Search

The research groups in computer vision, graphics, and machine learning have dedicated a substantial amount of attention to the areas of 3D object reconstruction, augmentation, and registration. Deep learning is the predominant method used in artificial intelligence for addressing computer vision challenges. However, deep learning on three-dimensional data presents distinct obstacles and is now in its nascent phase. There have been significant advancements in deep learning specifically for three-dimensional data, offering a range of ways to address these issues. This study offers a comprehensive examination of the latest advancements in deep learning methodologies. We examine many benchmark models for the tasks of 3D object registration, augmentation, and reconstruction. We thoroughly analyse their architectures, advantages, and constraints. In summary, this report provides a comprehensive overview of recent advancements in three-dimensional deep learning and highlights unresolved research areas that will need to be addressed in the future. [ABSTRACT FROM AUTHOR]

Unsupervised domain adaptation (UDA) aims to reapply the classifier to be ever-trained on a labeled source domain to a related unlabeled target domain. Recent progress in this line has evolved with the advance of network architectures from convolutional neural networks (CNNs) to transformers or both hybrids. However, this advance has to pay the cost of high computational overheads or complex training processes. In this paper, we propose an efficient alternative hybrid architecture by marrying transformer to contextual convolution (TransConv) to solve UDA tasks. Different from previous transformer based UDA architectures, TransConv has two special aspects: (1) reviving the multilayer perception (MLP) of transformer encoders with Gaussian channel attention fusion for robustness, and (2) mixing contextual features to highly efficient dynamic convolutions for cross-domain interaction. As a result, TransConv enables to calibrate interdomain feature semantics from the global features and the local ones. Experimental results on five benchmarks show that TransConv attains remarkable results with high efficiency as compared to the existing UDA methods. [ABSTRACT FROM AUTHOR]

Block compressed sensing (BCS) is a promising method for resource-constrained image/video coding applications. However, the quantization of BCS measurements has posed a challenge, leading to significant quantization errors and encoding redundancy. In this paper, we propose a quantization method for BCS measurements using convolutional neural networks (CNN). The quantization process maps measurements to quantized data that follow a uniform distribution based on the measurements' distribution, which aims to maximize the amount of information carried by the quantized data. The dequantization process restores the quantized data to data that conform to the measurements' distribution. The restored data are then modified by the correlation information of the measurements drawn from the quantized data, with the goal of minimizing the quantization errors. The proposed method uses CNNs to construct quantization and dequantization processes, and the networks are trained jointly. The distribution parameters of each block are used as side information, which is quantized with 1 bit by the same method. Extensive experiments on four public datasets showed that, compared with uniform quantization and entropy coding, the proposed method can improve the PSNR by an average of 0.48 dB without using entropy coding when the compression bit rate is 0.1 bpp. [ABSTRACT FROM AUTHOR]

The applications of deep learning and artificial intelligence have permeated daily life, with time series prediction emerging as a focal area of research due to its significance in data analysis. The evolution of deep learning methods for time series prediction has progressed from the Convolutional Neural Network (CNN) and the Recurrent Neural Network (RNN) to the recently popularized Transformer network. However, each of these methods has encountered specific issues. Recent studies have questioned the effectiveness of the self-attention mechanism in Transformers for time series prediction, prompting a reevaluation of approaches to LTSF (Long Time Series Forecasting) problems. To circumvent the limitations present in current models, this paper introduces a novel hybrid network, Temporal Convolutional Network-Linear (TCN-Linear), which leverages the temporal prediction capabilities of the Temporal Convolutional Network (TCN) to enhance the capacity of LSTF-Linear. Time series from three classical chaotic systems (Lorenz, Mackey–Glass, and Rossler) and real-world stock data serve as experimental datasets. Numerical simulation results indicate that, compared to classical networks and novel hybrid models, our model achieves the lowest RMSE, MAE, and MSE with the fewest training parameters, and its R2 value is the closest to 1. [ABSTRACT FROM AUTHOR]

Quantum computing (QC) has opened the door to advancements in machine learning (ML) tasks that are currently implemented in the classical domain. Convolutional neural networks (CNNs) are classical ML architectures that exploit data locality and possess a simpler structure than a fully connected multi-layer perceptrons (MLPs) without compromising the accuracy of classification. However, the concept of preserving data locality is usually overlooked in the existing quantum counterparts of CNNs, particularly for extracting multifeatures in multidimensional data. In this paper, we present an multidimensional quantum convolutional classifier (MQCC) that performs multidimensional and multifeature quantum convolution with average and Euclidean pooling, thus adapting the CNN structure to a variational quantum algorithm (VQA). The experimental work was conducted using multidimensional data to validate the correctness and demonstrate the scalability of the proposed method utilizing both noisy and noise-free quantum simulations. We evaluated the MQCC model with reference to reported work on state-of-the-art quantum simulators from IBM Quantum and Xanadu using a variety of standard ML datasets. The experimental results show the favorable characteristics of our proposed techniques compared with existing work with respect to a number of quantitative metrics, such as the number of training parameters, cross-entropy loss, classification accuracy, circuit depth, and quantum gate count. [ABSTRACT FROM AUTHOR]

Learning in neural networks with locally-tuned neuron models such as radial Basis Function (RBF) networks is often seen as instable, in particular when multi-layered architectures are used. Furthermore, universal approximation theorems for single-layered RBF networks are very well established; therefore, deeper architectures are theoretically not required. Consequently, RBFs are mostly used in a single-layered manner. However, deep neural networks have proven their effectiveness on many different tasks. In this paper, we show that deeper RBF architectures with multiple radial basis function layers can be designed together with efficient learning schemes. We introduce an initialization scheme for deep RBF networks based on k-means clustering and covariance estimation. We further show how to make use of convolutions to speed up the calculation of the Mahalanobis distance in a partially connected way, which is similar to the convolutional neural networks (CNNs). Finally, we evaluate our approach on image classification as well as speech emotion recognition tasks. Our results show that deep RBF networks perform very well, with comparable results to other deep neural network types, such as CNNs. [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]

The fifth-generation (5G) mobile cellular network is vulnerable to various security threats. Radio frequency fingerprint (RFF) identification is an emerging physical layer authentication technique which can be used to detect spoofing and distributed denial of service attacks. In this paper, the performance of RFF identification is studied for 5G mobile phones. The differential constellation trace figure (DCTF) is extracted from the physical random access channel (PRACH) preamble. When the database of all 64 PRACH preambles is available at the gNodeB (gNB), an index-based DCTF identification scheme is proposed, and the classification accuracy reaches 92.78% with a signal-to-noise ratio of 25 dB. Moreover, due to the randomness in the selection of preamble sequences in the random access procedure, when only a portion of the preamble sequences can be trained, a group-based DCTF identification scheme is proposed. The preamble sequences generated from the same root value are grouped together, and the untrained sequences can be identified based on the trained sequences within the same group. The classification accuracy of the group-based scheme is 89.59%. An experimental system has been set up using six 5G mobile phones of three models. The 5G gNB is implemented on the OpenAirInterface platform. [ABSTRACT FROM AUTHOR]

