Search

Deep learning, a crucial technique for achieving artificial intelligence (AI), has been successfully applied in many fields. The gradual application of the latest architectures of deep learning in the field of time series forecasting (TSF), such as Transformers, has shown excellent performance and results compared to traditional statistical methods. These applications are widely present in academia and in our daily lives, covering many areas including forecasting electricity consumption in power systems, meteorological rainfall, traffic flow, quantitative trading, risk control in finance, sales operations and price predictions for commercial companies, and pandemic prediction in the medical field. Deep learning-based TSF tasks stand out as one of the most valuable AI scenarios for research, playing an important role in explaining complex real-world phenomena. However, deep learning models still face challenges: they need to deal with the challenge of large-scale data in the information age, achieve longer forecasting ranges, reduce excessively high computational complexity, etc. Therefore, novel methods and more effective solutions are essential. In this paper, we review the latest developments in deep learning for TSF. We begin by introducing the recent development trends in the field of TSF and then propose a new taxonomy from the perspective of deep neural network models, comprehensively covering articles published over the past five years. We also organize commonly used experimental evaluation metrics and datasets. Finally, we point out current issues with the existing solutions and suggest promising future directions in the field of deep learning combined with TSF. This paper is the most comprehensive review related to TSF in recent years and will provide a detailed index for researchers in this field and those who are just starting out. [ABSTRACT FROM AUTHOR]

In recent years, deep learning methods have achieved remarkable success in hyperspectral image classification (HSIC), and the utilization of convolutional neural networks (CNNs) has proven to be highly effective. However, there are still several critical issues that need to be addressed in the HSIC task, such as the lack of labeled training samples, which constrains the classification accuracy and generalization ability of CNNs. To address this problem, a deep multi-scale attention fusion network (DMAF-NET) is proposed in this paper. This network is based on multi-scale features and fully exploits the deep features of samples from multiple levels and different perspectives with an aim to enhance HSIC results using limited samples. The innovation of this article is mainly reflected in three aspects: Firstly, a novel baseline network for multi-scale feature extraction is designed with a pyramid structure and densely connected 3D octave convolutional network enabling the extraction of deep-level information from features at different granularities. Secondly, a multi-scale spatial–spectral attention module and a pyramidal multi-scale channel attention module are designed, respectively. This allows modeling of the comprehensive dependencies of coordinates and directions, local and global, in four dimensions. Finally, a multi-attention fusion module is designed to effectively combine feature mappings extracted from multiple branches. Extensive experiments on four popular datasets demonstrate that the proposed method can achieve high classification accuracy even with fewer labeled samples. [ABSTRACT FROM AUTHOR]

Deep learning (DL) networks are a popular choice for short-term load forecasting (STLF) in the residential sector. Hybrid DL methodologies based on convolutional neural networks (CNNs) and long short-term memory networks (LSTMs) have a higher forecasting accuracy than conventional statistical STLF techniques for different types of single-household load series. However, existing load forecasting methodologies are often inefficient when a high load demand persists for a few hours in a day. Peak load consumption is explicitly depicted as a tail in the probability distribution function (PDF) of the load series. Due to the diverse and uncertain nature of peak load demands, DL methodologies have difficulty maintaining consistent forecasting results, particularly when the PDF of the load series has a longer tail. This paper proposes a multihousehold load forecasting strategy based on the collective moment measure (CMM) (which is obtained from the PDF of the load series), data augmentation, and a CNN. Each load series was compared and ordered through CMM indexing, which helped maintain a minimum or constant shifting variance in the dataset inputted to the CNN. Data augmentation was used to enlarge the input dataset and solve the existing data requirement issues of the CNN. With the ordered load series and data augmentation strategy, the simulation results demonstrated a significant improvement in the performance of both single-household and multihousehold load forecasting. The proposed method predicts day-ahead multihousehold loads simultaneously and compares the results based on a single household. The forecasting performance of the proposed method for six different household groups with 10, 20, 30, 50, 80, and 100 household load series was evaluated and compared with those of existing methodologies. The mean absolute percentage error of the prediction results for each multihousehold load series could be improved by more than 3%. This study can help advance the application of DL methods for household load prediction under high-load-demand conditions. [ABSTRACT FROM AUTHOR]

The monitoring of the lifetime of cutting tools often faces problems such as life data loss, drift, and distortion. The prediction of the lifetime in this situation is greatly compromised with respect to the accuracy. The recent rise of deep learning, such as Gated Recurrent Unit Units (GRUs), Hidden Markov Models (HMMs), Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Attention networks, and Transformers, has dramatically improved the data problems in tool lifetime prediction, substantially enhancing the accuracy of tool wear prediction. In this paper, we introduce a novel approach known as PCHIP-Enhanced ConvGRU (PECG), which leverages multiple—feature fusion for tool wear prediction. When compared to traditional models such as CNNs, the CNN Block, and GRUs, our method consistently outperformed them across all key performance metrics, with a primary focus on the accuracy. PECG addresses the challenge of missing tool wear measurement data in relation to sensor data. By employing PCHIP interpolation to fill in the gaps in the wear values, we have developed a model that combines the strengths of both CNNs and GRUs with data augmentation. The experimental results demonstrate that our proposed method achieved an exceptional relative accuracy of 0.8522, while also exhibiting a Pearson's Correlation Coefficient (PCC) exceeding 0.95. This innovative approach not only predicts tool wear with remarkable precision, but also offers enhanced stability. [ABSTRACT FROM AUTHOR]

In recent years, radar emitter signal recognition has enjoyed a wide range of applications in electronic support measure systems and communication security. More and more deep learning algorithms have been used to improve the recognition accuracy of radar emitter signals. However, complex deep learning algorithms and data preprocessing operations have a huge demand for computing power, which cannot meet the requirements of low power consumption and high real-time processing scenarios. Therefore, many research works have remained in the experimental stage and cannot be actually implemented. To tackle this problem, this paper proposes a resource reuse computing acceleration platform based on field programmable gate arrays (FPGA), and implements a one-dimensional (1D) convolutional neural network (CNN) and long short-term memory (LSTM) neural network (NN) model for radar emitter signal recognition, directly targeting the intermediate frequency (IF) data of radar emitter signal for classification and recognition. The implementation of the 1D-CNN-LSTM neural network on FPGA is realized by multiplexing the same systolic array to accomplish the parallel acceleration of 1D convolution and matrix vector multiplication operations. We implemented our network on Xilinx XCKU040 to evaluate the effectiveness of our proposed solution. Our experiments show that the system can achieve 7.34 giga operations per second (GOPS) data throughput with only 5.022 W power consumption when the radar emitter signal recognition rate is 96.53%, which greatly improves the energy efficiency ratio and real-time performance of the radar emitter recognition system. [ABSTRACT FROM AUTHOR]

In order to more effectively monitor and interfere with enemy signals, it is particularly important to accurately and efficiently identify the intercepted signals and estimate their parameters in the increasingly complex electromagnetic environment. Therefore, in non-cooperative situations, it is of great practical significance to study how to accurately detect direct sequence spread spectrum (DSSS) signals in real time and estimate their parameters. The traditional time-delay correlation algorithm encounters the challenges such as peak energy leakage and false peak interference. As an alternative, this paper introduces a Pseudo-Noise (PN) code period estimation method utilizing a one-dimensional (1D) convolutional neural network based on the residual network (CNN-ResNet). This method transforms the problem of spread spectrum code period estimation into a multi-classification problem of spread spectrum code length estimation. Firstly, the In-phase/Quadrature(I/Q) two-way of the received DSSS signals is directly input into the CNN-ResNet model, which will automatically learn the characteristics of the DSSS signal with different PN code lengths and then estimate the PN code length. Simulation experiments are conducted using a data set with DSSS signals ranging from −20 to 10 dB in terms of signal-to-noise ratios (SNRs). Upon training and verifying the model using BPSK modulation, it is then put to the test with QPSK-modulated signals, and the estimation performance was analyzed through metrics such as loss function, accuracy rate, recall rate, and confusion matrix. The results demonstrate that the 1D CNN-ResNet proposed in this paper is capable of effectively estimating the PN code period of the non-cooperative DSSS signal, exhibiting robust generalization abilities. [ABSTRACT FROM AUTHOR]

