Search

Few-shot learning is attracting extensive research because of its ability to classify only a few co-trainable samples. Current few-shot learning approaches focus on learning class prototypes representation to solve problems by a simple averaging approach, but this approach ignores intra-class differences. In this paper, we propose a new weighted prototype network for few-shot learning. Our model consists of two modules, feature extraction and prototype modification. We first construct graphs from the embeddings obtained from the feature extraction module. Then we fed these graphs into graph neural networks in order to explore the contribution of each sample to its class prototype from the graph structure. The experimental results on three benchmark datasets show that our proposed model is comparable to the state-of-the-art few-shot learning approach. • Introduced a novel approach for few-shot image classification. • Innovatively transformed the process of assessing sample importance into graph label propagation. • Achieved performance improvements on few-shot benchmark datasets. [ABSTRACT FROM AUTHOR]

Deep learning is widely used in Bearing Fault Diagnosis (BFD). Nonetheless, practical industrial production often generates a large amount of industrial noise. These noises exhibit randomness and complexity, which puts forward higher requirements for diagnosis algorithms. Certain studies have tackled the issue of anti-interference in high-noise environments (S N R ≤ 0 dB) by increasing the complexity of the model. However, due to the excessive number of parameters and computation, such models cannot be deployed on low-end edge devices. Balancing resource consumption and accuracy has become a major challenge in BFD modeling research. To solve the above problems, this paper proposes a new Transformer architecture model called LTFAFormer. The LTFAFormer is capable of achieving high-precision diagnostics on low-end edge devices and shows greater noise resistance. In terms of processing sequence information, Transformer has proven to be superior to other solutions. However, when dealing with longer sensor signal data containing complex noise, the traditional self-attention mechanism not only cannot effectively extract fault features, but also generates more computational complexity than CNN. To address this issue, we propose a novel time–frequency dual-channel parallel attention mechanism. Our approach enhances the feature extraction capability of the model by expanding the attention computation scale and reduces the computational resource consumption of the model by optimizing the model structure. To validate the effectiveness of LTFAFormer, we present two cases to demonstrate that LTFAFormer has higher prediction accuracy while satisfying lightweight. Especially in high-noise environments, LTFAFormer has stronger robustness. In this paper provides a new set of feasible strategies for the practical deployment of BFD models in practical industrial environments. • The LTFAFormer, a lightweight and noise-robust model, is ideal for low-end devices. • Added a Dual-Channel Attention Module to expand traditional self-attention scale. • Developed a Cross-fusion mechanism for attention to enhance feature fusion. • Optimized the model's embedding module based on sensor signal traits. • Our model outperforms others in bearing fault diagnosis and is robust in noise. [Display omitted] [ABSTRACT FROM AUTHOR]

Non invasive load monitoring (NILM) is beneficial for enhancing the monitoring capability of the distribution network and is crucial for improving the safety of smart grid operation. However, household appliances involve a variety of devices and a large amount of data, making it difficult to achieve high-precision load identification. Especially when a large amount of loads are running simultaneously, problems such as feature overlap and reduced distinguish ability may occur, and thus increasing the difficulty of load identification. To solve these problems, this paper proposes a new NILM method based on dynamic time warping (DTW) optimization and event detection. Firstly, a feature extraction algorithm of STFT-SSAE is constructed by using short-time Fourier transform (STFT) to extract time–frequency features from the load, and then by sparse stack autoencoder (SSAE) to extract important features from time–frequency information. Secondly, the above features are input into Bi-LSTM and DTW models respectively, and a new probabilistic model is established. A Bi-LSTM-DTW load recognition architecture is built by combining the two models. Finally, the load identification model of SSAE-Bi-LSTM based on DTW optimization (DOSL) is trained by the preset combined data, which ensures the high confidence of the DOSL model in various complex operating scenarios. This paper introduces the public data set PLAID for experimental verification. The results show that the selected load features has good discrimination, and the proposed algorithm has the best identification effect with the accuracy rate of 0.9412. The proposed algorithm is also validated for generalization ability on the public dataset UK-DALE, with a identification accuracy rate of 0.9306, and the error of different datasets is only about 1%. • A complete multiple house-hold appliances NILM framework. • A novel STFT-SSAE algorithm for load feature extraction. • Bi-LSTM-DTW load identification architecture combined with output probability vector. [ABSTRACT FROM AUTHOR]

The past decade has witnessed significant progress in detecting objects by using enormous features of deep learning models. But, most of the existing models are unable to detect x-small and dense objects, due to the futility of feature extraction, and substantial misalignments between anchor boxes and axis-aligned convolution features, which leads to the discrepancy between the categorization score and positioning accuracy. This paper introduces an anchor regenerative-based transformer module in a feature refinement network to solve this problem. The anchor-regenerative module can generate anchor scales based on the semantic statistics of the objects present in the image, which avoids the inconsistency between the anchor boxes and axis-aligned convolution features. Whereas, the Multi-Head-Self-Attention (MHSA) based transformer module extracts the in-depth information from the feature maps based on the query, key, and value parameter information. This proposed model is experimentally verified on the VisDrone, VOC, and SKU-110K datasets. This model generates different anchor scales for these three datasets and achieves higher mAP, precision, and recall values on three datasets. These tested results prove that the suggested model has outstanding achievements compared with existing models in detecting x-small objects as well as dense objects. Finally, we evaluated the performance of these three datasets by using accuracy, kappa coefficient, and ROC metrics. These evaluated metrics demonstrate that our model is a good fit for VOC, and SKU-110K datasets. [ABSTRACT FROM AUTHOR]