In this investigation, a synthesis of Convolutional Neural Networks (CNNs) and Bayesian inference is presented, leading to a novel approach to the problem of Multiple Hypothesis Testing (MHT). Diverging from traditional paradigms, this study introduces a sequence-based uncalibrated Bayes factor approach to test many hypotheses using the same family of sampling parametric models. A two-step methodology is employed: initially, a learning phase is conducted utilizing simulated datasets encompassing a wide spectrum of null and alternative hypotheses, followed by a transfer phase applying this fitted model to real-world experimental sequences. The outcome is a CNN model capable of navigating the complex domain of MHT with improved precision over traditional methods, also demonstrating robustness under varying conditions, including the number of true nulls and dependencies between tests. Although indications of empirical evaluations are presented and show that the methodology will prove useful, more work is required to provide a full evaluation from a theoretical perspective. The potential of this innovative approach is further illustrated within the critical domain of genomics. Although formal proof of the consistency of the model remains elusive due to the inherent complexity of the algorithms, this paper also provides some theoretical insights and advocates for continued exploration of this methodology. [ABSTRACT FROM AUTHOR]

Automatic image description, also known as image captioning, aims to describe the elements included in an image and their relationships. This task involves two research fields: computer vision and natural language processing; thus, it has received much attention in computer science. In this review paper, we follow the Kitchenham review methodology to present the most relevant approaches to image description methodologies based on deep learning. We focused on works using convolutional neural networks (CNN) to extract the characteristics of images and recurrent neural networks (RNN) for automatic sentence generation. As a result, 53 research articles using the encoder-decoder approach were selected, focusing only on supervised learning. The main contributions of this systematic review are: (i) to describe the most relevant image description papers implementing an encoder-decoder approach from 2014 to 2022 and (ii) to determine the main architectures, datasets, and metrics that have been applied to image description. [ABSTRACT FROM AUTHOR]

In this paper, we investigate the problem of graph neural network quantization. Despite the great success on convolutional neural networks, directly applying current network quantization approaches to graph neural networks faces two challenges. First, the fixed-scale parameter in the current methods cannot flexibly fit diverse tasks and network architectures. Second, the variations of node degree in a graph leads to uneven responses, limiting the accuracy of the quantizer. To address these two challenges, we introduce learnable scale parameters that can be optimized jointly with the graph networks. In addition, we propose degree-aware normalization to process nodes with different degrees. Experiments on different tasks, baselines, and datasets demonstrate the superiority of our method against previous state-of-the-art ones. [ABSTRACT FROM AUTHOR]

As one of the most critical tasks in legal artificial intelligence, legal judgment prediction (LJP) has garnered growing attention, especially in the civil law system. However, current methods often overlook the challenge of imbalanced label distributions, treating each label with equal importance, which can lead the model to be biased toward labels with high frequency. In this paper, we propose a label-enhanced prototypical network (LPN) suitable for LJP, that adopts a strategy of uniform encoding and separate decoding. Specifically, LPN adopts a multi-scale convolutional neural network to uniformly encode case factual description to capture long-distance features of the document. At the decoding end, a prototypical network incorporating label semantic features is used to guide the learning of prototype representations of high-frequency and low-frequency labels, respectively. At the same time, we also propose a prototype-prototype loss to optimize the prototypical representation. We conduct extensive experiments on two real datasets and show that our proposed method effectively improves the performance of LJP, with an average F1 of 1.23% and 1.13% higher than the state-of-the-art model on two subtasks, respectively. [ABSTRACT FROM AUTHOR]

Deep convolution neural networks have proven their powerful ability in comparing many tasks of computer vision due to their strong data learning capacity. In this paper, we propose a novel end-to-end denoising network, termed Fourier embedded U-shaped network (FEUSNet). By analyzing the amplitude spectrum and phase spectrum of Fourier coefficients, we find that low-frequency features of an image are in the former while noise features are in the latter. To make full use of this characteristic, Fourier features are learned and are concatenated as a prior module that is embedded into a U-shaped network to reduce noise while preserving multi-scale fine details. In the experiments, we first present ablation studies on the Fourier coefficients' learning networks and loss function. Then, we compare the proposed FEUSNet with the state-of-the-art denoising methods in quantization and qualification. The experimental results show that our FEUSNet performs well in noise suppression and preserves multi-scale enjoyable structures, even outperforming advanced denoising approaches. [ABSTRACT FROM AUTHOR]

Contact fatigue is one of the most common failure forms of typical basic components such as bearings and gears. Accurate prediction of contact fatigue performance degradation trends in components is conducive to the scientific formulation of maintenance strategies and health management of equipment, which is of great significance for industrial production. In this paper, to realize the performance degradation trend prediction accurately, a prediction method based on multi-domain features and temporal convolutional networks (TCNs) was proposed. Firstly, a multi-domain and high-dimensional feature set of vibration signals was constructed, and performance degradation indexes with good sensitivity and strong trends were initially screened using comprehensive evaluation indexes. Secondly, the kernel principal component analysis (KPCA) method was used to eliminate redundant information among multi-domain features and construct health indexes (HIs) based on a convolutional autoencoder (CAE) network. Then, the performance degradation trend prediction model based on TCN was constructed, and the degradation trend prediction for the monitored object was realized using direct multi-step prediction. On this basis, the effectiveness of the proposed method was verified using a bearing common-use data set, and it was successfully applied to performance degradation trend prediction for rolling contact fatigue specimens. The results show that using KPCA can reduce the feature set from 14 dimensions to 4 dimensions and retain 98.33% of the information in the original preferred feature set. The method of constructing the HI based on CAE is effective, and change processes versus time of the constructed HI can truly reflect the degradation process of rolling contact fatigue specimen performance; this method has obvious advantages over the two commonly used methods for constructing HIs including auto-encoding (AE) networks and gaussian mixture models (GMMs). The model based on TCN can accurately predict the performance degradation of rolling contact fatigue specimens. Compared with prediction models based on long short-term memory (LSTM) networks and gating recurrent units (GRUs), the model based on TCN has better performance and higher prediction accuracy. The RMS error and average absolute error for a prediction step of 3 are 0.0146 and 0.0105, respectively. Overall, the proposed method has universal significance and can be applied to predict the performance degradation trend of other mechanical equipment/parts. [ABSTRACT FROM AUTHOR]