With the wide application of direct sequence spread spectrum (DSSS) signals, the comprehensive performance of DSSS communication systems has been continuously improved, making the electronic reconnaissance link in communication countermeasures more difficult. Electronic reconnaissance technology, as the fundamental means of modern electronic warfare, mainly includes signal detection, recognition, and parameter estimation. At present, research on DSSS detection algorithms is mostly based on the correlation characteristics of DSSS signals, and autocorrelation algorithm is the most mature and widely used method in practical engineering. With the continuous development of deep learning, deep-learning-based methods have gradually been introduced to replace traditional algorithms in the field of signal processing. This paper proposes a spread spectrum signal detection method based on convolutional neural network (CNN). Through experimental analysis, the detection performance of the CNN model proposed in this paper on DSSS signals in various situations has been compared and analyzed with traditional autocorrelation detection methods for different signal-to-noise ratios. The experiments verified the estimation performance of the model in this paper under different signal-to-noise ratios, different spreading code lengths, different spreading code types, and different modulation methods and compared it with the autocorrelation detection algorithm. It was found that the detection performance of the model in this paper was higher than that of the autocorrelation detection method, and the overall performance was improved by 4 dB. [ABSTRACT FROM AUTHOR]

In the field of image fusion, spatial detail blurring and color distortion appear in synthetic aperture radar (SAR) images and multispectral (MS) during the traditional fusion process due to the difference in sensor imaging mechanisms. To solve this problem, this paper proposes a fusion method for SAR images and MS images based on a convolutional neural network. In order to make use of the spatial information and different scale feature information of high-resolution SAR image, a dual-channel feature extraction module is constructed to obtain a SAR image feature map. In addition, different from the common direct addition strategy, an attention-based feature fusion module is designed to achieve spectral fidelity of the fused images. In order to obtain better spectral and spatial retention ability of the network, an unsupervised joint loss function is designed to train the network. In this paper, the Sentinel 1 SAR images and Landsat 8 MS images are used as datasets for experiments. The experimental results show that the proposed algorithm has better performance in quantitative and visual representation when compared with traditional fusion methods and deep learning algorithms. [ABSTRACT FROM AUTHOR]

This paper proposed a real-time fault diagnostic method for hydraulic systems using data collected from multiple sensors. The method is based on a proposed multi-sensor convolutional neural network (MS-CNN) that incorporates feature extraction, sensor selection, and fault diagnosis into an end-to-end model. Both the sensor selection process and fault diagnosis process are based on abstract fault-related features learned by a CNN deep learning model. Therefore, compared with the traditional sensor-and-feature selection method, the proposed MS-CNN can find the sensor channels containing higher-level fault-related features, which provides two advantages for diagnosis. First, the sensor selection can reduce the redundant information and improve the diagnostic performance of the model. Secondly, the reduced number of sensors simplifies the model, reducing communication burden and computational complexity. These two advantages make the MS-CNN suitable for real-time hydraulic system fault diagnosis, in which the multi-sensor feature extraction and the computation speed are both significant. The proposed MS-CNN approach is evaluated experimentally on an electric-hydraulic subsea control system test rig and an open-source dataset. The proposed method shows obvious superiority in terms of both diagnosis accuracy and computational speed when compared with traditional CNN models and other state-of-the-art multi-sensor diagnostic methods. [ABSTRACT FROM AUTHOR]

The application of Transformer in computer vision has had the most significant influence of all the deep learning developments over the past five years. In addition to the exceptional performance of convolutional neural networks (CNN) in hyperspectral image (HSI) classification, Transformer has begun to be applied to HSI classification. However, for the time being, Transformer has not produced satisfactory results in HSI classification. Recently, in the field of image classification, the creators of Sequencer have proposed a Sequencer structure that substitutes the Transformer self-attention layer with a BiLSTM2D layer and achieves satisfactory results. As a result, this paper proposes a unique network called SquconvNet, that combines CNN with Sequencer block to improve hyperspectral classification. In this paper, we conducted rigorous HSI classification experiments on three relevant baseline datasets to evaluate the performance of the proposed method. The experimental results show that our proposed method has clear advantages in terms of classification accuracy and stability. [ABSTRACT FROM AUTHOR]

With the continuous advancement of target detection technology in remote sensing, target detection technology in images captured by drones has performed well. However, object detection in drone imagery is still a challenge under rainy conditions. Rain is a common severe weather condition, and rain streaks often degrade the image quality of sensors. The main issue of rain streaks removal from a single image is to prevent over smoothing (or underclearing) phenomena. Aiming at the above problems, this paper proposes a deep learning (DL)-based rain streaks removal framework called GSDerainNet, which properly formulates the single image rain streaks removal problem; rain streaks removal is aptly described as a Gaussian Shannon (GS) filter-based image decomposition problem. The GS filter is a novel filter proposed by us, which consists of a parameterized Gaussian function and a scaled Shannon function. Two-dimensional GS filters exhibit high stability and effectiveness in dividing an image into low- and high-frequency parts. In our framework, an input image is first decomposed into a low-frequency part and a high-frequency part by using the GS filter. Rain streaks are located in the high-frequency part. We extract and separate the rain features of the high-frequency part through a deep convolutional neural network (CNN). The experimental results obtained on synthetic data and real data show that the proposed method can better suppress the morphological artifacts caused by filtering. Compared with state-of-the-art single image rain streaks removal methods, the proposed method retains finer image object structures while removing rain streaks. [ABSTRACT FROM AUTHOR]

Synthetic aperture radar (SAR) sensor often produces a shadow in pairs with the target due to its slant-viewing imaging. As a result, shadows in SAR images can provide critical discriminative features for classifiers, such as target contours and relative positions. However, shadows possess unique properties that differ from targets, such as low intensity and sensitivity to depression angles, making it challenging to extract depth features from shadows directly using convolutional neural networks (CNN). In this paper, we propose a new SAR image-classification framework to utilize target and shadow information comprehensively. First, we design a SAR image segmentation method to extract target regions and shadow masks. Second, based on SAR projection geometry, we propose a data-augmentation method to compensate for the geometric distortion of shadows due to differences in depression angles. Finally, we introduce a feature-enhancement module (FEM) based on depthwise separable convolution (DSC) and convolutional block attention module (CBAM), enabling deep networks to fuse target and shadow features adaptively. The experimental results on the Moving and Stationary Target Acquisition and Recognition (MSTAR) dataset show that when only using target and shadow information, the published deep-learning models can still achieve state-of-the-art performance after embedding the FEM. [ABSTRACT FROM AUTHOR]

Hyperspectral image (HSI) classification has been extensively applied for analyzing remotely sensed images. HSI data consist of multiple bands that provide abundant spatial information. Convolutional neural networks (CNNs) have emerged as powerful deep learning methods for processing visual data. In recent work, CNNs have shown impressive results in HSI classification. In this paper, we propose a hierarchical neural network architecture called feature extraction with hybrid spectral CNN (FE-HybridSN) to extract superior spectral–spatial features. FE-HybridSN effectively captures more spectral–spatial information while reducing computational complexity. Considering the prevalent issue of class imbalance in experimental datasets (IP, UP, SV) and real-world hyperspectral datasets, we apply the focal loss to mitigate these problems. The focal loss reconstructs the loss function and facilitates effective achievement of the aforementioned goals. We propose a framework (FEHN-FL) that combines FE-HybridSN and the focal loss for HSI classification and then conduct extensive HSI classification experiments using three remote sensing datasets: Indian Pines (IP), University of Pavia (UP), and Salinas Scene (SV). Using cross-entropy loss as a baseline, we assess the hyperspectral classification performance of various backbone networks and examine the influence of different spatial sizes on classification accuracy. After incorporating focal loss as our loss function, we not only compare the classification performance of the FE-HybridSN backbone network under different loss functions but also evaluate their convergence rates during training. The proposed classification framework demonstrates satisfactory performance compared to state-of-the-art end-to-end deep-learning-based methods, such as 2D-CNN, 3D-CNN, etc. [ABSTRACT FROM AUTHOR]