Person re-identification is to recognize a pedestrian who has been observed at different places in monitoring system, which is done by computer automatically. Important practical significance can be found in this paper on improving the intelligence of monitoring system. There are two challenging problems needed to be solved for person re-identification, which is feature representation and matching. Designing a suitable feature representation method is both challenging and crucial. Ideally, the extracted features should be robust enough to cope with the various situations of the monitoring environment, such as posture difference, illumination changing, and shooting angle difference, etc. There are a lot of challenges for feature matching too, such as small sample size, inter-class confusion, and intra-class variation, etc. In order to meet these challenges, a new person re-identification scheme was put forward in this paper. The following three aspects would be used to describe it. First, an improved BOF feature extraction algorithm based on SURF was put forwarded, SURF algorithm extracts the preliminary feature descriptor and generates visual dictionary, so the influence factors such as illumination and scale invariants can be deal with. And then, covariance descriptor was adopted by the algorithm. It has high robustness, and matching accuracy can be improved too for the case of small sample. Finally, an effective classifier was designed by LIBSVM based on improved BOF algorithm, so the efficiency of the person re-identification algorithm can be improved. The proposed method was compared with the current mainstream algorithm through experiment, and it can be found through the experimental results that it is effective to solve the difficult problems for person re-identification. [ABSTRACT FROM AUTHOR]

The rapid advancement of intelligent theories and models, exemplified by deep learning, has achieved remarkable success across numerous fields. However, given that the complexity and variability of offshore wind turbine systems, particularly in the context of high-power variable-frequency control of insulated bearings in offshore wind turbines, the problem of fault identification has become a recognized technical challenge. Additionally, fully exploiting the temporal characteristics of insulated bearing fault data poses a key problem that demands urgent resolution. To address these gaps, this paper pioneers a novel Lightweight Temporal Feature-focused framework, named LTFM-Net, aimed at solving the difficult problem of identifying insulated bearing faults in offshore wind turbines in practical engineering applications. Specifically, this framework enables the intelligent identification of insulated bearing faults under harsh operating conditions such as alternating voltage and variable loads, marking a first in this field. Furthermore, an innovative strategy named Weighted Diminish Recurrent Unit (WDRU) was developed, along with the derivation of its backpropagation formula, which is innovatively applied to the feature extraction module of the LTFM-Net framework for the first time. Thus, an efficient method for acquiring fault data from insulated bearings in offshore wind turbines was proposed. Based on a unified dataset, the diagnostic performance of the LTFM-Net framework was evaluated and compared with seven advanced methods. The results demonstrate that the LTFM-Net achieves precise identification of insulated bearing faults, confirming its excellent generalization, robustness, and superiority. The introduction of t-SNE for visualizing the fault characteristics of insulated bearings uncovered by the LTFM-Net framework further enhances its reliability, accuracy, and credibility. • This study pioneered innovation by developing a lightweight temporal feature focusing framework, aimed at addressing the challenge of identifying faults. • This research facilitates the first intelligent identification of such faults, offering fresh insights into fault detection for insulated bearings and failure modes of insulated bearings. • It innovatively develops the Weight Reducing Recursive Unit (WDRU) strategy combined with an updated back propagation formulation. [ABSTRACT FROM AUTHOR]

• The SAM was first applied to domain-adaptive semantic segmentation. A new semantic segmentation method for mechanical assembly was proposed, and a U-SAM framework was designed to extract shallow features and improve convergence efficiency. • A strategy of replacing regular boxes with oblique boxes and an iterative loop strategy was proposed to address the prompt issue in SAM. • A line discriminator module and a trainable line-guided filtering module were proposed to further improve the accuracy of edge segmentation. Semantic segmentation of assembly images can effectively identify and prevent non-standard behaviors such as wrong assembly and missing assembly in the assembly process. However, supervised methods require massive training samples and are time-consuming to label. Unsupervised domain adaptation methods can address this issue without labeling the target data. Recently, the large-scale Segment Anything Model (SAM) has achieved significant success in semantic segmentation owing to its powerful feature extraction and instant interaction capabilities. Based on this, this paper proposed a new framework called U-SAM for unsupervised domain-adaptive semantic segmentation. Compared to other frameworks, U-SAM can obtain more segmented features while maintaining shallower features. Meanwhile, this study proposed an iterative loop strategy for the prompt module in U-SAM. This strategy gradually improves the accuracy of target prompt labels and solves the problem of no prompt labels in the target data. Additionally, a novel method of using oblique boxes instead of regular boxes was developed to achieve higher box prompt quality. Finally, a line discriminator module and a trainable line-guided filtering module were designed to improve the quality of edge segmentation. The experimental results indicate that the U-SAM network framework achieves a Dice accuracy of 84.4% for semantic segmentation on the mechanical assembly image dataset. This confirmed that our proposed method can be applied to the assembly monitoring process of target data. [ABSTRACT FROM AUTHOR]