In recent years, the number of traffic accidents caused by road defects has increased dramatically all over the world, and the repair and prevention of road defects is an urgent task. Researchers in different countries have proposed many models to deal with this task, but most of them are either highly accurate and slow in detection, or the accuracy is low and the detection speed is high. The accuracy and speed have achieved good results, but the generalization of the model to other datasets is poor. Given this, this paper takes YOLOv5s as a benchmark model and proposes an optimization model to solve the problem of road defect detection. First, we significantly reduce the parameters of the model by pruning the model and removing unimportant modules, propose an improved Spatial Pyramid Pooling-Fast (SPPF) module to improve the feature signature fusion ability, and finally add an attention module to focus on the key information. The activation function, sampling method, and other strategies were also replaced in this study. The test results on the Global Road Damage Detection Challenge (GRDDC) dataset show that the FPS of our proposed model is not only faster than the baseline model but also improves the MAP by 2.08%, and the size of this model is also reduced by 6.07 M. [ABSTRACT FROM AUTHOR]

An effective fault diagnosis method of bearing is the key to predictive maintenance of modern industrial equipment. With the single use of equipment failure mechanism or operation of data, it is hard to resolve multiple complex variable working conditions, multiple types of fault and equipment malfunctions and failures related to knowledge and data. In order to solve these problems, a fault diagnosis method based on the fusion of deep learning with a knowledge graph is proposed in this paper. Firstly, the knowledge rules of bearing data is used for entity extraction. Next, the multiscale optimized convolutional neural network (MOCNN) proposed in this paper is used for fault classification to achieve relationship extraction. Finally, the fault diagnosis graph of the bearing is constructed for fault-assisted decision-making as well as the detailed display of fault information. According to experiment analysis, the fault diagnosis model based on MOCNN proposed in this paper, which integrates the end-to-end convolutional neural network and the attention mechanism, still achieves an accuracy of 97.86% under the data set of 160 types of faults. Compared with the deep learning models such as Resnet and Inception in the noise environment of multiple working conditions and variable working conditions, the model proposed in this paper not only shows a faster convergence speed and stable performance, but also a higher accuracy in evaluation indicators, which is beneficial to practical use. [ABSTRACT FROM AUTHOR]

In this paper, to improve the slow processing speed of the rule-based visible and NIR (near-infrared) image synthesis method, we present a fast image fusion method using DenseFuse, one of the CNN (convolutional neural network)-based image synthesis methods. The proposed method applies a raster scan algorithm to secure visible and NIR datasets for effective learning and presents a dataset classification method using luminance and variance. Additionally, in this paper, a method for synthesizing a feature map in a fusion layer is presented and compared with the method for synthesizing a feature map in other fusion layers. The proposed method learns the superior image quality of the rule-based image synthesis method and shows a clear synthesized image with better visibility than other existing learning-based image synthesis methods. Compared with the rule-based image synthesis method used as the target image, the proposed method has an advantage in processing speed by reducing the processing time to three times or more. [ABSTRACT FROM AUTHOR]

In image classification with Deep Convolutional Neural Networks (DCNNs), the number of parameters in pointwise convolutions rapidly grows due to the multiplication of the number of filters by the number of input channels that come from the previous layer. Existing studies demonstrated that a subnetwork can replace pointwise convolutional layers with significantly fewer parameters and fewer floating-point computations, while maintaining the learning capacity. In this paper, we propose an improved scheme for reducing the complexity of pointwise convolutions in DCNNs for image classification based on interleaved grouped filters without divisibility constraints. The proposed scheme utilizes grouped pointwise convolutions, in which each group processes a fraction of the input channels. It requires a number of channels per group as a hyperparameter C h . The subnetwork of the proposed scheme contains two consecutive convolutional layers K and L, connected by an interleaving layer in the middle, and summed at the end. The number of groups of filters and filters per group for layers K and L is determined by exact divisions of the original number of input channels and filters by C h . If the divisions were not exact, the original layer could not be substituted. In this paper, we refine the previous algorithm so that input channels are replicated and groups can have different numbers of filters to cope with non exact divisibility situations. Thus, the proposed scheme further reduces the number of floating-point computations (11%) and trainable parameters (10%) achieved by the previous method. We tested our optimization on an EfficientNet-B0 as a baseline architecture and made classification tests on the CIFAR-10, Colorectal Cancer Histology, and Malaria datasets. For each dataset, our optimization achieves a saving of 76%, 89%, and 91% of the number of trainable parameters of EfficientNet-B0, while keeping its test classification accuracy. [ABSTRACT FROM AUTHOR]

Algorithms for converting 2D to 3D are gaining importance following the hiatus brought about by the discontinuation of 3D TV production; this is due to the high availability and popularity of virtual reality systems that use stereo vision. In this paper, several depth image-based rendering (DIBR) approaches using state-of-the-art single-frame depth generation neural networks and inpaint algorithms are proposed and validated, including a novel very fast inpaint (FAST). FAST significantly exceeds the speed of currently used inpaint algorithms by reducing computational complexity, without degrading the quality of the resulting image. The role of the inpaint algorithm is to fill in missing pixels in the stereo pair estimated by DIBR. Missing estimated pixels appear at the boundaries of areas that differ significantly in their estimated distance from the observer. In addition, we propose parameterizing DIBR using a singular, easy-to-interpret adaptable parameter that can be adjusted online according to the preferences of the user who views the visualization. This single parameter governs both the camera parameters and the maximum binocular disparity. The proposed solutions are also compared with a fully automatic 2D to 3D mapping solution. The algorithm proposed in this work, which features intuitive disparity steering, the foundational deep neural network MiDaS, and the FAST inpaint algorithm, received considerable acclaim from evaluators. The mean absolute error of the proposed solution does not contain statistically significant differences from state-of-the-art approaches like Deep3D and other DIBR-based approaches using different inpaint functions. Since both the source codes and the generated videos are available for download, all experiments can be reproduced, and one can apply our algorithm to any selected video or single image to convert it. [ABSTRACT FROM AUTHOR]