A brain–computer interface (BCI) is a computer-based system that allows for communication between the brain and the outer world, enabling users to interact with computers using neural activity. This brain signal is obtained from electroencephalogram (EEG) signals. A significant obstacle to the development of BCIs based on EEG is the classification of subject-independent motor imagery data since EEG data are very individualized. Deep learning techniques such as the convolutional neural network (CNN) have illustrated their influence on feature extraction to increase classification accuracy. In this paper, we present a multi-branch (five branches) 2D convolutional neural network that employs several hyperparameters for every branch. The proposed model achieved promising results for cross-subject classification and outperformed EEGNet, ShallowConvNet, DeepConvNet, MMCNN, and EEGNet_Fusion on three public datasets. Our proposed model, EEGNet Fusion V2, achieves 89.6% and 87.8% accuracy for the actual and imagined motor activity of the eegmmidb dataset and scores of 74.3% and 84.1% for the BCI IV-2a and IV-2b datasets, respectively. However, the proposed model has a bit higher computational cost, i.e., it takes around 3.5 times more computational time per sample than EEGNet_Fusion. [ABSTRACT FROM AUTHOR]

Video anomaly event detection (VAED) is one of the key technologies in computer vision for smart surveillance systems. With the advent of deep learning, contemporary advances in VAED have achieved substantial success. Recently, weakly supervised VAED (WVAED) has become a popular VAED technical route of research. WVAED methods do not depend on a supplementary self-supervised substitute task, yet they can assess anomaly scores straightway. However, the performance of WVAED methods depends on pretrained feature extractors. In this paper, we first address taking advantage of two pretrained feature extractors for CNN (e.g., C3D and I3D) and ViT (e.g., CLIP), for effectively extracting discerning representations. We then consider long-range and short-range temporal dependencies and put forward video snippets of interest by leveraging our proposed temporal self-attention network (TSAN). We design a multiple instance learning (MIL)-based generalized architecture named CNN-ViT-TSAN, by using CNN- and/or ViT-extracted features and TSAN to specify a series of models for the WVAED problem. Experimental results on publicly available popular crowd datasets demonstrated the effectiveness of our CNN-ViT-TSAN. [ABSTRACT FROM AUTHOR]

Although measuring worker productivity is crucial, the measurement of the productivity of each worker is challenging due to their dispersion across various construction jobsites. This paper presents a framework for measuring productivity based on an inertial measurement unit (IMU) and activity classification. Two deep learning algorithms and three sensor combinations were utilized to identify and analyze the feasibility of the framework in masonry work. Using the proposed method, worker activity classification could be performed with a maximum accuracy of 96.70% using the convolutional neural network model with multiple sensors, and a minimum accuracy of 72.11% using the long short-term memory (LSTM) model with a single sensor. Productivity could be measured with an accuracy of up to 96.47%. The main contributions of this study are the proposal of a method for classifying detailed activities and an exploration of the effect of the number of IMU sensors used in measuring worker productivity. [ABSTRACT FROM AUTHOR]

Sea fog is a weather hazard along the coast and over the ocean that seriously threatens maritime activities. In the deep learning approach, it is difficult for convolutional neural networks (CNNs) to fully consider global context information in sea fog research due to their own limitations, and the recognition of sea fog edges is relatively vague. To solve the above problems, this paper puts forward an ECA-TransUnet model for daytime sea fog recognition, which consists of a combination of a CNN and a transformer. By designing a two-branch feed-forward network (FFN) module and introducing an efficient channel attention (ECA) module, the model can effectively take into account long-range pixel interactions and feature channel information to capture the global contextual information of sea fog data. Meanwhile, to solve the problem of insufficient existing sea fog detection datasets, we investigated sea fog events occurring in the Yellow Sea and Bohai Sea and their territorial waters, extracted remote sensing images from Moderate Resolution Imaging Spectroradiometer (MODIS) data at corresponding times, and combined data from the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO), cloud and sea fog texture features, and waveband feature information to produce a manually annotated sea fog dataset. Our experiments showed that the proposed model achieves 94.5% accuracy and an 85.8% F1 score. Compared with the existing models relying only on CNNs such as UNet, FCN8s, and DeeplabV3+, it achieves state-of-the-art performance in sea fog recognition. [ABSTRACT FROM AUTHOR]

Due to the imaging mechanism of synthetic aperture radar (SAR), it is difficult and costly to acquire abundant labeled SAR images. Moreover, a typical matched filtering (MF) based image faces the problems of serious noise, sidelobes, and clutters, which will bring down the accuracy of SAR target classification. Different from the MF-based result, a sparse image shows better quality with less noise and higher image signal-to-noise ratio (SNR). Therefore, theoretically using it for target classification will achieve better performance. In this paper, a novel contrastive domain adaptation (CDA) based sparse SAR target classification method is proposed to solve the problem of insufficient samples. In the proposed method, we firstly construct a sparse SAR image dataset by using the complex image based iterative soft thresholding (BiIST) algorithm. Then, the simulated and real SAR datasets are simultaneously sent into an unsupervised domain adaptation framework to reduce the distribution difference and obtain the reconstructed simulated SAR images for subsequent target classification. Finally, the reconstructed simulated images are manually labeled and fed into a shallow convolutional neural network (CNN) for target classification along with a small number of real sparse SAR images. Since the current definition of the number of small samples is still vague and inconsistent, this paper defines few-shot as less than 20 per class. Experimental results based on MSTAR under standard operating conditions (SOC) and extended operating conditions (EOC) show that the reconstructed simulated SAR dataset makes up for the insufficient information from limited real data. Compared with other typical deep learning methods based on limited samples, our method is able to achieve higher accuracy especially under the conditions of few shots. [ABSTRACT FROM AUTHOR]

Recently, attention has been paid to the convolutional neural network (CNN) based synthetic aperture radar (SAR) target recognition method. Because of its advantages of automatic feature extraction and the preservation of translation invariance, the recognition accuracies are stronger than traditional methods. However, similar to other deep learning models, CNN is a "black-box" model, whose working process is vague. It is difficult to locate the decision reasons. Because of this, we focus on the process analysis of a pre-trained CNN model. The role of the processing to feature extraction and final recognition decision is discussed. The discussed components of CNN models are convolution, activation function, and full connection. Here, the convolution processing can be deemed as image filtering. The activation function provides a nonlinear element of processing. Moreover, the fully connected layers can also further extract features. In the experiment, four classical CNN models, i.e., AlexNet, VGG16, GoogLeNet, and ResNet-50, are trained by public MSTAR data, which can realize ten-category SAR target recognition. These pre-trained CNN models are processing objects of the proposed process analysis method. After the analysis, the content of the SAR image target features concerned by these pre-trained CNN models is further clarified. In summary, we provide a paradigm to process the analysis of pre-trained CNN models used for SAR target recognition in this paper. To some degree, the adaptability of these models to SAR images is verified. [ABSTRACT FROM AUTHOR]

Ground Control Points (GCPs) are of great significance for applications involving the registration and fusion of heterologous remote sensing images (RSIs). However, utilizing low-level information rather than deep features, traditional methods based on intensity and local image features turn out to be unsuitable for heterologous RSIs because of the large nonlinear radiation difference (NRD), inconsistent resolutions, and geometric distortions. Additionally, the limitations of current heterologous datasets and existing deep-learning-based methods make it difficult to obtain enough precision GCPs from different kinds of heterologous RSIs, especially for thermal infrared (TIR) images that present low spatial resolution and poor contrast. In this paper, to address the problems above, we propose a convolutional neural network-based (CNN-based) layer-adaptive GCPs extraction method for TIR RSIs. Particularly, the constructed feature extraction network is comprised of basic and layer-adaptive modules. The former is used to achieve the coarse extraction, and the latter is designed to obtain high-accuracy GCPs by adaptively updating the layers in the module to capture the fine communal homogenous features of the heterologous RSIs until the GCP precision meets the preset threshold. Experimental results evaluated on TIR images of SDGSAT-1 TIS and near infrared (NIR), short wave infrared (SWIR), and panchromatic (PAN) images of Landsat-8 OLI show that the matching root-mean-square error (RMSE) of TIS images with SWIR and NIR images could reach 0.8 pixels and an even much higher accuracy of 0.1 pixels could be reached between TIS and PAN images, which performs better than those of the traditional methods, such as SIFT, RIFT, and the CNN-based method like D2-Net. [ABSTRACT FROM AUTHOR]

