper utilizes optimization to select optimal features, which makes the detection more precise. As a novelty, a new hybrid algorithm Particle-Assisted Moth Search Algorithm (PA-MSA) is introduced that hybridizes the concept of Moth Search Algorithm (MSA) and Particle Swarm Optimization (PSO), respectively. For the classification process, the optimally chosen features (F opt ) are fed as input, where Deep Convolution Neural Network (DCNN) is used. Finally, a performance-based comparative analysis is conducted among the proposed PA-MSA as well as the existing models with respect to various measures. [ABSTRACT FROM AUTHOR]

Featured Application: The aim is to distinguish maturity, reduce labor costs, and solve the occurrence of false or missed detections caused by strawberry leaf occlusion. In view of the time-consuming and laborious manual picking and sorting of strawberries, the direct impact of image recognition accuracy on automatic picking and the rapid development of deep learning(DL), a Faster Regions with Convolutional Neural Network features (R-CNN) strawberry recognition method that combines Mixup data augmentation, a ResNet(Residual Network)50 backbone feature extraction network and a Soft-NMS (Non-Maximum Suppression) algorithm, named the MRS Faster R-CNN, is proposed. In this paper, the transfer learning backbone feature extraction network VGG (Visual Geometry Group) 16 and ResNet50 are compared, and the superior ResNet50 is selected as the backbone network of MRS Faster R-CNN. The data augmentation method of Mixup image fusion is used to improve the learning and generalization ability of the model. The redundant bboxes (bounding boxes) are removed through Soft-NMS to obtain the best region proposal. The freezing phase is added to the training process, effectively reducing the occupation of video memory and shortening the training time. After experimental verification, the optimized model improved the AP (Average Precision) values of mature and immature strawberries by 0.26% and 5.34%, respectively, and the P(Precision) values by 0.81% and 6.34%, respectively, compared to the original model (R Faster R-CNN). Therefore, the MRS Faster R-CNN model proposed in this paper has great potential in the field of strawberry recognition and maturity classification and improves the recognition rate of small fruit and overlapping occluded fruit, thus providing an excellent solution for mechanized picking and sorting. [ABSTRACT FROM AUTHOR]

To address the challenges of complex backgrounds, small defect sizes, and diverse defect types in defect detection of wire bonding X-ray images, this paper proposes a convolutional-neural-network-based defect detection method called YOLO-CSS. This method designs a novel feature extraction network that effectively captures semantic features from different gradient information. It utilizes a self-adaptive weighted multi-scale feature fusion module called SMA which adaptively weights the contribution of detection results based on different scales of feature maps. Simultaneously, skip connections are employed at the bottleneck of the network to ensure the integrity of feature information. Experimental results demonstrate that on the wire bonding X-ray defect image dataset, the proposed algorithm achieves mAP 0.5 and mAP 0.5–0.95 values of 97.3% and 72.1%, respectively, surpassing the YOLO series algorithms. It also exhibits certain advantages in terms of model size and detection speed, effectively balancing detection accuracy and speed. [ABSTRACT FROM AUTHOR]

Because of its long running time, complex working environment, and for other reasons, a gear is prone to failure, and early failure is difficult to detect by direct observation; therefore, fault diagnosis of gears is very necessary. Neural network algorithms have been widely used to realize gear fault diagnosis, but the structure of the neural network model is complicated, the training time is long and the model is not easy to converge. To solve the above problems and combine the advantages of the ResNeXt50 model in the extraction of image features, this paper proposes a gearbox fault detection method that integrates the convolutional block attention module (CBAM). Firstly, the CBAM is embedded in the ResNeXt50 network to enhance the extraction of image channels and spatial features. Secondly, the different time–frequency analysis method was compared and analyzed, and the method with the better effect was selected to convert the one-dimensional vibration signal in the open data set of the gearbox into a two-dimensional image, eliminating the influence of the redundant background noise, and took it as the input of the model for training. Finally, the accuracy and the average training time of the model were obtained by entering the test set into the model, and the results were compared with four other classical convolutional neural network models. The results show that the proposed method performs well both in fault identification accuracy and average training time under two working conditions, and it also provides some references for existing gear failure diagnosis research. [ABSTRACT FROM AUTHOR]