To address the diverse needs of enterprise users and the cold-start issue of recommendation system, this paper proposes a quality-service demand classification method—1D-CNN-CrossEntorpyLoss, based on cross-entropy loss and one-dimensional convolutional neural network (1D-CNN) with the comprehensive enterprise quality portrait labels. The main idea of 1D-CNN-CrossEntorpyLoss is to use cross-entropy to minimize the loss of 1D-CNN model and enhance the performance of the enterprise quality-service demand classification. The transaction data of the enterprise quality-service platform are selected as the data source. Finally, the performance of 1D-CNN-CrossEntorpyLoss is compared with XGBoost, SVM, and logistic regression models. From the experimental results, it can be found that 1D-CNN-CrossEntorpyLoss has the best classification results with an accuracy of 72.44%. In addition, compared to the results without the enterprise-quality portrait, the enterprise-quality portrait improves the accuracy and recall of 1D-CNN-CrossEntorpyLoss model. It is also verified that the enterprise-quality portrait can further improve the classification ability of enterprise quality-service demand, and 1D-CNN-CrossEntorpyLoss is better than other classification methods, which can improve the precision service of the comprehensive quality service platform for MSMEs. [ABSTRACT FROM AUTHOR]

In this paper, we propose a novel method for point cloud complementation called PADPNet. Our approach uses a combination of global and local information to infer missing elements in the point cloud. We achieve this by dividing the input point cloud into uniform local regions, called perceptual fields, which are abstractly understood as special convolution kernels. The set of point clouds in each local region is represented as a feature vector and transformed into N uniform perceptual fields as the input to our transformer model. We also designed a geometric density-aware block to better exploit the inductive bias of the point cloud's 3D geometric structure. Our method preserves sharp edges and detailed structures that are often lost in voxel-based or point-based approaches. Experimental results demonstrate that our approach outperforms other methods in reducing the ambiguity of output results. Our proposed method has important applications in 3D computer vision and can efficiently recover complete 3D object shapes from missing point clouds. [ABSTRACT FROM AUTHOR]

This paper proposes an intelligent diagnosis method for rotating machinery faults based on improved variational mode decomposition (IVMD) and CNN to process the rotating machinery non-stationary signal. Firstly, to solve the problem of time-domain feature extraction for fault diagnosis, this paper proposes an improved variational mode decomposition method with automatic optimization of the number of modes. This method overcomes the problems of the traditional VMD method, in that each parameter is set by experience and is greatly influenced by subjective experience. Secondly, the decomposed signal components are analyzed by correlation, and then high correlated components with the original signal are selected to reconstruct the original signal. The continuous wavelet transform (CWT) is employed to extract the two-dimensional time–frequency domain feature map of the fault signal. Finally, the deep learning method is used to construct a convolutional neural network. After feature extraction, the two-dimensional time-frequency image is applied to the neural network to identify fault features. Experiments verify that the proposed method can adapt to rotating machinery faults in complex environments and has a high recognition rate. [ABSTRACT FROM AUTHOR]

AMC (automatic modulation classification) plays a vital role in spectrum monitoring and electromagnetic abnormal signal detection. Up to now, few studies have focused on the complementarity between features of different modalities and the importance of the feature fusion mechanism in the AMC method. This paper proposes a dual-modal feature fusion convolutional neural network (DMFF-CNN) for AMC to use the complementarity between different modal features fully. DMFF-CNN uses the gram angular field (GAF) image coding and intelligence quotient (IQ) data combined with CNN. Firstly, the original signal is converted into images by GAF, and the GAF images are used as the input of ResNet50. Secondly, it is converted into IQ data and as the complex value network (CV-CNN) input to extract features. Furthermore, a dual-modal feature fusion mechanism (DMFF) is proposed to fuse the dual-modal features extracted by GAF-ResNet50 and CV-CNN. The fusion feature is used as the input of DMFF-CNN for model training to achieve AMC of multi-type signals. In the evaluation stage, the advantages of the DMFF mechanism proposed in this paper and the accuracy improvement compared with other feature fusion algorithms are discussed. The experiment shows that our method performs better than others, including some state-of-the-art methods, and has superior robustness at a low signal-to-noise ratio (SNR), and the average classification accuracy of the dataset signals reaches 92.1%. The DMFF-CNN proposed in this paper provides a new path for the AMC field. [ABSTRACT FROM AUTHOR]

Offline hand-drawn diagram recognition is concerned with digitizing diagrams sketched on paper or whiteboard to enable further editing. Some existing models can identify the individual objects like arrows and symbols, but they become involved in the dilemma of being unable to understand a diagram's structure. Such a shortage may be inconvenient to digitalization or reconstruction of a diagram from its hand-drawn version. Other methods can accomplish this goal, but they live on stroke temporary information and time-consuming post-processing, which somehow hinders the practicability of these methods. Recently, Convolutional Neural Networks (CNN) have been proved that they perform the state-of-the-art across many visual tasks. In this paper, we propose DrawnNet, a unified CNN-based keypoint-based detector, for recognizing individual symbols and understanding the structure of offline hand-drawn diagrams. DrawnNet is designed upon CornerNet with extensions of two novel keypoint pooling modules which serve to extract and aggregate geometric characteristics existing in polygonal contours such as rectangle, square, and diamond within hand-drawn diagrams, and an arrow orientation prediction branch which aims to predict which direction an arrow points to through predicting arrow keypoints. We conducted wide experiments on public diagram benchmarks to evaluate our proposed method. Results show that DrawnNet achieves 2.4 % , 2.3 % , and 1.7 % recognition rate improvements compared with the state-of-the-art methods across benchmarks of FC-A, FC-B, and FA, respectively, outperforming existing diagram recognition systems on each metric. Ablation study reveals that our proposed method can effectively enable hand-drawn diagram recognition. [ABSTRACT FROM AUTHOR]

The remaining useful life (RUL) prediction of rolling bearings based on vibration signals has attracted widespread attention. It is not satisfactory to adopt information theory (such as information entropy) to realize RUL prediction for complex vibration signals. Recent research has used more deep learning methods based on the automatic extraction of feature information to replace traditional methods (such as information theory or signal processing) to obtain higher prediction accuracy. Convolutional neural networks (CNNs) based on multi-scale information extraction have demonstrated promising effectiveness. However, the existing multi-scale methods significantly increase the number of model parameters and lack efficient learning mechanisms to distinguish the importance of different scale information. To deal with the issue, the authors of this paper developed a novel feature reuse multi-scale attention residual network (FRMARNet) for the RUL prediction of rolling bearings. Firstly, a cross-channel maximum pooling layer was designed to automatically select the more important information. Secondly, a lightweight feature reuse multi-scale attention unit was developed to extract the multi-scale degradation information in the vibration signals and recalibrate the multi-scale information. Then, end-to-end mapping between the vibration signal and the RUL was established. Finally, extensive experiments were used to demonstrate that the proposed FRMARNet model can improve prediction accuracy while reducing the number of model parameters, and it outperformed other state-of-the-art methods. [ABSTRACT FROM AUTHOR]