Apple plays a vitally important role in human life and is considered one of the most nutritious fruits. However, the quality and production of the apple industry are seriously restricted by apple leaf diseases and the disease lesions are hard to detect because they often have various scales and deformable geometry. To solve the above problem, this paper proposed a novel Multi-scale Constrained Deformable Convolution Network(MCDCNet), which takes advantage of multi-branch convolution and deformable convolution. Firstly, the novel two-branch convolution network is presented to enhance the discriminatory ability of models for extracting different scales of apple leaf disease. Secondly, different offset intervals are applied to the two kernels of the dual convolution channel separately, which makes the proposed model pay more attention to the deformable geometry features of the lesions and avoid extra weight parameters. Finally, a feature fusion module is constructed to achieve automatic detection of multi-scale apple leaf disease, which combines the output features from the dual convolution channels and performs dimensional operations on the channel dimensions of the feature map. Under the complex natural environment, the accuracy value of the proposed model can reach 66.8%, which is an improvement of 3.85% compared to the existing SOTA models. The experiment results established that MCDCNet has a better feature extraction capability and can efficiently and accurately detect 5 common apple leaf diseases in the natural environment. • The proposed MCDCNet can extract more reliable features of apple leaf diseases with various scales and geometry which effectively improve the discriminative ability of the network. • A novel Dual-constrained deformable convolution module is proposed to help the network get flexible receptive fields and help network dealing with apple leaf disease with various geometry and size. • A novel Feature fusion module is proposed to fuse outputs from dual branches which helps the MCDCNet automatically selecting appreciate geometry and scales of apple leaf disease. [ABSTRACT FROM AUTHOR]

As tidal stream energy is clean, sustainable, and can be extracted by tidal stream turbine (TST), many countries have paid more attention to it. However, the imbalance faults, caused by the rotor attachment of TST, have an adverse effect on power generation efficiency. Therefore, rotor attachment detection is of great importance for the operation and maintenance of TST. This paper proposes a new method based on asymmetric depthwise convolution and convolution block attention module (ADC-CBAM) to solve this problem. First, rotor images of TST are preprocessed by normalization and centering. Then, asymmetric depthwise convolution (ADC) separates the convolution structure and combines depthwise convolution to extract features and reduce the convolution computation. Furthermore, the convolution block attention module (CBAM) focuses on the feature region to reduce redundant information. Finally, the proposed method is used to detect the rotor attachment. The experimental results demonstrate that the method has higher accuracy and is more lightweight than other state-of-the-art lightweight methods. Moreover, the generalization is verified on the public dataset CIFAR-10. • A rotor attachment detection method based on ADC-CBAM is proposed for TST by combining the ADC and CBAM. • The ADC is proposed to solve the problem of heavy convolution computation and feature extraction of rotor attachment. • The CBAM is studied in the rotor attachment method to focus on the feature region and reduce the redundant information of rotor images. [ABSTRACT FROM AUTHOR]

Ancient Chinese character recognition is a challenging problem in the field of pattern recognition. It is difficult to collect all character classes during the training stage due to the numerous classes of ancient Chinese characters and the likelihood of discovering new characters over time. A solution to address this problem is continual learning. However, most continual learning methods are not well-suited for large-scale applications, making them insufficient for solving the problem of ancient Chinese character recognition. Although saving raw data for old classes is a good approach for continual learning to address large-scale problems, it is often infeasible due to the lack of data accessibility in reality. To solve these problems, we propose a large-scale continual learning framework based on the convolutional prototype network (CPN), which does not save raw data for old classes. In this paper, several basic strategies have been proposed for the initial training stage to enhance the feature extraction ability and robustness of the network, which can improve the performance of the model in continual learning. In addition, we propose two practical methods in varying feature space (parameters of feature extractor are changeable) and fixed feature space (parameters of feature extractor are fixed), which enable the model to carry out large-scale continual learning. The proposed method does not save the raw data of old classes and enables simultaneous classification of all existing classes without knowing the incremental batch number. Experiments on the CASIA-AHCDB dataset with 5000 character classes demonstrate the effectiveness and superiority of the proposed method. • We propose a prototype-based framework for large-scale continual learning. • The proposed strategies in the initial stage are good for continual learning. • We propose two continual learning methods in varying and fixed feature space. • The proposed methods achieve SOTA results on the CASIA-AHCDB dataset with 5000 classes. [ABSTRACT FROM AUTHOR]

Most of the current tunnel defect detection methods do not perform repeated inspection of the defects taken at different times at the same location, which is a major deficiency for detecting the development trend of the defects. To solve the problem of repeated inspection of tunnel lining defects, this paper proposes a new method based on the combination of enhanced visual odometer and a Sequence-Siamese network(Seq-Siamese-Net) to identify and locate tunnel lining defects. We first detect the defects in the tunnel lining surface video by You Only Look Once network, then estimate the general location of the defects by using the proposed enhanced visual odometer, and finally obtain the most matching defects in the vicinity of the defects by Seq-Siamese-Net to achieve the repeated inspection of the tunnel lining surface defects. Experiments show that our proposed enhanced visual odometer method outperforms the original visual odometer method in terms of localization accuracy, while the feature extracted form Seq-Siamese-Network is more effective than the average-pooled feature in the tunnel defect image sequence matching task. Furthermore, experiments in a real tunnel demonstrate the practical value of the proposed method. • A network-based method for repeated detection of tunnel lining defects is proposed. • Average error of localization method reached about 1% at real tunnel experiment. • Sequence-Siamese-Network surpasses averaged feature in image sequence matching task. [ABSTRACT FROM AUTHOR]