Object detection is a fundamental task in computer vision. Over the past several years, convolutional neural network (CNN)-based object detection models have significantly improved detection accuracyin terms of average precision (AP). Furthermore, feature pyramid networks (FPNs) are essential modules for object detection models to consider various object scales. However, the AP for small objects is lower than the AP for medium and large objects. It is difficult to recognize small objects because they do not have sufficient information, and information is lost in deeper CNN layers. This paper proposes a new FPN model named ssFPN (scale sequence (S 2 ) feature-based feature pyramid network) to detect multi-scale objects, especially small objects. We propose a new scale sequence (S 2 ) feature that is extracted by 3D convolution on the level of the FPN. It is defined and extracted from the FPN to strengthen the information on small objects based on scale-space theory. Motivated by this theory, the FPN is regarded as a scale space and extracts a scale sequence (S 2 ) feature by three-dimensional convolution on the level axis of the FPN. The defined feature is basically scale-invariant and is built on a high-resolution pyramid feature map for small objects. Additionally, the deigned S 2 feature can be extended to most object detection models based on FPNs. We also designed a feature-level super-resolution approach to show the efficiency of the scale sequence (S 2 ) feature. We verified that the scale sequence (S 2 ) feature could improve the classification accuracy for low-resolution images by training a feature-level super-resolution model. To demonstrate the effect of the scale sequence (S 2 ) feature, experiments on the scale sequence (S 2 ) feature built-in object detection approach including both one-stage and two-stage models were conducted on the MS COCO dataset. For the two-stage object detection models Faster R-CNN and Mask R-CNN with the S 2 feature, AP improvements of up to 1.6% and 1.4%, respectively, were achieved. Additionally, the AP S of each model was improved by 1.2% and 1.1%, respectively. Furthermore, the one-stage object detection models in the YOLO series were improved. For YOLOv4-P5, YOLOv4-P6, YOLOR-P6, YOLOR-W6, and YOLOR-D6 with the S 2 feature, 0.9%, 0.5%, 0.5%, 0.1%, and 0.1% AP improvements were observed. For small object detection, the AP S increased by 1.1%, 1.1%, 0.9%, 0.4%, and 0.1%, respectively. Experiments using the feature-level super-resolution approach with the proposed scale sequence (S 2 ) feature were conducted on the CIFAR-100 dataset. By training the feature-level super-resolution model, we verified that ResNet-101 with the S 2 feature trained on LR images achieved a 55.2% classification accuracy, which was 1.6% higher than for ResNet-101 trained on HR images. [ABSTRACT FROM AUTHOR]

Arrhythmias are defined as irregularities in the heartbeat rhythm, which may infrequently occur in a human's life. These arrhythmias may cause potentially fatal complications, which may lead to an immediate risk of life. Thus, the detection and classification of arrhythmias is a pertinent issue for cardiac diagnosis. (1) Background: To capture these sporadic events, an electrocardiogram (ECG), a register containing the heart's electrical function, is considered the gold standard. However, since ECG carries a vast amount of information, it becomes very complex and challenging to extract the relevant information from visual analysis. As a result, designing an efficient (automated) system to analyse the enormous quantity of data possessed by ECG is critical. (2) Method: This paper proposes a hybrid deep learning-based approach to automate the detection and classification process. This paper makes two-fold contributions. First, 1D ECG signals are translated into 2D Scalogram images to automate the noise filtering and feature extraction. Then, based on experimental evidence, by combining two learning models, namely 2D convolutional neural network (CNN) and the Long Short-Term Memory (LSTM) network, a hybrid model called 2D-CNN-LSTM is proposed. (3) Result: To evaluate the efficacy of the proposed 2D-CNN-LSTM approach, we conducted a rigorous experimental study using the widely adopted MIT–BIH arrhythmia database. The obtained results show that the proposed approach provides ≈98.7%, 99%, and 99% accuracy for Cardiac Arrhythmias (ARR), Congestive Heart Failure (CHF), and Normal Sinus Rhythm (NSR), respectively. Moreover, it provides an average sensitivity of the proposed model of 98.33% and a specificity value of 98.35%, for all three arrhythmias. (4) Conclusions: For the classification of arrhythmias, a robust approach has been introduced where 2 D scalogram images of ECG signals are trained over the CNN-LSTM model. The results obtained are better as compared to the other existing techniques and will greatly reduce the amount of intervention required by doctors. For future work, the proposed method can be applied over some live ECG signals and Bi-LSTM can be applied instead of LSTM. [ABSTRACT FROM AUTHOR]

Reliable and timely crop-yield prediction and crop mapping are crucial for food security and decision making in the food industry and in agro-environmental management. The global coverage, rich spectral and spatial information and repetitive nature of remote sensing (RS) data have made them effective tools for mapping crop extent and predicting yield before harvesting. Advanced machine-learning methods, particularly deep learning (DL), can accurately represent the complex features essential for crop mapping and yield predictions by accounting for the nonlinear relationships between variables. The DL algorithm has attained remarkable success in different fields of RS and its use in crop monitoring is also increasing. Although a few reviews cover the use of DL techniques in broader RS and agricultural applications, only a small number of references are made to RS-based crop-mapping and yield-prediction studies. A few recently conducted reviews attempted to provide overviews of the applications of DL in crop-yield prediction. However, they did not cover crop mapping and did not consider some of the critical attributes that reveal the essential issues in the field. This study is one of the first in the literature to provide a thorough systematic review of the important scientific works related to state-of-the-art DL techniques and RS in crop mapping and yield estimation. This review systematically identified 90 papers from databases of peer-reviewed scientific publications and comprehensively reviewed the aspects related to the employed platforms, sensors, input features, architectures, frameworks, training data, spatial distributions of study sites, output scales, evaluation metrics and performances. The review suggests that multiple DL-based solutions using different RS data and DL architectures have been developed in recent years, thereby providing reliable solutions for crop mapping and yield prediction. However, challenges related to scarce training data, the development of effective, efficient and generalisable models and the transparency of predictions should be addressed to implement these solutions at scale for diverse locations and crops. [ABSTRACT FROM AUTHOR]

In this paper, a structural health monitoring (SHM) system is proposed to provide automatic early warning for detecting damage and its location in composite pipelines at an early stage. The study considers a basalt fiber reinforced polymer (BFRP) pipeline with an embedded Fiber Bragg grating (FBG) sensory system and first discusses the shortcomings and challenges with incorporating FBG sensors for accurate detection of damage information in pipelines. The novelty and the main focus of this study is, however, a proposed approach that relies on designing an integrated sensing-diagnostic SHM system that has the capability to detect damage in composite pipelines at an early stage via implementation of an artificial intelligence (AI)-based algorithm combining deep learning and other efficient machine learning methods using an Enhanced Convolutional Neural Network (ECNN) without retraining the model. The proposed architecture replaces the softmax layer by a k-Nearest Neighbor (k-NN) algorithm for inference. Finite element models are developed and calibrated by the results of pipe measurements under damage tests. The models are then used to assess the patterns of the strain distributions of the pipeline under internal pressure loading and under pressure changes due to bursts, and to find the relationship of strains at different locations axially and circumferentially. A prediction algorithm for pipe damage mechanisms using distributed strain patterns is also developed. The ECNN is designed and trained to identify the condition of pipe deterioration so the initiation of damage can be detected. The strain results from the current method and the available experimental results in the literature show excellent agreement. The average error between the ECNN data and FBG sensor data is 0.093%, thus confirming the reliability and accuracy of the proposed method. The proposed ECNN achieves high performance with 93.33 % accuracy (P%), 91.18 % regression rate (R%) and a 90.54 % F1-score (F%). [ABSTRACT FROM AUTHOR]