At present, the fault diagnosis methods for rolling bearings are all based on research with fewer fault categories, without considering the problem of multiple faults. In practical applications, the coexistence of multiple operating conditions and faults can lead to an increase in classification difficulty and a decrease in diagnostic accuracy. To solve this problem, a fault diagnosis method based on an improved convolution neural network is proposed. The convolution neural network adopts a simple structure of three-layer convolution. The average pooling layer is used to replace the common maximum pooling layer, and the global average pooling layer is used to replace the full connection layer. The BN layer is used to optimize the model. The collected multi-class signals are used as the input of the model, and the improved convolution neural network is used for fault identification and classification of the input signals. The experimental data of XJTU-SY and Paderborn University show that the method proposed in this paper has a good effect on the multi-classification of bearing faults. [ABSTRACT FROM AUTHOR]

The safe and comfortable operation of high-speed trains has attracted extensive attention. With the operation of the train, the performance of high-speed train bogie components inevitably degrades and eventually leads to failures. At present, it is a common method to achieve performance degradation estimation of bogie components by processing high-speed train vibration signals and analyzing the information contained in the signals. In the face of complex signals, the usage of information theory, such as information entropy, to achieve performance degradation estimations is not satisfactory, and recent studies have more often used deep learning methods instead of traditional methods, such as information theory or signal processing, to obtain higher estimation accuracy. However, current research is more focused on the estimation for a certain component of the bogie and does not consider the bogie as a whole system to accomplish the performance degradation estimation task for several key components at the same time. In this paper, based on soft parameter sharing multi-task deep learning, a multi-task and multi-scale convolutional neural network is proposed to realize performance degradation state estimations of key components of a high-speed train bogie. Firstly, the structure takes into account the multi-scale characteristics of high-speed train vibration signals and uses a multi-scale convolution structure to better extract the key features of the signal. Secondly, considering that the vibration signal of high-speed trains contains the information of all components, the soft parameter sharing method is adopted to realize feature sharing in the depth structure and improve the utilization of information. The effectiveness and superiority of the structure proposed by the experiment is a feasible scheme for improving the performance degradation estimation of a high-speed train bogie. [ABSTRACT FROM AUTHOR]

Source acquisition device identification from recorded audio aims to identify the source recording device by analyzing the intrinsic characteristics of audio, which is a challenging problem in audio forensics. In this paper, we propose a spatiotemporal representation learning framework with multi-attention mechanisms to tackle this problem. In the deep feature extraction stage of recording devices, a two-branch network based on residual dense temporal convolution networks (RD-TCNs) and convolutional neural networks (CNNs) is constructed. The spatial probability distribution features of audio signals are employed as inputs to the branch of the CNN for spatial representation learning, and the temporal spectral features of audio signals are fed into the branch of the RD-TCN network for temporal representation learning. This achieves simultaneous learning of long-term and short-term features to obtain an accurate representation of device-related information. In the spatiotemporal feature fusion stage, three attention mechanisms—temporal, spatial, and branch attention mechanisms—are designed to capture spatiotemporal weights and achieve effective deep feature fusion. The proposed framework achieves state-of-the-art performance on the benchmark CCNU_Mobile dataset, reaching an accuracy of 97.6% for the identification of 45 recording devices, with a significant reduction in training time compared to other models. [ABSTRACT FROM AUTHOR]

In recent years, with the development of artificial intelligence, smart catering has become one of the most popular research fields, where ingredients identification is a necessary and significant link. The automatic identification of ingredients can effectively reduce labor costs in the acceptance stage of the catering process. Although there have been a few methods for ingredients classification, most of them are of low recognition accuracy and poor flexibility. In order to solve these problems, in this paper, we construct a large-scale fresh ingredients database and design an end-to-end multi-attention-based convolutional neural network model for ingredients identification. Our method achieves an accuracy of 95.90% in the classification task, which contains 170 kinds of ingredients. The experiment results indicate that it is the state-of-the-art method for the automatic identification of ingredients. In addition, considering the sudden addition of some new categories beyond our training list in actual applications, we introduce an open-set recognition module to predict the samples outside the training set as the unknown ones. The accuracy of open-set recognition reaches 74.6%. Our algorithm has been deployed successfully in smart catering systems. It achieves an average accuracy of 92% in actual use and saves 60% of the time compared to manual operation, according to the statistics of actual application scenarios. [ABSTRACT FROM AUTHOR]

This paper presents a method for the transparent, robust, and highly capacitive watermarking of video signals using an information mapper. The proposed architecture is based on the use of deep neural networks to embed the watermark in the luminance channel in the YUV color space. An information mapper was used to enable the transformation of a multi-bit binary signature of varying capacitance reflecting the entropy measure of the system into a watermark embedded in the signal frame. To confirm the effectiveness of the method, tests were carried out for video frames with a resolution of 256 × 256 pixels, with a watermark capacity of 4 to 16,384 bits. Transparency metrics (SSIM and PSNR) and a robustness metric—the bit error rate (BER)—were used to assess the performance of the algorithms. [ABSTRACT FROM AUTHOR]

Logo detection is one of the crucial branches in computer vision due to various real-world applications, such as automatic logo detection and recognition, intelligent transportation, and trademark infringement detection. Compared with traditional handcrafted-feature-based methods, deep learning-based convolutional neural networks (CNNs) can learn both low-level and high-level image features. Recent decades have witnessed the great feature representation capabilities of deep CNNs and their variants, which have been very good at discovering intricate structures in high-dimensional data and are thereby applicable to many domains including logo detection. However, logo detection remains challenging, as existing detection methods cannot solve well the problems of a multiscale and large aspect ratios. In this paper, we tackle these challenges by developing a novel long-range dependence involutional network (LDI-Net). Specifically, we designed a strategy that combines a new operator and a self-attention mechanism via rethinking the intrinsic principle of convolution called long-range dependence involution (LD involution) to alleviate the detection difficulties caused by large aspect ratios. We also introduce a multilevel representation neural architecture search (MRNAS) to detect multiscale logo objects by constructing a novel multipath topology. In addition, we implemented an adaptive RoI pooling module (ARM) to improve detection efficiency by addressing the problem of logo deformation. Comprehensive experiments on four benchmark logo datasets demonstrate the effectiveness and efficiency of the proposed approach. [ABSTRACT FROM AUTHOR]