Aiming at the problems of difficult feature extraction and low recognition accuracy of coal and gangue, this paper proposes a recognition method based on dielectric and geometric features, which uses variational mode decomposition (VMD) to denoise the response signal to obtain its dielectric features; uses the marked watershed method to segment the image boundary to obtain its geometric features. The dielectric and geometric characteristic parameters of coal and gangue are extracted to construct the classification model of support vector machines (SVM). It takes only 0.001483 s for the secondary SVM classification model to process 60 groups of test samples, and the recognition accuracy is 98.33%, which is superior to the constructed SVM classification model, k-nearest neighbor (KNN) classification model, and back propagation neural network (BP) classification model. This method opens up a new way to solve the problem of coal and gangue identification. [Display omitted] • A recognition model of coal and gangue based on dielectric characteristics and geometric constraints is proposed. • The effective thickness and position of coal and gangue in capacitor should be strictly controlled. • The geometric features of coal and gangue are extracted by industrial camera. • A VMD- Kalman denoising method is proposed. • Quadratic support vector machine algorithm has excellent performance in processing eigenvectors. [ABSTRACT FROM AUTHOR]

Accurate and robust localization is one of the most basic functional modules of mobile robots. A recent promising approach is to employ a camera that captures ground textures to achieve high-precision localization. However, low robustness and real-time performance are disadvantages of the system due to the time-consuming front-end feature extraction and search methods. To solve these problems, a comprehensive system that has high robustness and real-time performance is proposed, including offline mapping, global localization, local localization and map update. This paper summarizes a method for selecting the optimal front-end for different textures. A fast global localization method based on bag of words is proposed, which can quickly and effectively determine the local positioning area. Experiments are carried out on public datasets and actual indoor scenes to verify the practicability of the system. [ABSTRACT FROM AUTHOR]

During recent years, deep convolutional neural networks (CNNs) have become increasingly popular in sensor-based human activity recognition (HAR), due to powerful feature extraction capacity. However, the required high computational cost has hindered their practical deployment in real-world applications. To address this issue, this paper presents a novel approach named ABCSearch to prune channels through an artificial bee colony (ABC) algorithm, which is a meta-heuristic optimization algorithm inspired by the foraging behavior of honeybees. Unlike traditional pruning strategies that prioritize channel importance, the proposed method seeks to identify the optimal number of channels at each layer, i.e., the optimal pruning structure. To lessen human interference, we transform the search problem of the optimal pruning structure into a simple optimization problem and solve it automatically through the ABC algorithm, which can significantly decrease both model size and computational cost without compromising performance. Specifically, we validate the effectiveness of ABCSearch for a standard five-layer CNN backbone architecture, which achieves 97.05%, 98.45%, 91.71%, 78.89% classification accuracies on four HAR benchmarks, i.e., UCI-HAR, WISDM, PAMAP2, UniMiB-SHAR. Our approach can yield 87.44%, 89.41%, 89.06%, 89.59% FLOPs reductions, and 89.54%, 88.70%, 88.38%, 90.45% less memory consumptions, as well as 4.4 × , 5.0 × , 5.8 × , 6.1 × practical speedups on real hardware devices, while maintaining comparable accuracy. [ABSTRACT FROM AUTHOR]

It is one of the hot topics in the field of computer vision to estimate the density of vehicles on the road by using UAV equipped camera. Because the vehicle scale in video images is variable, the scene is complex and the change of vehicle density has strong randomness, the estimation results of vehicle density obtained by the existing methods have large errors. To solve these problems, this paper proposes a vehicle density estimation network based on multi-scale sensing of vehicle density (DSNet). The network adaptively senses the vehicle density in video images through an innovatively designed density sensing module and generates feature maps with different densities. Combined with the context information extracted from the network, the appropriate weight for each density-specific feature map is generated. By fusing the feature maps with different weights to enhance the feature of vehicle. Finally, the enhanced feature map is further refined to obtain a high-resolution density map. In addition, to solve the problem of the global counting loss (L C) that the global estimation results are good but the local estimation errors exist, we design and add the new region-aware loss (L G) to enhance the local feature extraction ability of the network and improve the robustness of the network. Experimental results show that the GAME of DSNet on the VisDrone 2019 dataset and DETRAC dataset is reduced to 5.45 and 6.07, respectively. Compared with the current state-of-the-art vehicle density estimation methods, DSNet achieves higher accuracy in the vehicle density estimation tasks of different scenes. [ABSTRACT FROM AUTHOR]

• Proposes an end-to-end network for biphasic image change detection. • Can better solve the problem of detection errors. • High degree of automation in the calculation process. • Enhanced feature extraction module was inroduced. Change detection is an effective means of monitoring surface changes. In order to solve the problem of reducing detection accuracy in remote sensing image change detection methods that only focus on extracting deep semantic information while ignoring high-resolution shallow information, this paper proposes an end-to-end multi-side output fusion deep supervised network for high-resolution biphasic remote sensing image change detection. The enhanced feature extraction module is used to increase the Receptive field and capture multi-scale features. The dense skip connection module utilizes skip connections to fuse shallow position information and deep semantic information, alleviating the problem of feature information loss. The deep supervision module with multiple side-outputs fusion utilizes the information complementarity between different scale feature maps, combined with multi-scale prediction maps, to improve the robustness of target scale changes. The combined Loss function is used to alleviate the problem of imbalance between positive and negative samples, so that more attention is paid to the learning of changing characteristics in the process of training the network. The proposed method was validated on CDD dataset and DSIFN dataset, with F1 scores of 91.58% and 86.69%, respectively, which were improved by 1.00% and 5.53% compared to the suboptimal network. [ABSTRACT FROM AUTHOR]