Genitourinary syndrome of menopause (GSM) is a group of syndromes, including atrophy of the reproductive tract and urinary tract, and sexual dysfunction, caused by decreased levels of hormones, such as estrogen, in women during the transition to, or late stage of, menopause. GSM symptoms can gradually become severe with age and menopausal time, seriously affecting the safety, and physical and mental health, of patients. Optical coherence tomography (OCT) systems can obtain images similar to "optical slices" in a non-destructive manner. This paper presents a neural network, called RVM-GSM, to implement automatic classification tasks for different types of GSM-OCT images. The RVM-GSM module uses a convolutional neural network (CNN) and a vision transformer (ViT) to capture local and global features of the GSM-OCT images, respectively, and, then, fuses the two features in a multi-layer perception module to classify the image. In accordance with the practical needs of clinical work, lightweight post-processing is added to the final surface of the RVM-GSM module to compress the module. Experimental results showed that the accuracy rate of RVM-GSM in the GSM-OCT image classification task was 98.2%. This result is better than those of the CNN and Vit models, demonstrating the promise and potential of the application of RVM-GSM in the physical health and hygiene fields for women. [ABSTRACT FROM AUTHOR]

It is considered that 1 in 10 adults worldwide have diabetes. Diabetic foot ulcers are some of the most common complications of diabetes, and they are associated with a high risk of lower-limb amputation and, as a result, reduced life expectancy. Timely detection and periodic ulcer monitoring can considerably decrease amputation rates. Recent research has demonstrated that computer vision can be used to identify foot ulcers and perform non-contact telemetry by using ulcer and tissue area segmentation. However, the applications are limited to controlled lighting conditions, and expert knowledge is required for dataset annotation. This paper reviews the latest publications on the use of artificial intelligence for ulcer area detection and segmentation. The PRISMA methodology was used to search for and select articles, and the selected articles were reviewed to collect quantitative and qualitative data. Qualitative data were used to describe the methodologies used in individual studies, while quantitative data were used for generalization in terms of dataset preparation and feature extraction. Publicly available datasets were accounted for, and methods for preprocessing, augmentation, and feature extraction were evaluated. It was concluded that public datasets can be used to form a bigger, more diverse datasets, and the prospects of wider image preprocessing and the adoption of augmentation require further research. [ABSTRACT FROM AUTHOR]

The semantic segmentation of remote sensing images is a significant research direction in digital image processing. The complex background environment, irregular size and shape of objects, and similar appearance of different categories of remote sensing images have brought great challenges to remote sensing image segmentation tasks. Traditional convolutional-neural-network-based models often ignore spatial information in the feature extraction stage and pay less attention to global context information. However, spatial context information is important in complex remote sensing images, which means that the segmentation effect of traditional models needs to be improved. In addition, neural networks with a superior segmentation performance often suffer from the problem of high computational resource consumption. To address the above issues, this paper proposes a combination model of a modified multiscale deformable convolutional neural network (mmsDCNN) and dense conditional random field (DenseCRF). Firstly, we designed a lightweight multiscale deformable convolutional network (mmsDCNN) with a large receptive field to generate a preliminary prediction probability map at each pixel. The output of the mmsDCNN model is a coarse segmentation result map, which has the same size as the input image. In addition, the preliminary segmentation result map contains rich multiscale features. Then, the multi-level DenseCRF model based on the superpixel level and the pixel level is proposed, which can make full use of the context information of the image at different levels and further optimize the rough segmentation result of mmsDCNN. To be specific, we converted the pixel-level preliminary probability map into a superpixel-level predicted probability map according to the simple linear iterative clustering (SILC) algorithm and defined the potential function of the DenseCRF model based on this. Furthermore, we added the pixel-level potential function constraint term to the superpixel-based Gaussian potential function to obtain a combined Gaussian potential function, which enabled our model to consider the features of various scales and prevent poor superpixel segmentation results from affecting the final result. To restore the contour of the object more clearly, we utilized the Sketch token edge detection algorithm to extract the edge contour features of the image and fused them into the potential function of the DenseCRF model. Finally, extensive experiments on the Potsdam and Vaihingen datasets demonstrated that the proposed model exhibited significant advantages compared to the current state-of-the-art models. [ABSTRACT FROM AUTHOR]

This paper aims to explore and evaluate aerial imagery and deep learning technology in pavement condition evaluation. A convolutional neural network (CNN) model, named PCIer, was designed to process aerial images and produce pavement condition index (PCI) estimations, which are classified into four scales of Good (PCI ≥ 70), Fair (50 ≤ PCI < 70), Poor (25 ≤ PCI < 50), and Very Poor (PCI < 25). In the experiment, the PCI datasets were retrieved from the published pavement condition report by the City of Sacramento, CA. Following the retrieved datasets, the authors also collected the corresponding aerial image datasets containing 100 images for each PCI grade from Google Earth. An 80% proportion of datasets were used for PCIer model training, and the remaining were used for testing. Comparisons showed using a 128-channel heatmap layer in the proposed PCIer model and saving the PCIer model with the best validation accuracy would yield the best performance, with a testing accuracy of 0.97, and a weighted average precision, recall, and F1-score of 0.98, 0.97, and 0.97, respectively. Moreover, future research recommendations are provided in the discussion for improving the effectiveness of pavement evaluation via aerial imagery and deep learning. [ABSTRACT FROM AUTHOR]

In brain–computer interface (BCI) systems, motor imagery electroencephalography (MI-EEG) signals are commonly used to detect participant intent. Many factors, including low signal-to-noise ratios and few high-quality samples, make MI classification difficult. In order for BCI systems to function, MI-EEG signals must be studied. In pattern recognition and other fields, deep learning approaches have recently been successfully applied. In contrast, few effective deep learning algorithms have been applied to BCI systems, especially MI-based systems. In this paper, we address these problems from two aspects based on the characteristics of EEG signals: first, we proposed a combined time–frequency domain data enhancement method. This method guarantees that the size of the training data is effectively increased while maintaining the intrinsic composition of the data. Second, our design consists of a parallel CNN that takes both raw EEG images and images transformed through continuous wavelet transform (CWT) as inputs. We conducted classification experiments on a public data set to verify the effectiveness of the algorithm. According to experimental results based on the BCI Competition IV Dataset2a, the average classification accuracy is 97.61%. A comparison of the proposed algorithm with other algorithms shows that it performs better in classification. The algorithm can be used to improve the classification performance of MI-based BCIs and BCI systems created for people with disabilities. [ABSTRACT FROM AUTHOR]

Cloud occlusion phenomena are widespread in optical remote sensing (RS) images, leading to information loss and image degradation and causing difficulties in subsequent applications such as land surface classification, object detection, and land change monitoring. Therefore, thin cloud removal is a key preprocessing procedure for optical RS images, and has great practical value. Recent deep learning-based thin cloud removal methods have achieved excellent results. However, these methods have a common problem in that they cannot obtain large receptive fields while preserving image detail. In this paper, we propose a novel wavelet-integrated convolutional neural network for thin cloud removal (WaveCNN-CR) in RS images that can obtain larger receptive fields without any information loss. WaveCNN-CR generates cloud-free images in an end-to-end manner based on an encoder–decoder-like architecture. In the encoding stage, WaveCNN-CR first extracts multi-scale and multi-frequency components via wavelet transform, then further performs feature extraction for each high-frequency component at different scales by multiple enhanced feature extraction modules (EFEM) separately. In the decoding stage, WaveCNN-CR recursively concatenates the processed low-frequency and high-frequency components at each scale, feeds them into EFEMs for feature extraction, then reconstructs the high-resolution low-frequency component by inverse wavelet transform. In addition, the designed EFEM consisting of an attentive residual block (ARB) and gated residual block (GRB) is used to emphasize the more informative features. ARB and GRB enhance features from the perspective of global and local context, respectively. Extensive experiments on the T-CLOUD, RICE1, and WHUS2-CR datasets demonstrate that our WaveCNN-CR significantly outperforms existing state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Some recent studies use a convolutional neural network (CNN) or long short-term memory (LSTM) to extract gait features, but the methods based on the CNN and LSTM have a high loss rate of time-series and spatial information, respectively. Since gait has obvious time-series characteristics, while CNN only collects waveform characteristics, and only uses CNN for gait recognition, this leads to a certain lack of time-series characteristics. LSTM can collect time-series characteristics, but LSTM results in performance degradation when processing long sequences. However, using CNN can compress the length of feature vectors. In this paper, a sequential convolution LSTM network for gait recognition using multimodal wearable inertial sensors is proposed, which is called SConvLSTM. Based on 1D-CNN and a bidirectional LSTM network, the method can automatically extract features from the raw acceleration and gyroscope signals without a manual feature design. 1D-CNN is first used to extract the high-dimensional features of the inertial sensor signals. While retaining the time-series features of the data, the dimension of the features is expanded, and the length of the feature vectors is compressed. Then, the bidirectional LSTM network is used to extract the time-series features of the data. The proposed method uses fixed-length data frames as the input and does not require gait cycle detection, which avoids the impact of cycle detection errors on the recognition accuracy. We performed experiments on three public benchmark datasets: UCI-HAR, HuGaDB, and WISDM. The results show that SConvLSTM performs better than most of those reporting the best performance methods, at present, on the three datasets. [ABSTRACT FROM AUTHOR]