Copyright of Application Research of Computers / Jisuanji Yingyong Yanjiu is the property of Application Research of Computers Edition and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Summary: Convolutional neural networks (CNNs) have shown remarkable advantages in a wide range of domains at the expense of huge parameters and computations. Modern CNNs still tend to be more complex and larger to achieve better inference accuracy. However, the complex and large structures of CNNs could slow down the inference speed. Recently, Compressing the convolutional weights to be sparse by pruning the unimportant parameters has been demonstrated as an efficient way to reduce the computations of CNNs. On the other hand, field‐programmable gate arrays (FPGAs) have been a popular hardware platform to accelerate CNN inference. In this paper, we propose an algorithm/hardware co‐optimized method for accelerating CNN inference on FPGAs. For the algorithm, we take advantage of unstructured and structured parameter sparsifying methods to achieve high sparsity and keep the regularity of convolutional weights. Correspondingly, hardware‐friendly index representations of sparse convolutional weights are proposed. For the hardware architecture, we propose row‐wise input‐stationary dataflow, which is tightly coupled with the algorithm. A row‐wise computing engine (RConv Engine) is proposed, which is based on the dataflow. Inside the RConv Engine, the scalar‐vector structure is applied to implement the basic processing elements (PEs). To flexibly calculate the feature map with various sizes, the PEs are organized in a 2D structure with two work modes. The experimental results demonstrate that our co‐optimized method implements high sparsity of convolutional weights, and the computing engine achieves high computation efficiency. Compared with other accelerators, our co‐optimized method implements a 10.9×$$ \times $$ speedup on FPS at most with the highest sparsity of convolutional weights and negligible accuracy loss. [ABSTRACT FROM AUTHOR]

The accurate detection of traffic signs is a critical component of self-driving systems, enabling safe and efficient navigation. In the literature, various methods have been investigated for traffic sign detection, among which deep learning-based approaches have demonstrated superior performance compared to other techniques. This paper justifies the widespread adoption of deep learning due to its ability to provide highly accurate results. However, the current research challenge lies in addressing the need for high accuracy rates and real-time processing requirements. In this study, we propose a convolutional neural network based on the YOLOv8 algorithm to overcome the aforementioned research challenge. Our approach involves generating a custom dataset with diverse traffic sign images, followed by conducting training, validation, and testing sets to ensure the robustness and generalization of the model. Experimental results and performance evaluation demonstrate the effectiveness of the proposed method. Extensive experiments show that our model achieved remarkable accuracy rates in traffic sign detection, meeting the real-time requirements of the input data. [ABSTRACT FROM AUTHOR]

Hand-drawn is one of the few visual descriptors that can directly represent visual content, and has significant research in the area of computer vision. Aiming at the problem of sparse features in the realm of hand-drawn image retrieval, hand-drawn images, and the easy deformation of hand-drawn images, this paper proposes a feature extraction method of grid resource sharing collaborative algorithm, which can be obtained utilizing precisely extracted semantic characteristics from hand-drawn images through computer multimedia-aided design Efficient and accurate retrieval results. First, the fundamental framework for obtaining semantic features is algorithm; then the attention model mechanism is the grid resource sharing collaborative introduced in the process of supervised training, and the attention structure block is introduced after the convolutional neural network's bottom layer. To locate effective semantic features, In order to accomplish high-precision retrieval, the attention structure block combines channel attention structure and spatial attention structure to build the attention structure block. The last feature descriptor is then created by combining various semantic feature levels. The proposed strategy is practical and efficient, as demonstrated by the experimental findings on the comparison database Flickr15k. In addition, in the task of hand-drawn image classification, the proposed attention mechanism greatly improves the classification accuracy. [ABSTRACT FROM AUTHOR]