Graph-based semi-supervised dimension reduction can use the inherent graph structure of samples to propagate label information, and has become a hot research field in machine learning. However, most current methods have strict linear constraints and cannot handle data sampled from nonlinear manifolds; and rely on predefined graphs, which cannot capture the local structure information of data and cannot handle complex non-Gaussian data. To address these issues, this paper proposes a new locality-preserved flexible dimension reduction framework, called Semi-supervised Flexible Adaptive Graph Embedding (SFAG), which learns the embedding space that can preserve the local neighborhood structure by constructing a k 1 -nearest neighbor graph over labeled samples. Then, another k 2 -nearest neighbor graph is constructed on all samples to adaptively construct the optimal graph, clustering labeled and unlabeled embedding sample points with neighborhood relations into the same sub-manifold sharing the same label information. Last but not least, the hard linear projection constraint is relaxed by adding residual terms to obtain not only the nonlinear embedding of the training samples, but also the linear projection matrix applied directly to the out-of-sample. In addition, two different semi-supervised dimension reduction methods for adaptive construction of optimal graphs are proposed based on the SFAG framework. Several evaluation experiments validate the effectiveness of our method in exploring manifold structures and classification tasks. • A novel semi-supervised adaptive flexible dimension reduction method is proposed. • The local and global structure of the data is explored by constructing kNN graphs. • The hard linear constraints on the projection are relaxed by adding a residual term. • An algorithm is designed to solve the norm-constrained problem of feature extraction. [ABSTRACT FROM AUTHOR]

Tomato disease image identification plays a very important role in the field of agricultural production. Aiming at the problems of large intraclass differences, small inter-class differences and difficult feature extraction of some tomato leaf diseases, this paper proposes an identification of tomato leaf diseases based on LMBRNet. Firstly, a comprehensive grouped differentiated residual (CGDR) is built,The multi-branch structure of CGDR is used to capture the diversified feature information of tomato leaf diseases in different dimensions and receptive fields. then, a multiple residual connection scheme is adopted,using residuals to connect all layers, to ensure the maximum information transmission between layers in the network and to solve the problems of network degradation and gradient disappearance in the network training process. Secondly,the visual enhancement effectively fuses the results obtained by three different downsampling strategies using average pooling, max pooling, and 1*1 convolution. Avoid the loss of information caused by downsampling and improve the accuracy of the network. Moreover, deep separable convolution is used to optimize the network structure, reduce the amount of model parameters and reduce the computational resources occupied by training and deploying the model.we found that the depthwise separable convolution with a kernel size of 1*1 can slightly improve the efficiency of the model under the premise that it has little effect on the number of model parameters. The application results of more than 8000 images show that the overall identification accuracy is about 99.7%,higher than ResNet50(97.48%),GoogleNet(98.96%) etc. conventional models. The parameter amount of LMBRNet is 4.1M. Less than ResNet50(23M),GoogleNet(5.7M) etc. conventional models. It is worth noting that the accuracy of LMBRNet(99.7%) is similar to that of InceptionResNetV2(99.68%), but the amount of parameters of LMBRNet(4.1M) is much lower than that of InceptionResNetV2(54M). Moreover, the parameter amount of LMBRNet (4.1M) is slightly lower than that of MobileNetV2(2.2M), but the accuracy rate of LMBRNet(99.7%) is higher than that of MobileNetV2(97.87%). LMBRNet was tested on RS, SIW, Plantvillage-corn public datasets, all obtained high recognition accuracy, 82.32% on RS, 88.37% on SIW and 97.25% on Plantvillage-corn, indicating that LMBRNet has good generalization. Compare LMBRNet with advanced methods. In four different classification tasks, the performance of LMBRNet is similar to ResMLP12 and DCCAM-MRNet, and the difference of recognition accuracy between LMBRNet and ResMLP12 and DCCAM-MRNet is not more than 1%. However, the parameters of LMBRNet (4.1M) are lower than ResMLP12 (14.94M) and DCCAM MRNet (22.8M). LMBRNet is compared with MobileNetV3, an advanced lightweight classification model. LMBRNet(88.37% on SIW,82.32% ON RS) is used on certain datasets and performs better than MobileNetV3S(83.76% on SIW,75 on RS) and MobileNetV3L(84.34 on SIW,73.39 on RS). The parameters of LMBRNet(4.1M) are lower than MobileNetV3L(5.4M) and slightly higher than MobileNetV3S(2.9M). This indicates that LMBRNet has good generality even though it has a small number of parameters. • Complementary Grouped Dilated Residual(CGDR) solves the problem of in tomato leaf disease images. • Visual Enhancement ensures the integrity of tomato image information in downsampling module. • Deep separable convolution in the network model to reduce the number of parameters of the model. • We study the 1 * 1 deep separable convolution, which can slightly improve the efficiency of the model with less influence. [ABSTRACT FROM AUTHOR]