In this paper, we propose a novel technique for the inspection of high-density polyethylene (HDPE) pipes using ultrasonic sensors, signal processing, and deep neural networks (DNNs). Specifically, we propose a technique that detects whether there is a diversion on a pipe or not. The proposed model transmits ultrasound signals through a pipe using a custom-designed array of piezoelectric transmitters and receivers. We propose to use the Zadoff–Chu sequence to modulate the input signals, then utilize its correlation properties to estimate the pipe channel response. The processed signal is then fed to a DNN that extracts the features and decides whether there is a diversion or not. The proposed technique demonstrates an average classification accuracy of 90.3 % (when one sensor is used) and 99.6 % (when two sensors are used) on 3 4 inch pipes. The technique can be readily generalized for pipes of different diameters and materials. [ABSTRACT FROM AUTHOR]

Transient stability (TS) and short-term voltage stability (STVS) assessment are of fundamental importance for the operation security of power systems. Both phenomena can be mutually influenced in weak power systems due to the proliferation of power electronic interface devices and the phase-out of conventional heavy machines (e.g., thermal power plants). There is little research on the assessment of both types of stability together, despite the fact that they develop over the same short-term period, and that they can have a major influence on the overall transient performance driven by large electrical disturbances (e.g., short circuits). This work addresses this open research challenge by proposing a methodology for the joint assessment of TS and STVS. The methodology aims at estimating the resulting short-term stability state (STSS) in stable, or unstable conditions, following critical events, such as the synchronism loss of synchronous generators (SG) or the stalling of induction motors (IM). The estimations capture the mechanisms responsible for the degradations of TS and STVS, respectively. The paper overviews the off-line design of the data-driven STSS classification methodology, which supports the design and training of a hybrid deep neural network RCNN (recurrent convolutional neural network). The RCNN can automatically capture spatial and temporal features from the power system through a time series of selected physical variables, which results in a high estimation degree for STSS in real-time applications. The methodology is tested on the New England 39-bus system, where the results demonstrate the superiority of the proposed methodology over other traditional and deep learning-based methodologies. For reference purposes, the numerical tests also illustrate the classification performance in special situations, when the training is performed by exclusively using measurements from generation and motor load buses, which constitute locations where the investigated stability can be observed. [ABSTRACT FROM AUTHOR]

The spatial distribution of remote-sensing scene images is highly complex in character, so how to extract local key semantic information and discriminative features is the key to making it possible to classify accurately. However, most of the existing convolutional neural network (CNN) models tend to have global feature representations and lose the shallow features. In addition, when the network is too deep, gradient disappearance and overfitting tend to occur. To solve these problems, a lightweight, multi-instance CNN model for remote sensing scene classification is proposed in this paper: MILRDA. In the instance extraction and classifier part, more discriminative features are extracted by the constructed residual dense attention block (RDAB) while retaining shallow features. Then, the extracted features are transformed into instance-level vectors and the local information associated with bag-level labels is highlighted by the proposed channel-attention-based multi-instance pooling, while suppressing the weights of useless objects or backgrounds. Finally, the network is constrained by the cross-entropy loss function to output the final prediction results. The experimental results on four public datasets show that our proposed method can achieve comparable results to other state-of-the-art methods. Moreover, the visualization of feature maps shows that MILRDA can find more effective features. [ABSTRACT FROM AUTHOR]

The development of the smartphone and computer vision technique provides customers with a convenient approach to identify tea species, as well as qualities. However, the prediction model may not behave robustly due to changes in illumination conditions. Fluorescence imaging can induce the fluorescence signal from typical components, and thus may improve the prediction accuracy. In this paper, a tea classification method based on fluorescence imaging and convolutional neural networks (CNN) is proposed. Ultra-violet (UV) LEDs with a central wavelength of 370 nm were utilized to induce the fluorescence of tea samples so that the fluorescence images could be captured. Five kinds of tea were included and pre-processed. Two CNN-based classification models, e.g., the VGG16 and ResNet-34, were utilized for model training. Images captured under the conventional fluorescent lamp were also tested for comparison. The results show that the accuracy of the classification model based on fluorescence images is better than those based on the white-light illumination images, and the performance of the VGG16 model is better than the ResNet-34 model in our case. The classification accuracy of fluorescence images reached 97.5%, which proves that the LED-induced fluorescence imaging technique is promising to use in our daily life. [ABSTRACT FROM AUTHOR]

Vehicle fault detection and diagnosis (VFDD) along with predictive maintenance (PdM) are indispensable for early diagnosis in order to prevent severe accidents due to mechanical malfunction in urban environments. This paper proposes an early voiceprint driving fault identification system using machine learning algorithms for classification. Previous studies have examined driving fault identification, but less attention has focused on using voiceprint features to locate corresponding faults. This research uses 43 different common vehicle mechanical malfunction condition voiceprint signals to construct the dataset. These datasets were filtered by linear predictive coefficient (LPC) and wavelet transform(WT). After the original voiceprint fault sounds were filtered and obtained the main fault characteristics, the deep neural network (DNN), convolutional neural network (CNN), and long short-term memory (LSTM) architectures are used for identification. The experimental results show that the accuracy of the CNN algorithm is the best for the LPC dataset. In addition, for the wavelet dataset, DNN has the best performance in terms of identification performance and training time. After cross-comparison of experimental results, the wavelet algorithm combined with DNN can improve the identification accuracy by up to 16.57% compared with other deep learning algorithms and reduce the model training time by up to 21.5% compared with other algorithms. Realizing the cross-comparison of recognition results through various machine learning methods, it is possible for the vehicle to proactively remind the driver of the real-time potential hazard of vehicle machinery failure. [ABSTRACT FROM AUTHOR]

This paper presents an automatic recognition system for classifying stones belonging to different Calabrian quarries (Southern Italy). The tool for stone recognition has been developed in the SILPI project (acronym of "Sistema per l'Identificazione di Lapidei Per Immagini"), financed by POR Calabria FESR-FSE 2014-2020. Our study is based on the Convolutional Neural Network (CNNs) that is used in literature for many different tasks such as speech recognition, neural language processing, bioinformatics, image classification and much more. In particular, we propose a two-stage hybrid approach based on the use of a model of Deep Learning (DL), in our case the CNN, in the first stage and a model of Machine Learning (ML) in the second one. In this work, we discuss a possible solution to stones classification which uses a CNN for the feature extraction phase and the Softmax or Multinomial Logistic Regression (MLR), Support Vector Machine (SVM), k-Nearest Neighbors (kNN), Random Forest (RF) and Gaussian Naive Bayes (GNB) ML techniques in order to perform the classification phase basing our study on the approach called Transfer Learning (TL). We show the image acquisition process in order to collect adequate information for creating an opportune database of the stone typologies present in the Calabrian quarries, also performing the identification of quarries in the considered region. Finally, we show a comparison of different DL and ML combinations in our Two-Stage Hybrid Model solution. [ABSTRACT FROM AUTHOR]