In order to solve the most difficult problem of the architectural model established by CNN in solving specific problems, which results in parameter overflow and inefficient training, an optimization algorithm for nonlinear convolutional neural networks based on improved chameleon swarm algorithm is proposed. This article mainly introduces the use of Chameleon Swarm Optimization (PSO) algorithm to research the parameters of CNN architecture, solve them, and achieve the optimization of the optimization model.Although the number of parameters that need to be set up in CNN is very large, this method can find better testing space for Alexnet samples with 5 different images. In order to improve the performance of the improved pruning algorithms, two candidate pruning algorithms are also proposed. The experimental results show that compared with the traditional Alexnet model, the improved pruning method improves the image recognition ability of the Caffe primary parameter set from 1.3% to 5.7%. Conclusion: This method has wide applicability and can be applied to most neural networks which do not require any special functional modules of the Alexnet network model. [ABSTRACT FROM AUTHOR]

The features of the kernel extreme learning machine—efficient processing, improved performance, and less human parameter setting—have allowed it to be effectively used to batch multi-label classification tasks. These classic classification algorithms must at present contend with accuracy and space–time issues as a result of the vast and quick, multi-label, and concept drift features of the developing data streams in the practical application sector. The KELM training procedure still has a difficulty in that it has to be repeated numerous times independently in order to maximize the model's generalization performance or the number of nodes in the hidden layer. In this paper, a kernel extreme learning machine multi-label data classification method based on the butterfly algorithm optimized by particle swarm optimization is proposed. The proposed algorithm, which fully accounts for the optimization of the model generalization ability and the number of hidden layer nodes, can train multiple KELM hidden layer networks at once while maintaining the algorithm's current time complexity and avoiding a significant number of repeated calculations. The simulation results demonstrate that, in comparison to the PSO-KELM, BBA-KELM, and BOA-KELM algorithms, the PSOBOA-KELM algorithm proposed in this paper can more effectively search the kernel extreme learning machine parameters and more effectively balance the global and local performance, resulting in a KELM prediction model with a higher prediction accuracy. [ABSTRACT FROM AUTHOR]

Inspired by insects in nature, an increasing number of soft robots have been proposed to mimic their locomotion patterns. As a wireless actuation method, the magnetic actuation technique has been widely applied to drive soft magnetic robots for diverse applications. Although recent works on soft materials have stimulated the development of soft robots, it is challenging to achieve the efficient movement of soft robots for in vivo biomedical application. Inspired by centipede locomotion, a soft octopodal robot is designed in this paper. The robot is fabricated by mixing magnetic particles with silicone polymers, which is then magnetized by a specific magnetic field. The prototypes can be actuated by an external magnetic field (5–8 mT) produced by custom-made electromagnetic coils. Experimental results show that the soft robot can move at a high speed in the range of 0.536–1.604 mm/s on different surfaces, including paper, wood, and PMMA. This indicates that the soft robot can achieve comparable speeds to other robots, while being driven by a lower magnitude, resulting in energy savings. Furthermore, it achieves a high speed of 0.823 mm/s on the surface of a pig colon. The fine capabilities of the soft robot in terms of crossing uneven biological surfaces and carrying external loads are demonstrated. The results indicate that the reported soft robot exhibits promising applications in the biomedical field. [ABSTRACT FROM AUTHOR]

The purpose of image dehazing is to remove the interference from weather factors in degraded images and enhance the clarity and color saturation of images to maximize the restoration of useful features. Single image dehazing is one of the most important tasks in the field of image restoration. In recent years, due to the progress of deep learning, single image dehazing has made great progress. With the success of Transformer in advanced computer vision tasks, some research studies also began to apply Transformer to image dehazing tasks and obtained surprising results. However, both the deconvolution-neural-network-based dehazing algorithm and Transformer based dehazing algorithm magnify their advantages and disadvantages separately. Therefore, this paper proposes a novel Transformer–Convolution fusion dehazing network (TCFDN), which uses Transformer's global modeling ability and convolutional neural network's local modeling ability to improve the dehazing ability. In the Transformer–Convolution fusion dehazing network, the classic self-encoder structure is used. This paper proposes a Transformer–Convolution hybrid layer, which uses an adaptive fusion strategy to make full use of the Swin-Transformer and convolutional neural network to extract and reconstruct image features. On the basis of previous research, this layer further improves the ability of the network to remove haze. A series of contrast experiments and ablation experiments not only proved that the Transformer–Convolution fusion dehazing network proposed in this paper exceeded the more advanced dehazing algorithm, but also provided solid and powerful evidence for the basic theory on which it depends. [ABSTRACT FROM AUTHOR]