• A multiview method, GRMDA, is proposed for extracting the robust and sparse features, which reconstructs within-class scatter term and between-class scatter term in multiview scenario using L 2 , 1 norm to enhance robustness. • We conduct the analysis of small-class problem and adopt another criterion inspired by MMC to solve this problem. In terms of overfitting problem, the L 2 , 1 norm is also imposed on the regularization term for extracting sparse features. • The paper presents an iterative algorithm to compute optimal linear transform for each view. The poof of convergence is provided, and the computational complexity is analyzed. Several experiments were conducted to evaluate the performance of GRMDA Multiview data leads to the demand for classifying samples from various views, and the large gap between different views makes the classification task challenging. Recently, researchers have extended linear discriminant analysis (LDA) to multi-view scenarios. However, the extended methods are generally associated with the small-class problem, that is, the projection size is limited by the number of classes. In addition, they are sensitive to variations in images or outliers. To solve these problems, this study proposes a generalized robust multiview discriminant analysis (GRMDA) to obtain a linear transform for each view and for learning multiview jointly sparse discriminant common subspace. GRMDA aims to achieve both maximal between-class and minimal within-class variation for data from multiple views in a common space. Instead of formulating the ratio trace problem, we reformulate GRMDA inspired by maximum margin criterion (MMC) to address the small-class problem. Moreover, the proposed method achieves stronger robustness by reconstructing the within-class and between-class scatter terms from the definition of L 2 , 1 norm. Furthermore, GRMDA ensures joint sparsity using the L 2 , 1 norm-based regularization term. Additionally, we present an iterative algorithm, convergence proof, and complexity analysis. Experiments on six popular databases, that is, COIL100, USPS/MNIST, Extended Yale Face B, AR, BBCSport, and multiple feature datasets, were conducted to evaluate the performance of GRMDA against the state-of-the-art multiview methods. The experimental results demonstrate that the proposed method can achieve a significant performance with strong robustness and fast convergence. [ABSTRACT FROM AUTHOR]

In recent years, many deep transfer learning methods have been widely used in bearing fault diagnosis under varying working conditions to solve the problem of data distribution shift. However, the deep transfer learning methods used in bearing fault diagnosis have low accuracy when source domain data differ much and feature distribution misaligns, so a deep multi-source transfer learning model is proposed in this paper. To solve the problem of the large discrepancy in source domain data, we use maximum mean discrepancy (MMD) to select suitable source domain data to form a new source domain to help model training, we utilize independent domain-specific feature extractors to extract domain features to avoid poor source domains affecting feature extraction of other domains. At the same time, classifier output is aligned using Wasserstein distance to reduce the probability of misclassification of boundary samples, and the weighted decision is used to impose more weight on the better source domains. In addition, aiming at the problem of data distribution misalignment, we propose an alignment method combining MMD, local maximum mean discrepancy (LMMD), and triplet loss. Experimental results show that the proposed model can achieve more than the accuracy of 95% in the four working conditions of the Paderborn bearing fault dataset, and is suitable for bearing fault diagnosis under varying working conditions. [ABSTRACT FROM AUTHOR]

• The GS-TQWT UGW defect features extraction model is proposed. • A simulation signal conforming to the characteristics of UGW is constructed. • The adaptive parameters selection of the GS-TQWT model is completed. • The non-convex penalty function is used for optimization. Ultrasonic guided wave （UGW） is suitable for defect detection of long weld, but it is difficult to extract defect echo features due to dispersion, multi-mode, background noise and structural noise. To solve this problem, a group sparse tunable Q-factor wavelet transform (GS-TQWT) model is proposed in this paper. Firstly, it is revealed that the defect echo of UGW nondestructive testing (NDT) has the characteristic of group sparsity. Based on this, the UGW defect features extraction model of the GS-TQWT is established. Then, a simulation signal is constructed according to the dispersion and attenuation characteristics of UGW, and the adaptive selection of the GS-TQWT optimal parameters has been completed based on the simulation signal. Moreover, the majorization-minimization (MM) algorithm is used to solve the model. Finally, the experiment of UGW weld defect detection was carried out to verify the effectiveness of the GS-TQWT model. [ABSTRACT FROM AUTHOR]

In this paper, we propose ACINR-MVSNet, an end-to-end trainable framework with adaptive group-wise correlation and implicit neural depth refinement for multi-view stereo (MVS). Previous learning-based MVS methods have demonstrated their outstanding performance, and most of them estimate depth maps in a coarse-to-fine manner. However, in a commonly used multi-stage cascaded framework, the previous wrong estimation might lead to error propagation. In contrast, we focus on another coarse-to-fine structure, i.e., one-stage MVS architecture followed by refinement modules. Inspired by implicit neural representation, we propose an implicit neural refinement module to refine the coarse depth map. Guided by the corresponding reference image, it can better recover finer details, especially those in boundary areas. To solve the visibility problem in complex scenarios while maintaining efficiency, we propose an adaptive group-wise correlation similarity measure for cost volume construction. Besides, we present a pyramid-based feature extraction network with a repeated top-down and bottom-up structure to gather more context-aware information, which can better meet the challenges in ill-posed regions. This novel feature extractor is also utilized to construct an enhanced Gauss-Newton refinement module for further upsampling and optimizing. Extensive experiments on the DTU, the Tanks & Temples and the BlendedMVS datasets demonstrate the effectiveness and generalization of our approach, which can achieve better or competitive results compared to state-of-the-art methods. The code will be available at https://github.com/BoyangSONG/ACINR-MVSNet. [ABSTRACT FROM AUTHOR]