Data-driven fault diagnosis methods for rotating machinery have developed rapidly with the help of deep learning methods. However, traditional intelligent fault diagnosis methods still have some limitations in fault feature extraction and the latest object detection theory has not been applied in fault diagnosis. To this end, a fault diagnosis method based on a sparse short-term Fourier transform (SSTFT) and object detection theory is developed in this paper. First, a sparse constraint is introduced in time-frequency analysis to improve the time-frequency resolution of the model without cross-term interference and proximal gradient descent (PGD) is adopted to quickly and effectively optimize the model to obtain a high-quality time-frequency representation (TFR). Second, a fault diagnosis model based on a region-based convolutional neural network (RCNN) is built; the model can extract multiple regions that can characterize fault features from the TFR. This process avoids the interference of irrelevant vibration components and improves the interpretability of the fault diagnosis model. Finally, multicategory rolling bearing fault identification is realized. The effectiveness of the proposed method is validated by simulation signals and bearing experiments. The results indicate that the proposed method is more effective than existing methods. [ABSTRACT FROM AUTHOR]

In this work, we formulate the image in-painting as a matrix completion problem. Traditional matrix completion methods are generally based on linear models, assuming that the matrix is low rank. When the original matrix is large scale and the observed elements are few, they will easily lead to over-fitting and their performance will also decrease significantly. Recently, researchers have tried to apply deep learning and nonlinear techniques to solve matrix completion. However, most of the existing deep learning-based methods restore each column or row of the matrix independently, which loses the global structure information of the matrix and therefore does not achieve the expected results in the image in-painting. In this paper, we propose a deep matrix factorization completion network (DMFCNet) for image in-painting by combining deep learning and a traditional matrix completion model. The main idea of DMFCNet is to map iterative updates of variables from a traditional matrix completion model into a fixed depth neural network. The potential relationships between observed matrix data are learned in a trainable end-to-end manner, which leads to a high-performance and easy-to-deploy nonlinear solution. Experimental results show that DMFCNet can provide higher matrix completion accuracy than the state-of-the-art matrix completion methods in a shorter running time. [ABSTRACT FROM AUTHOR]

In this paper, we present the multi-stage attentive network (MSAN), an efficient and good generalization performance convolutional neural network (CNN) architecture for motion deblurring. We build a multi-stage encoder–decoder network with self-attention and use the binary cross-entropy loss to train our model. In MSAN, there are two core designs. First, we introduce a new attention-based end-to-end method on top of multi-stage networks, which applies group convolution to the self-attention module, effectively reducing the computing cost and improving the model's adaptability to different blurred images. Secondly, we propose using binary cross-entropy loss instead of pixel loss to optimize our model to minimize the over-smoothing impact of pixel loss while maintaining a good deblurring effect. We conduct extensive experiments on several deblurring datasets to evaluate the performance of our solution for deblurring. Our MSAN achieves superior performance while also generalizing and compares well with state-of-the-art methods. [ABSTRACT FROM AUTHOR]

After high-intensity exercise, athletes have a greatly increased possibility of pneumonia infection due to the immune function of athletes decreasing. Diseases caused by pulmonary bacterial or viral infections can have serious consequences on the health of athletes in a short period of time, and can even lead to their early retirement. Therefore, early diagnosis is the key to athletes' early recovery from pneumonia. Existing identification methods rely too much on professional medical knowledge, which leads to inefficient diagnosis due to the shortage of medical staff. To solve this problem, this paper presents an optimized convolutional neural network recognition method based on an attention mechanism after image enhancement. For the collected images of athlete pneumonia, we first use contrast boost to adjust the coefficient distribution. Then, the edge coefficient is extracted and enhanced to highlight the edge information, and enhanced images of the athlete lungs are obtained by using the inverse curvelet transformation. Finally, an optimized convolutional neural network with an attention mechanism is used to identify the athlete lung images. A series of experimental results show that, compared with the typical image recognition methods based on DecisionTree and RandomForest, the proposed method has higher recognition accuracy for lung images. [ABSTRACT FROM AUTHOR]

In recent years, convolutional neural network (CNN)-based object detection algorithms have made breakthroughs, and much of the research corresponds to hardware accelerator designs. Although many previous works have proposed efficient FPGA designs for one-stage detectors such as Yolo, there are still few accelerator designs for faster regions with CNN features (Faster R-CNN) algorithms. Moreover, CNN's inherently high computational complexity and high memory complexity bring challenges to the design of efficient accelerators. This paper proposes a software-hardware co-design scheme based on OpenCL to implement a Faster R-CNN object detection algorithm on FPGA. First, we design an efficient, deep pipelined FPGA hardware accelerator that can implement Faster R-CNN algorithms for different backbone networks. Then, an optimized hardware-aware software algorithm was proposed, including fixed-point quantization, layer fusion, and a multi-batch Regions of interest (RoIs) detector. Finally, we present an end-to-end design space exploration scheme to comprehensively evaluate the performance and resource utilization of the proposed accelerator. Experimental results show that the proposed design achieves a peak throughput of 846.9 GOP/s at the working frequency of 172 MHz. Compared with the state-of-the-art Faster R-CNN accelerator and the one-stage YOLO accelerator, our method achieves 10 × and 2.1 × inference throughput improvements, respectively. [ABSTRACT FROM AUTHOR]

In the context of "double carbon", as a traditional high energy consumption industry, the textile industry is facing the severe challenges of energy saving and emission reduction. To improve production efficiency in the textile industry, we propose the use of content-based image retrieval technology to shorten the fabric production cycle. However, fabric retrieval has high requirements for results, which makes it difficult for common retrieval methods to be directly applied to fabric retrieval. This paper presents a novel method for fabric image retrieval. Firstly, we define a fine-grained similarity to measure the similarity between two fabric images. Then, a convolutional neural network with a compact structure and cross-domain connections is designed to narrow the gap between fabric images and similarities. To overcome the problems of probabilistic missing and difficult training in classical hashing, we introduce a variational network module and structural module into the hashing model, which is called DVSH. We employ list-wise learning to perform similarity embedding. The experimental results demonstrate the superiority and efficiency of the proposed hashing model, DVSH. [ABSTRACT FROM AUTHOR]

Recently, emotional electroencephalography (EEG) has been of great importance in brain–computer interfaces, and it is more urgent to realize automatic emotion recognition. The EEG signal has the disadvantages of being non-smooth, non-linear, stochastic, and susceptible to background noise. Additionally, EEG signal processing network models have the disadvantages of a large number of parameters and long training time. To address the above issues, a novel model is presented in this paper. Initially, a deep sparse autoencoder network (DSAE) was used to remove redundant information from the EEG signal and reconstruct its underlying features. Further, combining a convolutional neural network (CNN) with long short-term memory (LSTM) can extract relevant features from task-related features, mine the correlation between the 32 channels of the EEG signal, and integrate contextual information from these frames. The proposed DSAE + CNN + LSTM (DCRNN) model was experimented with on the public dataset DEAP. The classification accuracies of valence and arousal reached 76.70% and 81.43%, respectively. Meanwhile, we conducted experiments with other comparative methods to further demonstrate the effectiveness of the DCRNN method. [ABSTRACT FROM AUTHOR]