Convolutional neural networks (CNNs) play an important role in hyperspectral image (HSI) classification due to their powerful feature extraction ability. Multiscale information is an important means of enhancing the feature representation ability. However, current HSI classification models based on deep learning only use fixed patches as the network input, which may not well reflect the complexity and richness of HSIs. While the existing methods achieve good classification performance for large-scale scenes, the classification of boundary locations and small-scale scenes is still challenging. In addition, dimensional dislocation often exists in the feature fusion process, and the up/downsampling operation for feature alignment may introduce extra noise or result in feature loss. Aiming at the above issues, this paper deeply explores multiscale features, proposes an adaptive attention constraint fusion module for different scale features, and designs a semantic feature enhancement module for high-dimensional features. First, HSI data of two different spatial scales are fed into the model. For the two inputs, we upsample them using bilinear interpolation to obtain their subpixel data. The proposed multiscale feature extraction module is intended to extract the features of the above four parts of the data. For the extracted features, the multiscale attention fusion module is used for feature fusion, and then, the fused features are fed into the high-level feature semantic enhancement module. Finally, based on the fully connected layer and softmax layer, the prediction results of the proposed model are obtained. Experimental results on four public HSI databases verify that the proposed method outperforms several state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Weather detection systems (WDS) have an indispensable role in supporting the decisions of autonomous vehicles, especially in severe and adverse circumstances. With deep learning techniques, autonomous vehicles can effectively identify outdoor weather conditions and thus make appropriate decisions to easily adapt to new conditions and environments. This paper proposes a deep learning (DL)-based detection framework to categorize weather conditions for autonomous vehicles in adverse or normal situations. The proposed framework leverages the power of transfer learning techniques along with the powerful Nvidia GPU to characterize the performance of three deep convolutional neural networks (CNNs): SqueezeNet, ResNet-50, and EfficientNet. The developed models have been evaluated on two up-to-date weather imaging datasets, namely, DAWN2020 and MCWRD2018. The combined dataset has been used to provide six weather classes: cloudy, rainy, snowy, sandy, shine, and sunrise. Experimentally, all models demonstrated superior classification capacity, with the best experimental performance metrics recorded for the weather-detection-based ResNet-50 CNN model scoring 98.48%, 98.51%, and 98.41% for detection accuracy, precision, and sensitivity. In addition to this, a short detection time has been noted for the weather-detection-based ResNet-50 CNN model, involving an average of 5 (ms) for the time-per-inference step using the GPU component. Finally, comparison with other related state-of-art models showed the superiority of our model which improved the classification accuracy for the six weather conditions classifiers by a factor of 0.5–21%. Consequently, the proposed framework can be effectively implemented in real-time environments to provide decisions on demand for autonomous vehicles with quick, precise detection capacity. [ABSTRACT FROM AUTHOR]

This paper presents an automatic ECG signal classification system that applied the Deep Learning (DL) model to classify four types of ECG signals. In the first part of our work, we present the model development. Four different classes of ECG signals from the PhysioNet open-source database were selected and used. This preliminary study used a Deep Learning (DL) technique namely Convolutional Neural Network (CNN) to classify and predict the ECG signals from four different classes: normal, sudden death, arrhythmia, and supraventricular arrhythmia. The classification and prediction process includes pulse extraction, image reshaping, training dataset, and testing process. In general, the training accuracy achieved up to 95% after 100 epochs. However, the prediction of each ECG single type shows a differentiation. Among the four classes, the results show that the predictions for sudden death ECG waveforms are the highest, i.e., 80 out of 80 samples are correct (100% accuracy). In contrast, the lowest is the prediction for normal sinus ECG waveforms, i.e., 74 out of 80 samples are correct (92.5% accuracy). This is due to the image features of normal sinus ECG waveforms being almost similar to the image features of supraventricular arrhythmia ECG waveforms. However, the model has been tuned to achieve an optimal prediction. In the second part, we presented the hardware implementation with the predictive model embedded in an NVIDIA Jetson Nanoprocessor for the online and real-time classification of ECG waveforms. [ABSTRACT FROM AUTHOR]

With the development of technology and the rise of the meta-universe concept, the brain-computer interface (BCI) has become a hotspot in the research field, and the BCI based on motor imagery (MI) EEG has been widely concerned. However, in the process of MI-EEG decoding, the performance of the decoding model needs to be improved. At present, most MI-EEG decoding methods based on deep learning cannot make full use of the temporal and frequency features of EEG data, which leads to a low accuracy of MI-EEG decoding. To address this issue, this paper proposes a two-branch convolutional neural network (TBTF-CNN) that can simultaneously learn the temporal and frequency features of EEG data. The structure of EEG data is reconstructed to simplify the spatio-temporal convolution process of CNN, and continuous wavelet transform is used to express the time-frequency features of EEG data. TBTF-CNN fuses the features learned from the two branches and then inputs them into the classifier to decode the MI-EEG. The experimental results on the BCI competition IV 2b dataset show that the proposed model achieves an average classification accuracy of 81.3% and a kappa value of 0.63. Compared with other methods, TBTF-CNN achieves a better performance in MI-EEG decoding. The proposed method can make full use of the temporal and frequency features of EEG data and can improve the decoding accuracy of MI-EEG. [ABSTRACT FROM AUTHOR]

Object detection is one of the vital and challenging tasks of computer vision. It supports a wide range of applications in real life, such as surveillance, shipping, and medical diagnostics. Object detection techniques aim to detect objects of certain target classes in a given image and assign each object to a corresponding class label. These techniques proceed differently in network architecture, training strategy and optimization function. In this paper, we focus on animal species detection as an initial step to mitigate the negative impacts of wildlife--human and wildlife--vehicle encounters in remote wilderness regions and on highways. Our goal is to provide a summary of object detection techniques based on R-CNN models, and to enhance the performance of detecting animal species in accuracy and speed, by using four different R-CNN models and a deformable convolutional neural network. Each model is applied on three wildlife datasets, results are compared and analyzed by using four evaluation metrics. Based on the evaluation, an animal species detection system is proposed. [ABSTRACT FROM AUTHOR]

Based on experimental observations, there is a correlation between time and consecutive gaze positions in visual behaviors. Previous studies on gaze point estimation usually use images as the input for model trainings without taking into account the sequence relationship between image data. In addition to the spatial features, the temporal features are considered to improve the accuracy in this paper by using videos instead of images as the input data. To be able to capture spatial and temporal features at the same time, the convolutional neural network (CNN) and long short-term memory (LSTM) network are introduced to build a training model. In this way, CNN is used to extract the spatial features, and LSTM correlates temporal features. This paper presents a CNN Concatenating LSTM network (CCLN) that concatenates spatial and temporal features to improve the performance of gaze estimation in the case of time-series videos as the input training data. In addition, the proposed model can be optimized by exploring the numbers of LSTM layers, the influence of batch normalization (BN) and global average pooling layer (GAP) on CCLN. It is generally believed that larger amounts of training data will lead to better models. To provide data for training and prediction, we propose a method for constructing datasets of video for gaze point estimation. The issues are studied, including the effectiveness of different commonly used general models and the impact of transfer learning. Through exhaustive evaluation, it has been proved that the proposed method achieves a better prediction accuracy than the existing CNN-based methods. Finally, 93.1% of the best model and 92.6% of the general model MobileNet are obtained. [ABSTRACT FROM AUTHOR]

In this paper, a deep learning-based successive interference cancellation (SIC) scheme for use in nonorthogonal multiple access (NOMA) communication systems is investigated. NOMA has become a notable technique in the field of mobile wireless communication because of its capacity to overcome orthogonality, unlike a conventional orthogonal frequency division multiple access (OFDMA) communication system. In NOMA communication systems, SIC is one of the decoding schemes applied at receivers for downlink NOMA transmissions. In this paper, a convolutional neural network (CNN)-based SIC scheme is proposed to improve performance of the single base station and multiuser NOMA scheme. In contrast to existing SIC schemes, the proposed CNN-based SIC scheme can effectively mitigate losses resulting from imperfections of the SIC. The simulation results indicate that the CNN-based SIC method can successfully relieve conventional SIC impairments and achieve good detection performance. Consequently, a CNN-based SIC scheme can be considered as a potential technique for use in NOMA detection schemes. [ABSTRACT FROM AUTHOR]