Dual‐layer LCDs have the advantage of high contrast ratio, but due to the physical distance between the two LC panels, parallax error may appear when viewed off‐axis, affecting the display quality. In this paper, we propose a convolutional neural network‐based algorithm to mitigate parallax error and improve the display quality. First, a suitable training data set is selected and the input image is processed by a convolutional neural network to output a front image and a rear image. These two images are staggered and multiplied to obtain reproduced images at different angles. The reproduced image is compared with the original image by means of a loss function, and the parameters in the convolution are continuously updated to achieve the best display quality. The algorithm proposed in this paper can effectively mitigate the parallax error phenomenon of a dual‐layer LCD and improve the display quality, and the computational efficiency is also higher than the traditional algorithms. [ABSTRACT FROM AUTHOR]

Long-staple cotton from Xinjiang is renowned for its exceptional quality. However, it is susceptible to contamination with plastic film during mechanical picking. To address the issue of tricky removal of film in seed cotton, a technique based on hyperspectral images and AlexNet-PCA is proposed to identify the colorless and transparent film of the seed cotton. The method consists of black and white correction of hyperspectral images, dimensionality reduction of hyperspectral data, and training and testing of convolutional neural network (CNN) models. The key technique is to find the optimal way to reduce the dimensionality of the hyperspectral data, thus reducing the computational cost. The biggest innovation of the paper is the combination of CNNs and dimensionality reduction methods to achieve high-precision intelligent recognition of transparent plastic films. Experiments with three dimensionality reduction methods and three CNN architectures are conducted to seek the optimal model for plastic film recognition. The results demonstrate that AlexNet-PCA-12 achieves the highest recognition accuracy and cost performance in dimensionality reduction. In the practical application sorting tests, the method proposed in this paper achieved a 97.02% removal rate of plastic film, which provides a modern theoretical model and effective method for high-precision identification of heteropolymers in seed cotton. [ABSTRACT FROM AUTHOR]

Crack detection is an important component of dam safety monitoring. Detection methods based on deep convolutional neural networks (DCNNs) are widely used for their high efficiency and safety. Most existing DCNNs with high accuracy are too complex for users to deploy for real-time detection. However, compressing models face the dilemma of sacrificing detection accuracy. To solve this dilemma, an improved residual neural network (ResNet)-based algorithm for concrete dam crack detection using dynamic knowledge distillation is proposed in this paper in order to obtain higher accuracy for small models. To see how well distillation works, preliminary experiments were carried out on mini-ImageNet. ResNet18 was trained by adding additional tasks to match soft targets generated by ResNet50 under dynamic high temperatures. Furthermore, these pre-trained teacher and student models were transferred to experiments on concrete crack detection. The results showed that the accuracy of the improved algorithm was up to 99.85%, an increase of 4.92%. [ABSTRACT FROM AUTHOR]

Convolutional neural networks have achieved good results in target detection in many application scenarios, but convolutional neural networks still face great challenges when facing scenarios with small target sizes and complex background environments. To solve the problem of low accuracy of infrared weak target detection in complex scenes, and considering the real-time requirements of the detection task, we choose the YOLOv5s target detection algorithm for improvement. We add the Bottleneck Transformer structure and CoordConv to the network to optimize the model parameters and improve the performance of the detection network. Meanwhile, a two-dimensional Gaussian distribution is used to describe the importance of pixel points in the target frame, and the normalized Guassian Wasserstein distance (NWD) is used to measure the similarity between the prediction frame and the true frame to characterize the loss function of weak targets, which will help highlight the targets with flat positional deviation transformation and improve the detection accuracy. Finally, through experimental verification, compared with other mainstream detection algorithms, the improved algorithm in this paper significantly improves the target detection accuracy, with the mAP reaching 96.7 percent, which is 2.2 percentage points higher compared with Yolov5s. [ABSTRACT FROM AUTHOR]