This paper proposes an accurate short-term prediction model of bike-sharing demand with the hybrid TCN-GRU method. The emergence of shared bicycles has provided people with a low-carbon, green and healthy way of transportation. However, the explosive growth and free-form development of bike-sharing has also brought about a series of problems in the area of urban governance, creating a new opportunity and challenge in the use of a large amount of historical data for regional bike-sharing traffic flow predictions. In this study, we built an accurate short-term prediction model of bike-sharing demand with the bike-sharing dataset from 2015 to 2017 in London. First, we conducted a multidimensional bike-sharing travel characteristics analysis based on explanatory variables such as weather, temperature, and humidity. This will help us to understand the travel characteristics of local people, will facilitate traffic management and, to a certain extent, improve traffic congestion. Then, the explanatory variables that help predict the demand for bike-sharing were obtained using the Granger causality with the entropy theory-based MIC method to verify each other. The Multivariate Temporal Convolutional Network (TCN) and Gated Recurrent Unit (GRU) model were integrated to build the prediction model, and this is abbreviated as the TCN-GRU model. The fitted coefficient of determination R2 and explainable variance score (EVar) of the dataset reached 98.42% and 98.49%, respectively. Meanwhile, the mean absolute error (MAE) and root mean square error (RMSE) were at least 1.98% and 2.4% lower than those in other models. The results show that the TCN-GRU model has strong efficiency and robustness. The model can be used to make short-term accurate predictions of bike-sharing demand in the region, so as to provide decision support for intelligent dispatching and urban traffic safety improvement, which will help to promote the development of green and low-carbon mobility in the future. [ABSTRACT FROM AUTHOR]

Attention mechanisms can improve the performance of neural networks, but the recent attention networks bring a greater computational overhead while improving network performance. How to maintain model performance while reducing complexity is a hot research topic. In this paper, a lightweight Mixture Attention (MA) module is proposed to improve network performance and reduce the complexity of the model. Firstly, the MA module uses multi-branch architecture to process the input feature map in order to extract the multi-scale feature information of the input image. Secondly, in order to reduce the number of parameters, each branch uses group convolution independently, and the feature maps extracted by different branches are fused along the channel dimension. Finally, the fused feature maps are processed using the channel attention module to extract statistical information on the channels. The proposed method is efficient yet effective, e.g., the network parameters and computational cost are reduced by 9.86% and 7.83%, respectively, and the Top-1 performance is improved by 1.99% compared with ResNet50. Experimental results on common-used benchmarks, including CIFAR-10 for classification and PASCAL-VOC for object detection, demonstrate that the proposed MA outperforms the current SOTA methods significantly by achieving higher accuracy while having lower model complexity. [ABSTRACT FROM AUTHOR]

A key element for reducing energy consumption and improving thermal comfort on high-speed rail is controlling air-conditioning temperature. Accurate prediction of air supply temperature is aimed at improving control effects. Existing studies of supply air temperature prediction models are interdisciplinary, involving heat transfer science and computer science, where the problem is defined as time-series prediction. However, the model is widely accepted as a complex model that is nonlinear and dynamic. That makes it difficult for existing statistical and deep learning methods, e.g., autoregressive integrated moving average model (ARIMA), convolutional neural network (CNN), and long short-term memory network (LSTM), to fully capture the interaction between these variables and provide accurate prediction results. Recent studies have shown the potential of the Transformer to increase the prediction capacity. This paper offers an improved temporal fusion transformers (TFT) prediction model for supply air temperature in high-speed train carriages to tackle these challenges, with two improvements: (i) Double-convolutional residual encoder structure based on dilated causal convolution; (ii) Spatio-temporal double-gated structure based on Gated Linear Units. Moreover, this study designs a loss function suitable for general long sequence time-series forecast tasks for temperature forecasting. Empirical simulations using a high-speed rail air-conditioning operation dataset at a specific location in China show that the temperature prediction of the two units using the improved TFT model improves the MAPE by 21.70% and 11.73%, respectively the original model. Furthermore, experiments demonstrate that the model effectively outperforms seven popular methods on time series computing tasks, and the attention of the prediction problem in the time dimension is analyzed. [ABSTRACT FROM AUTHOR]

With the development of convolutional neural networks, the effect of pedestrian detection has been greatly improved by deep learning models. However, the presence of pseudo pedestrians will lead to accuracy reduction in pedestrian detection. To solve the problem that the existing pedestrian detection algorithms cannot distinguish pseudo pedestrians from real pedestrians, a real and pseudo pedestrian detection method with CA-YOLOv5s based on stereo image fusion is proposed in this paper. Firstly, the two-view images of the pedestrian are captured by a binocular stereo camera. Then, a proposed CA-YOLOv5s pedestrian detection algorithm is used for the left-view and right-view images, respectively, to detect the respective pedestrian regions. Afterwards, the detected left-view and right-view pedestrian regions are matched to obtain the feature point set, and the 3D spatial coordinates of the feature point set are calculated with Zhengyou Zhang's calibration method. Finally, the RANSAC plane-fitting algorithm is adopted to extract the 3D features of the feature point set, and the real and pseudo pedestrian detection is achieved by the trained SVM. The proposed real and pseudo pedestrian detection method with CA-YOLOv5s based on stereo image fusion effectively solves the pseudo pedestrian detection problem and efficiently improves the accuracy. Experimental results also show that for the dataset with real and pseudo pedestrians, the proposed method significantly outperforms other existing pedestrian detection algorithms in terms of accuracy and precision. [ABSTRACT FROM AUTHOR]

Convolutional neural networks have greatly improved the performance of image super-resolution. However, perceptual networks have problems such as blurred line structures and a lack of high-frequency information when reconstructing image textures. To mitigate these issues, a generative adversarial network based on multiscale asynchronous learning is proposed in this paper, whereby a pyramid structure is employed in the network model to integrate high-frequency information at different scales. Our scheme employs a U-net as a discriminator to focus on the consistency of adjacent pixels in the input image and uses the LPIPS loss for perceptual extreme super-resolution with stronger supervision. Experiments on benchmark datasets and independent datasets Set5, Set14, BSD100, and SunHays80 show that our approach is effective in restoring detailed texture information from low-resolution images. [ABSTRACT FROM AUTHOR]