For acoustic event recognition (AER), it is important to extract the semantic feature that jointly considers the content information and the temporal ordering. Thus, our previous work proposed a method to obtain this type of semantic feature by independently learning a related subspace for each acoustic event. However, in this method, since different subspaces are employed to extract the semantic features of various events, the subspace consistency cannot be guaranteed, and this makes the differences between various semantic features come from not only the semantic characteristic but also the unnecessary subspace inconsistency. To solve the problem, we propose a common subspace learning (CSL) based method for extracting the semantic features of various events in this paper. To obtain the above subspace, the CSL minimizes an objective that can jointly consider the content information and the temporal orderings of various events in the subspace, and provides an efficient algorithm to get its optimization solution. Furthermore, in the obtained common subspace, the semantic feature of each event is represented by a corresponding projection matrix. Since various projection matrices are the mappings onto this shared subspace, the subspace consistency is ensured. To evaluate the performance of our method, experiments are conducted on the AudioEvent, the ESC-50, and the ESC-10 databases, and the results indicate that the CSL is better than our previous and the related state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Ensemble learning (EL) method which has high potential to improve the performance of single image classification model can be constructed in two steps: one is the generation of weak learners; the other is the combination of these learners. In this paper, an ensemble learning method based on deep neural network and group decision making (DNN-GDM-EL) is proposed, which uses deep neural networks (DNNs) to generate individual learners and exploits group decision making (GDM) to combine these learners. DNNs have demonstrated remarkable ability for image classification due to the powerful feature extraction ability. To ensure the diversity and accuracy, many different DNNs are used to generate individual learners. Furthermore, the individual learners are regarded as decision-makers (DMs), the categories are seen as alternatives, and the GDM aims to find an optimal alternative considering various suggestions of DMs. Specifically, a GDM model is established based on Bayesian theory which can reflect the complex relationship among the class of image, prior knowledge and output of DNN, and a GDM method based on TOPSIS is applied to solve this problem. Next, the index matrix consisted of DM's attributes is proposed, and an aggregation method based on 2-additive generalized Shapley AIVIFCA (2AGSAIVIFCA) operator is used to calculate the weights of DMs by fusing these matrixes. Further, state transition algorithm (STA) is applied to obtain the optimal weights of alternative's attributes. The effectiveness and superiority are verified in three public data sets and a real industrial problem by comparing DNN-GDM-EL method with other typical EL methods. [ABSTRACT FROM AUTHOR]

Fault detection is a fundamental and important research topic in automatic seismic interpretation since the geometry of faults usually reveals the accumulation and migration of geological resource. In the communities of machine learning, many methods convert the fault detection task into a patch-based binary classification problem to solve. Because there are only a few faults in seismic data, the related classification task is obviously imbalanced, i.e., non-fault samples heavily outnumber fault samples. On considering that the existing CNN models just transform the problem into a balanced two-class task via undersampling, in this paper we attempt to construct CNNs directly from the imbalanced patch data with the assistance of transfer learning. In particular, we use focal loss to robustly learn fault feature representations for both fault and non-fault patches. With the aim to make full use of available data, we adopt transfer learning to enhance the recognition ability of relatively complex patches. The experiments conducted with both synthetic and real seismic data (some slices of the Netherland offshore F3 block in the North Sea) show that the novel method performs very well in detecting faults in seismic data. In supplemental material, the features extracted by the CNNs are visualized for both fault and non-fault patches to get some insights for their good performance. • We apply CNN with the focal loss to detect seismic faults with imbalanced patches. • Transfer learning strategy is used to fine-tune the CNN to classify complex patches. • We consider more types of faults in generated data to improve the CNN' robustness. • The CNN' predicted results are post-processed by some methods to extract fault lines. [ABSTRACT FROM AUTHOR]

Most of the infrared and visible image fusion methods need to decompose the source image into two parts during the fusion process, and this will lead to deficient extraction of details and salient targets. In order to solve this problem, we first analyze VGG-19 and reserve five convolutional layers we need under the guidance of transfer learning. Then the fusion method proposed in this paper directly inputs the source image into the five layers for feature extraction. After that, we can obtain the activity level maps by L1-norm and average operator. On this basis, softmax function and up-sampling operator are used to obtain the weight maps. Then the final weight maps are convolved with infrared and visible images respectively to get five candidate fused images. Finally, we choose the maximum value from the five candidates for each position as the reconstruction of the final fused image. Experimental results show that the proposed method has better visual quality and less artifact and noise. It is also overwhelming in objective evaluation than some traditional or popular fusion methods. [ABSTRACT FROM AUTHOR]

As an essential subproblem in the face recognition field, occluded face recognition has received considerable attention in recent years. However, satisfactory recognition accuracy for occluded faces has not been achieved. Therefore, in order to solve this problem, this paper proposes a convolutional neural network that has multidimensional serial feature extraction modules for occluded faces and uses the deep learning method to improve the recognition rate. In order to improve the expression ability of the network, the method extracts features from two dimensions — space and channel — and built a multidimensional serial feature extraction module. By "multiscale" and "dependence" processing features, the generalization ability was significantly improved. Through the experimental test, the recognition accuracy of the multidimensional feature network (MFNet) reached 90.35%. This accuracy is 17.89% (72.36%) better than that of the traditional algorithm. Meanwhile, compared with other convolutional neural networks, this network has different degrees of improvement, which are 0.68% (ArcFace), 2.14% (Visual Geometry Group Face), and 4.83% (DeepID3). [ABSTRACT FROM AUTHOR]