The quality of rolling bearings is vital for the working state and rotation accuracy of the shaft. Timely and accurately acquiring bearing status and early fault diagnosis can effectively prevent losses, making it highly practical. To improve the accuracy of bearing fault diagnosis, this paper proposes a CNN-LSTM bearing fault diagnosis model optimized by hybrid particle swarm optimization (HPSO). The HPSO algorithm has a strong global optimization ability and can effectively solve nonlinear and multivariate optimization problems. It is used to optimize and match the parameters of the CNN-LSTM model and dynamically find the optimal value of the parameters. This model overcomes the problem that the parameters of the CNN-LSTM model depend on empirical settings and cannot be adjusted dynamically. This model is used for bearing fault diagnosis, and the accuracy rate of fault diagnosis classification reaches 99.2%. Compared with the traditional CNN, LSTM, and CNN-LSTM models, the accuracy rates are increased by 6.6%, 9.2%, and 5%, respectively. At the same time, comparing the models with different optimization parameters shows that the model proposed in this paper has the highest accuracy. The experimental results verified the superiority of the HPSO algorithm to optimize model parameters and the feasibility and accuracy of the HPSO-CNN-LSTM model for bearing fault diagnosis. [ABSTRACT FROM AUTHOR]

Dedicated equipment, which is widely used in many different types of vehicles, is the core system that determines the combat capability of special vehicles. Therefore, assuring the normal operation of dedicated equipment is crucial. With the increase in battlefield complexity, the demand for equipment functions is increasing, and the complexity of dedicated equipment is also increasing. To solve the problem of fault diagnosis of dedicated equipment, a fault diagnosis algorithm based on CNN-LSTM was proposed in this paper. CNN and LSTM are used in the model adopted by the algorithm to extract spatial and temporal features from the data. CBAM is used to enhance the model's accuracy in identifying faults for dedicated equipment. Data on dedicated equipment faults were obtained from a hardware-in-loop simulation platform to verify the model. It is demonstrated that the proposed fault diagnosis algorithm has high recognition ability for dedicated equipment by comparing it to other neural network models. [ABSTRACT FROM AUTHOR]

Applying ensemble learning to facial expression recognition is an important research field nowadays, but all may not be better than many, the redundant learners in the classifier pool may hinder the ensemble system's performance, so ensemble pruning is needed. Ensemble pruning selects the most suitable subset of classifiers to classify test samples according to the classifier competence. However, the noisy and redundant samples in the validation set will often adversely affect the evaluation of the classifier, making it impossible to select the most suitable classifier. In this paper, a novel ensemble pruning algorithm based on clustering soft label optimization and sorting for facial expression recognition is proposed. First, to increase classifier evaluation objectivity, the novel method uses the clustering optimization model to perform prototype selection and classifier clustering simultaneously. Then the accuracy-based ordering is employed to remove the redundant or poor quality learners, and keep a balance between diversity and accuracy of the ensemble system. Experimental results show that the proposed method outperforms or competes with some state-of-the-art methods on several typical facial expression datasets. [ABSTRACT FROM AUTHOR]

Very recently, the BERT graph convolutional network (BertGCN) model has attracted much attention from researchers due to its good text classification performance. However, just using original documents in the corpus to construct the topology of graphs for GCN-based models may lose some effective information. In this paper, we focus on sentiment classification, an important branch of text classification, and propose the multistream BERT graph convolutional network (MS-BertGCN) for sentiment classification based on cross-document learning. In the proposed method, we first combine the documents in the training set based on within-class similarity. Then, each heterogeneous graph is constructed using a group of combinations of documents for the single-stream BertGCN model. Finally, we construct multistream-BertGCN (MS-BertGCN) based on multiple heterogeneous graphs constructed from different groups of combined documents. The experimental results show that our MS-BertGCN model outperforms state-of-the-art methods on sentiment classification tasks. [ABSTRACT FROM AUTHOR]

Facility Management is a multi-disciplinary task in which coordination is key to attaining success during the building life cycle and for which technology assumes an increasing role. This sector is demanding more available and accurate tools to optimize the management process, decrease the probability of failure, and reduce the time spent on anomaly analysis. So, the present paper presents work developed to improve access to building anomaly recognition and to predict the building degradation state in an automatized way. The methodology applied to achieve this goal started with a survey and digital data acquisition from a case study, followed by the automatized detection of building anomalies using supervised classification in Deep Learning; then, the early diagnosis of threatening conditions for building degradation took place using degradation curves based on data records and regression algorithms. The results drive this study a step forward toward obtaining advanced tools for Facility Management based in Artificial Intelligence, able to provide the most appropriate moment at which to intervene according to the cost-benefit. The present work provided better results on the harmonic mean of precision and recall when compared with previous studies of image classification for the construction sector. Moreover, the mathematical functions for the prediction of future degradation based on the data field for each construction system were presented and can be applied to the typologies of other buildings. In the end, future developments and limitations are highlighted. [ABSTRACT FROM AUTHOR]