Usage of effective classification techniques on Magnetic Resonance Imaging (MRI) helps in the proper diagnosis of brain tumors. Previous studies have focused on the classification of normal (nontumorous) or abnormal (tumorous) brain MRIs using methods such as Support Vector Machine (SVM) and AlexNet. In this paper, deep learning architectures are used to classify brain MRI images into normal or abnormal. Gender and age are added as higher attributes for more accurate and meaningful classification. A deep learning Convolutional Neural Network (CNN)-based technique and a Deep Neural Network (DNN) are also proposed for effective classification. Other deep learning architectures such as LeNet, AlexNet, ResNet, and traditional approaches such as SVM are also implemented to analyze and compare the results. Age and gender biases are found to be more useful and play a key role in classification, and they can be considered essential factors in brain tumor analysis. It is also worth noting that, in most circumstances, the proposed technique outperforms both existing SVM and AlexNet. The overall accuracy obtained is 88% (LeNet Inspired Model) and 80% (CNN-DNN) compared to SVM (82%) and AlexNet (64%), with best accuracy of 100%, 92%, 92%, and 81%, respectively. [ABSTRACT FROM AUTHOR]

Sea level rise is an important and topical issue in the South Pacific region and needs an urgent assessment of trends for informed decision making. This paper presents mean sea level trend assessment using harmonic analysis and a hybrid deep learning (DL) model based on the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN) technique, Convolutional Neural Network (CNN), Gated Recurrent Unit (GRU) and Neighbourhood Component Analysis (NCA) to build a highly accurate sea level forecasting model for three small islands (Fiji, Marshall Island and Papua New Guinea (PNG)) in the South Pacific. For a 20-year period, the estimated mean sea level rise per year from the harmonic computation is obtained: 112 mm for PNG, 98 mm for Marshall Island and 52 mm for Fiji. The DL procedure uses climate and environment-based remote sensing satellite (MODIS, GLDAS-2.0, MODIS TERRA, MERRA-2) predictor variables with tide gauge base mean sea level (MSL) data for model training and development for forecasting. The developed CEEMDAN-CNN-GRU as the objective model is benchmarked by comparison to the hybrid model without data decomposition, CNN-GRU and standalone models, Decision Trees (DT) and Support Vector Regression (SVR). All model performances are evaluated using reliable statistical metrics. The CEEMDAN-CNN-GRU shows superior accuracy when compared with other standalone and hybrid models. It shows an accuracy of >96% for correlation coefficient and an error of <1% for all study sites. [ABSTRACT FROM AUTHOR]

The Cobb angle measurement of the scoliotic spine is prone to inter- and intra-observer variations in the clinical setting. This paper proposes a deep learning architecture for detecting spine vertebrae from X-ray images to evaluate the Cobb angle automatically. The public AASCE MICCAI 2019 anterior-posterior X-ray image dataset and local images were used to train and test the proposed convolutional neural network architecture. Sixty-eight landmark features of the spine were detected from the input image to obtain seventeen vertebrae on the spine. The vertebrae locations obtained were processed to automatically measure the Cobb angle. The proposed method can measure the Cobb angle with accuracies up to 93.6% and has excellent reliability compared to clinicians' measurement (intraclass correlation coefficient > 0.95). The proposed deep learning architecture may be used as a tool to augment Cobb angle measurement in X-ray images of patients with adolescent idiopathic scoliosis in a real-world clinical setting. [ABSTRACT FROM AUTHOR]

Accurate traffic status prediction is of great importance to improve the security and reliability of the intelligent transportation system. However, urban traffic status prediction is a very challenging task due to the tight symmetry among the Human–Vehicle–Environment (HVE). The recently proposed spatial–temporal 3D convolutional neural network (ST-3DNet) effectively extracts both spatial and temporal characteristics in HVE, but ignores the essential long-term temporal characteristics and the symmetry of historical data. Therefore, a novel spatial–temporal 3D residual correlation network (ST-3DRCN) is proposed for urban traffic status prediction in this paper. The ST-3DRCN firstly introduces the Pearson correlation coefficient method to extract a high correlation between traffic data. Then, a dynamic spatial feature extraction component is constructed by using 3D convolution combined with residual units to capture dynamic spatial features. After that, based on the idea of long short-term memory (LSTM), a novel architectural unit is proposed to extract dynamic temporal features. Finally, the spatial and temporal features are fused to obtain the final prediction results. Experiments have been performed using two datasets from Chengdu, China (TaxiCD) and California, USA (PEMS-BAY). Taking the root mean square error (RMSE) as the evaluation index, the prediction accuracy of ST-3DRCN on TaxiCD dataset is 21.4%, 21.3%, 11.7%, 10.8%, 4.7%, 3.6% and 2.3% higher than LSTM, convolutional neural network (CNN), 3D-CNN, spatial–temporal residual network (ST-ResNet), spatial–temporal graph convolutional network (ST-GCN), dynamic global-local spatial–temporal network (DGLSTNet), and ST-3DNet, respectively. [ABSTRACT FROM AUTHOR]

In recent years, supervised learning-based methods have achieved excellent performance for hyperspectral image (HSI) classification. However, the collection of training samples with labels is not only costly but also time-consuming. This fact usually causes the existence of weak supervision, including incorrect supervision where mislabeled samples exist and incomplete supervision where unlabeled samples exist. Focusing on the inaccurate supervision and incomplete supervision, the weakly supervised classification of HSI is investigated in this paper. For inaccurate supervision, complementary learning (CL) is firstly introduced for HSI classification. Then, a new method, which is based on selective CL and convolutional neural network (SeCL-CNN), is proposed for classification with noisy labels. For incomplete supervision, a data augmentation-based method, which combines mixup and Pseudo-Label (Mix-PL) is proposed. And then, a classification method, which combines Mix-PL and CL (Mix-PL-CL), is designed aiming at better semi-supervised classification capacity of HSI. The proposed weakly supervised methods are evaluated on three widely-used hyperspectral datasets (i.e., Indian Pines, Houston, and Salinas datasets). The obtained results reveal that the proposed methods provide competitive results compared to the state-of-the-art methods. For inaccurate supervision, the proposed SeCL-CNN has outperformed the state-of-the-art method (i.e., SSDP-CNN) by 0.92%, 1.84%, and 1.75% in terms of OA on the three datasets, when the noise ratio is 30%. And for incomplete supervision, the proposed Mix-PL-CL has outperformed the state-of-the-art method (i.e., AROC-DP) by 1.03%, 0.70%, and 0.82% in terms of OA on the three datasets, with 25 training samples per class. [ABSTRACT FROM AUTHOR]

This paper presents a wearable device, fitted on the waist of a participant that recognizes six activities of daily living (walking, walking upstairs, walking downstairs, sitting, standing, and laying) through a deep-learning algorithm, human activity recognition (HAR). The wearable device comprises a single-board computer (SBC) and six-axis sensors. The deep-learning algorithm employs three parallel convolutional neural networks for local feature extraction and for subsequent concatenation to establish feature fusion models of varying kernel size. By using kernels of different sizes, relevant local features of varying lengths were identified, thereby increasing the accuracy of human activity recognition. Regarding experimental data, the database of University of California, Irvine (UCI) and self-recorded data were used separately. The self-recorded data were obtained by having 21 participants wear the device on their waist and perform six common activities in the laboratory. These data were used to verify the proposed deep-learning algorithm on the performance of the wearable device. The accuracy of these six activities in the UCI dataset and in the self-recorded data were 97.49% and 96.27%, respectively. The accuracies in tenfold cross-validation were 99.56% and 97.46%, respectively. The experimental results have successfully verified the proposed convolutional neural network (CNN) architecture, which can be used in rehabilitation assessment for people unable to exercise vigorously. [ABSTRACT FROM AUTHOR]