Due to the actual operation of the wind turbine, the collected fault data sets are limited, and it is difficult to realize fault diagnosis through the correlation between variables. To solve this problem, a fault diagnosis method based on virtual expansion and spherical mapping model is proposed in this paper. Firstly, Hermite interpolation is applied to discrete wind power data samples to obtain an interpolation curve about the characteristics of the sample, and a synchronous sampling method is adopted for the interpolation curve to construct a virtual sample. Then, the features of the virtual sample are mapped to a three-dimensional space. Define the spherical data model and perform spherical fitting in a three-dimensional coordinate system. Finally, feature extraction is performed on the fitted spherical surface for training and testing extreme learning machine (ELM). The distribution law of fault data in the spherical model is summarized. Using the data generated based on the Bootstrap method as a control group, comparative experiments were carried out in back-propagation neural network (BP), Probabilistic Neural Network (PNN), General Regression Neural Network (GRNN), and Support Vector Machine (SVM), which verified the effectiveness of the proposed method. [ABSTRACT FROM AUTHOR]

The damage assessment of Carbon Fiber Reinforced Plastics (CFRP) laminate structure is very important to ensure the safety of the structure in service. Machine learning is a research hotspot in damage assessment of CFRP structures. However, there are problems such as noise interference, small number of training samples, and few damage types of training samples. In order to solve the above problems, the damage assessment method of CFRP laminate based on Synchrosqueezed Wavelet Transform and ensemble Principal Component Analysis was proposed. First, the damage feature extraction method based on Synchrosqueezed Wavelet Transform (SWT) and Singular Spectrum Analysis (SSA) was proposed to reduce the feature dimension and achieve effective extraction of damage features. Then, the damage assessment method based on ensemble Principal Component Analysis (PCA) was proposed to realize the damage assessment of CFRP structure. Finally, the simulation experiment was designed to verify the damage assessment results in the case of noise interference, small number of training samples, and few damage types of training samples. The results showed that the damage assessment method of CFRP laminate structure based on SWT, SSA and ensemble PCA was feasible. This paper provided a feasible method for evaluating the damage degree of CFRP laminate structure. [ABSTRACT FROM AUTHOR]

Band selection is a research hotspot in hyperspectral image processing. The continuity of the spectral bands causes the adjacent bands to be highly correlated, and correlation among long-range bands is possible with hundreds of spectral bands. Most existing deep learning methods fail to make full use of the inter-band correlation for band selection. In this paper, a novel dual-graph convolutional network based on band attention and a sparse constraint is proposed for band selection. The network consists of two branches. In the attention branch, band-based dual graphs are constructed to encode the contextual correlation of adjacent bands and the structural correlation of long-range bands into non-Euclidean space. Subsequently, the graph convolution-based band attention mechanism is devised to aggregate the band information in the band-based dual graphs and to generate the attention map for all bands. The band attention map is sparsely constrained and embedded as a mask into the trunk branch. In the trunk branch, sample-based dual graphs are constructed to represent the topological information of the samples in the spectral and spatial domains. Furthermore, a dense graph convolutional network is designed to extract and fuse the spatial–spectral and topological features from the shallow to deep layers for classification. A soft-shifting optimization strategy is implemented by defining a new loss from full bands and selected bands to solve the optimization problem caused by the sparse constraint. In this manner, band selection, feature extraction, and classification can be combined into an end-to-end trainable network. The experimental results on representative hyperspectral image datasets demonstrate the superiority of the proposed method over current state-of-the-art band selection methods. • An improved graph convolutional network is proposed for band selection. • Band-based and sample-based graphs are built with spatial and spectral information. • Design soft-shifting optimization strategy to overcome the optimization problem. • Mini batches of labeled and unlabeled samples are used for graph construction. [ABSTRACT FROM AUTHOR]

Semantic segmentation has achieved great success with the popularity of convolutional neural networks (CNNs). However, the huge computational burden restricts the application of most existing networks on edge devices with strict inference time constraints. To solve this problem, a weighted factorized-depthwise convolution network (WFDCNet) is presented in this paper, which contains full- dimensional continuous separation convolution (FCS) modules and a lateral asymmetric pyramid fusion (LAPF) module, aiming to obtain high accuracy without damaging inference speed. Specifically, the FCS module enables the calculation of each dimension to be completed independently in a continuous separation process and uses simplified SE (SSE) attention layer to adjust the channels, achieving the extensive extraction of feature information. The LAPF module is able to eliminate semantic divergence and fuse feature maps of three different scales to realize the combination of multiple information from the front-end and the back-end network. WFDCNet shows superior performance on Cityscapes, Camvid, Mapillary Vistas and COCO-Stuff datasets. Especially, the experimental results demonstrate that our network achieves 73.7% mIoU on Cityscapes dataset, with the inference speed of 102.6FPS on a single RTX 2080 Ti GPU, and 17.2FPS on Jetson TX2. • Separate calculation in all dimensions and use simplified attention layer. • Present the full-dimensional continuous separation convolution module. • Propose the lateral asymmetric pyramid fusion module. • Design an asymmetric encoder-decoder network for real-time semantic segmentation. [ABSTRACT FROM AUTHOR]