Due to the lack of other component information in traditional magnetic leakage signal defects and the low accuracy of prediction methods, this paper proposes an improved residual network for magnetic leakage detection defect recognition method that predicts defect size and different detection speeds. A new defect diagnosis method based on ResNet18 on the Convolutional Neural Network (CNN) is proposed in this study. This method transfers the pre-trained ResNet18 network and replaces the activation function in the transferred network structure. It extracts features from transformed two-dimensional images obtained by converting the original experimental signals and signals with added noise, removing the influence of manual features. The results demonstrated that the improved ResNet18 network model, after transfer learning, achieved 100% prediction accuracy for all 10,000 grayscale images generated with defect lengths of 50 mm; width of 2 mm; and depths of 2 mm, 4 mm, 6 mm, and 8 mm. Moreover, the prediction accuracies for the quasi-static, slow, compensated fast, and fast scanning speeds were 99.20%, 98.50%, 93.30%, and 94.00%, respectively, for defect depths of 2 mm, 4 mm, 6 mm, and 8 mm. These accuracies surpass those of other models, demonstrating the significant improvement in prediction accuracy achieved by this method. [ABSTRACT FROM AUTHOR]

With the popularization of the network and the expansion of its application scope, the problem of abnormal network traffic caused by network attacks, malicious software, traffic peaks, or network device failures is becoming increasingly prominent. This problem not only leads to a decline in network performance and service quality but also may pose a serious threat to network security. This paper proposes a hybrid data processing model based on deep learning for network anomaly detection to improve anomaly detection performance. First, the Grey Wolf optimization algorithm is improved to select high-quality data features, which are then converted to RGB images and input into an anomaly detection model. An anomaly detection model of network dataflow based on a convolutional neural network is designed to recognize network anomalies, including DoS (Denial of Service), R2L (Remote to Local), U2R (User to Root), and Probe (Probing). To verify the effectiveness of the improved Grey Wolf algorithm and the anomaly detection model, we conducted experiments on the KDD99 and UNSW-NB15 datasets. The proposed method achieves an average detection rate of 0.986, which is much higher than all the counterparts. Experimental results show that the accuracy and the detection rates of our method were improved, while the false alarm rate has been reduced, proving the effectiveness of our approach in network anomaly classification tasks. [ABSTRACT FROM AUTHOR]

Image processing algorithms are essential for clarifying the image and improving the ability to recognize distinct characteristics of the image. The field of digital image processing is widespread in several research and technology applications. In many of these applications, the existence of impulsive noise in the obtained images is one of the most frequent problems. The median filter is a strong method to remove the impulsive noise; it effectively eliminates salt and pepper noise from the image. The main target of this paper is to investigate efficient median filter units to be connected to a general-purpose processor (GPP) for FPGA-based embedded systems. The paper exposes three novel techniques, two of them specially for median filtering techniques and the third one is used to get the maximum number of any 9 elements array. The proposed algorithms are inspired by the Median Of Median (MOM) algorithm. The first two techniques are tested for filtering 3 × 3 image windows and optimized for producing the expected result in high accuracy, short time, and reduced number of comparisons. The last technique is tested for a 9 elements array for extracting the maximum number in same high efficiency manner. Furthermore, the three proposed techniques are implemented leveraging the advantage of the parallel processing and the FPGA flexible resources to satisfy the real-time processing constraints. A comparison between the first two proposed filtering units and their counterparts in the literature is included. The comparison reveals the superiority of the first technique in terms of accuracy with fewer comparators than previously published techniques. Besides, the paper illustrates how the concept beyond the proposed techniques can be used to perform the maximum pooling for convolution neural networks. [ABSTRACT FROM AUTHOR]