Recently, the rapid development of the Internet of Things has contributed to the generation of telemedicine. However, online diagnoses by doctors require the analyses of multiple multi-modal medical images, which are inconvenient and inefficient. Multi-modal medical image fusion is proposed to solve this problem. Due to its outstanding feature extraction and representation capabilities, convolutional neural networks (CNNs) have been widely used in medical image fusion. However, most existing CNN-based medical image fusion methods calculate their weight maps by a simple weighted average strategy, which weakens the quality of fused images due to the effect of inessential information. In this paper, we propose a CNN-based CT and MRI image fusion method (MMAN), which adopts a visual saliency-based strategy to preserve more useful information. Firstly, a multi-scale mixed attention block is designed to extract features. This block can gather more helpful information and refine the extracted features both in the channel and spatial levels. Then, a visual saliency-based fusion strategy is used to fuse the feature maps. Finally, the fused image can be obtained via reconstruction blocks. The experimental results of our method preserve more textual details, clearer edge information and higher contrast when compared to other state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Peer-to-peer lending (P2P lending) has proliferated in recent years thanks to Fintech and big data advancements. However, P2P lending platforms are not tightly governed by relevant laws yet, as their development speed has far exceeded that of regulations. Therefore, P2P lending operations are still subject to risks. This paper proposes prediction models to mitigate the risks of default and asymmetric information on P2P lending platforms. Specifically, we designed sophisticated procedures to pre-process mass data extracted from Lending Club in 2018 Q3–2019 Q2. After that, three statistical models, namely, Logistic Regression, Bayesian Classifier, and Linear Discriminant Analysis (LDA), and five AI models, namely, Decision Tree, Random Forest, LightGBM, Artificial Neural Network (ANN), and Convolutional Neural Network (CNN), were utilized for data analysis. The loan statuses of Lending Club's customers were rationally classified. To evaluate the models, we adopted the confusion matrix series of metrics, AUC-ROC curve, Kolmogorov–Smirnov chart (KS), and Student's t-test. Empirical studies show that LightGBM produces the best performance and is 2.91% more accurate than the other models, resulting in a revenue improvement of nearly USD 24 million for Lending Club. Student's t-test proves that the differences between models are statistically significant. [ABSTRACT FROM AUTHOR]

Automatic building semantic segmentation is the most critical and relevant task in several geospatial applications. Methods based on convolutional neural networks (CNNs) are mainly used in current building segmentation. The requirement of huge pixel-level labels is a significant obstacle to achieve the semantic segmentation of building by CNNs. In this paper, we propose a novel weakly supervised framework for building segmentation, which generates high-quality pixel-level annotations and optimizes the segmentation network. A superpixel segmentation algorithm can predict a boundary map for training images. Then, Superpixels-CRF built on the superpixel regions is guided by spot seeds to propagate information from spot seeds to unlabeled regions, resulting in high-quality pixel-level annotations. Using these high-quality pixel-level annotations, we can train a more robust segmentation network and predict segmentation maps. To iteratively optimize the segmentation network, the predicted segmentation maps are refined, and the segmentation network are retrained. Comparative experiments demonstrate that the proposed segmentation framework achieves a marked improvement in the building's segmentation quality while reducing human labeling efforts. [ABSTRACT FROM AUTHOR]

Manually designing a convolutional neural network (CNN) is an important deep learning method for solving the problem of image classification. However, most of the existing CNN structure designs consume a significant amount of time and computing resources. Over the years, the demand for neural architecture search (NAS) methods has been on the rise. Therefore, we propose a novel deep architecture generation model based on Aquila optimization (AO) and a genetic algorithm (GA). The main contributions of this paper are as follows: Firstly, a new encoding strategy representing the CNN coding structure is proposed, so that the evolutionary computing algorithm can be combined with CNN. Secondly, a new mechanism for updating location is proposed, which incorporates three typical operators from GA cleverly into the model we have designed so that the model can find the optimal solution in the limited search space. Thirdly, the proposed method can deal with the variable-length CNN structure by adding skip connections. Fourthly, combining traditional CNN layers and residual blocks and introducing a grouping strategy provides greater possibilities for searching for the optimal CNN structure. Additionally, we use two notable datasets, consisting of the MNIST and CIFAR-10 datasets for model evaluation. The experimental results show that our proposed model has good results in terms of search accuracy and time. [ABSTRACT FROM AUTHOR]

The methods based on the convolutional neural network have demonstrated its powerful information integration ability in image fusion. However, most of the existing methods based on neural networks are only applied to a part of the fusion process. In this paper, an end-to-end multi-focus image fusion method based on a multi-scale generative adversarial network (MsGAN) is proposed that makes full use of image features by a combination of multi-scale decomposition with a convolutional neural network. Extensive qualitative and quantitative experiments on the synthetic and Lytro datasets demonstrated the effectiveness and superiority of the proposed MsGAN compared to the state-of-the-art multi-focus image fusion methods. [ABSTRACT FROM AUTHOR]

To satisfy the requirements of the end-to-end fault diagnosis of rolling bearings, a hybrid model, based on optimal SWD and 1D-CNN, with the layer of multi-sensor data fusion, is proposed in this paper. Firstly, the BAS optimal algorithm is adopted to obtain the optimal parameters of SWD. After that, the raw signals from different channels of sensors are segmented and preprocessed by the optimal SWD, whose name is BAS-SWD. By which, the sensitive OCs with higher values of spectrum kurtosis are extracted from the raw signals. Subsequently, the improved 1D-CNN model based on VGG-16 is constructed, and the decomposed signals from different channels are fed into the independent convolutional blocks in the model; then, the features extracted from the input signals are fused in the fusion layer. Finally, the fused features are processed by the fully connected layers, and the probability of classification is calculated by the cross-entropy loss function. The result of comparative experiments, based on different datasets, indicates that the proposed model is accurate, effective, and has a good generalization ability. [ABSTRACT FROM AUTHOR]

(1) Background and objective: Cardiovascular disease is one of the most common causes of death in today's world. ECG is crucial in the early detection and prevention of cardiovascular disease. In this study, an improved deep learning method is proposed to diagnose abnormal and normal ECG accurately. (2) Methods: This paper proposes a CNN-FWS that combines three convolutional neural networks (CNN) and recursive feature elimination based on feature weights (FW-RFE), which diagnoses abnormal and normal ECG. F1 score and Recall are used to evaluate the performance. (3) Results: A total of 17,259 records were used in this study, which validated the diagnostic performance of CNN-FWS for normal and abnormal ECG signals in 12 leads. The experimental results show that the F1 score of CNN-FWS is 0.902, and the Recall of CNN-FWS is 0.889. (4) Conclusion: CNN-FWS absorbs the advantages of convolutional neural networks (CNN) to obtain three parts of different spatial information and enrich the learned features. CNN-FWS can select the most relevant features while eliminating unrelated and redundant features by FW-RFE, making the residual features more representative and effective. The method is an end-to-end modeling approach that enables an adaptive feature selection process without human intervention. [ABSTRACT FROM AUTHOR]