Multi-scale Convolutional Network Research Articles

A Novel Non-Contact Detection and Identification Method for the Post-Disaster Compression State of Injured Individuals Using UWB Bio-Radar.

Building collapse leads to mechanical injury, which is the main cause of injury and death, with crush syndrome as its most common complication. During the post-disaster search and rescue phase, if rescue personnel hastily remove heavy objects covering the bodies of injured individuals and fail to provide targeted medical care, ischemia-reperfusion injury may be triggered, leading to rhabdomyolysis. This may result in disseminated intravascular coagulation or acute respiratory distress syndrome, further leading to multiple organ failure, which ultimately leads to shock and death. Using bio-radar to detect vital signs and identify compression states can effectively reduce casualties during the search for missing persons behind obstacles. A time-domain ultra-wideband (UWB) bio-radar was applied for the non-contact detection of human vital sign signals behind obstacles. An echo denoising algorithm based on PSO-VMD and permutation entropy was proposed to suppress environmental noise, along with a wounded compression state recognition network based on radar-life signals. Based on training and testing using over 3000 data sets from 10 subjects in different compression states, the proposed multiscale convolutional network achieved a 92.63% identification accuracy. This outperformed SVM and 1D-CNN models by 5.30% and 6.12%, respectively, improving the casualty rescue success and post-disaster precision.

Multi-scale convolutional attention network for lightweight image super-resolution

Convolutional neural network (CNN) based methods have recently achieved extraordinary performance in single image super-resolution (SISR) tasks. However, most existing CNN-based approaches increase the model’s depth by stacking massive kernel convolutions, bringing expensive computational costs and limiting their application in mobile devices with limited resources. Furthermore, large kernel convolutions are rarely used in lightweight super-resolution designs. To alleviate the above problems, we propose a multi-scale convolutional attention network (MCAN), a lightweight and efficient network for SISR. Specifically, a multi-scale convolutional attention (MCA) is designed to aggregate the spatial information of different large receptive fields. Since the contextual information of the image has a strong local correlation, we design a local feature enhancement unit (LFEU) to further enhance the local feature extraction. Extensive experimental results illustrate that our proposed MCAN can achieve better performance with lower model complexity compared with other state-of-the-art lightweight methods.

MSFCA-Net: A Multi-Scale Feature Convolutional Attention Network for Segmenting Crops and Weeds in the Field

Weed control has always been one of the most important issues in agriculture. The research based on deep learning methods for weed identification and segmentation in the field provides necessary conditions for intelligent point-to-point spraying and intelligent weeding. However, due to limited and difficult-to-obtain agricultural weed datasets, complex changes in field lighting intensity, mutual occlusion between crops and weeds, and uneven size and quantity of crops and weeds, the existing weed segmentation methods are unable to perform effectively. In order to address these issues in weed segmentation, this study proposes a multi-scale convolutional attention network for crop and weed segmentation. In this work, we designed a multi-scale feature convolutional attention network for segmenting crops and weeds in the field called MSFCA-Net using various sizes of strip convolutions. A hybrid loss designed based on the Dice loss and focal loss is used to enhance the model’s sensitivity towards different classes and improve the model’s ability to learn from hard samples, thereby enhancing the segmentation performance of crops and weeds. The proposed method is trained and tested on soybean, sugar beet, carrot, and rice weed datasets. Comparisons with popular semantic segmentation methods show that the proposed MSFCA-Net has higher mean intersection over union (MIoU) on these datasets, with values of 92.64%, 89.58%, 79.34%, and 78.12%, respectively. The results show that under the same experimental conditions and parameter configurations, the proposed method outperforms other methods and has strong robustness and generalization ability.

Galaxy morphology classification using multiscale convolution capsule network

ABSTRACT Classification of galaxy morphology is a hot issue in astronomical research. Although significant progress has been made in the last decade in classifying galaxy morphology using deep learning technology, there are still some deficiencies in spatial feature representation and classification accuracy. In this study, we present a multiscale convolutional capsule network (MSCCN) model for the classification of galaxy morphology. First, this model improves the convolutional layers using a multibranch structure to extract the multiscale hidden features of galaxy images. In order to further explore the hidden information in the features, the multiscale features are encapsulated and fed into the capsule layer. Second, we use a sigmoid function to replace the softmax function in dynamic routing, which can enhance the robustness of MSCCN. Finally, the classification model achieves 97 per cent accuracy, 96 per cent precision, 98 per cent recall, and 97 per cent F1-score under macroscopic averaging. In addition, a more comprehensive model evaluation was accomplished in this study. We visualized the morphological features for the part of sample set that used the t-distributed stochastic neighbour embedding (t-SNE) algorithm. The results show that the model has a better generalization ability and robustness, and it can be effectively used in the galaxy morphological classification.

Multi-scale Generative Adversarial Deblurring Network with Gradient Guidance

<>With regards to the lack of crisp edges and a poor recovery of high frequency information such as details in deblurred motion pictures, this research proposes a multi-scale adversarial deblurring network with gradient guidance (MADN). The algorithm uses the classical generative adversarial network (GAN) framework, consisting of a generator and a discriminator. The generator includes a multi-scale convolutional network and a gradient feature extraction network. The multi-scale convolutional network extracts image features at different scales with a nested connection residual codec structure to improve the image edge structure recovery and to increase the perceptual field. This gradient network incorporates with intermediate scale features to extract the gradient features of blurred images to obtain their high frequency information. The generator combines the gradient and multiscale features to recover the remaining high-frequency information in a deblurred image. The loss function of MADN is formed in this research combining adversarial loss, pixel L2-norm loss and mean absolute error. Compared to those experimental results obtained from current deblurring algorithms, our experimental results indicate visually clearer images retaining more information such as edges and details. This MADN algorithm enhances the peak signal-to-noise ratio by an average of 3.32dB and the structural similarity by an average of 0.053.<>

Editorial Issue 34.2

Editorial Issue 34.2

Attention Multihop Graph and Multiscale Convolutional Fusion Network for Hyperspectral Image Classification

Convolutional neural networks (CNNs) for hyperspectral image (HSI) classification have generated good progress. Meanwhile, graph convolutional networks (GCNs) have also attracted considerable attention by using unlabeled data, broadly and explicitly exploiting correlations between adjacent parcels. However, the CNN with a fixed square convolution kernel is not flexible enough to deal with irregular patterns, while the GCN using the superpixel to reduce the number of nodes will lose the pixel-level features, and the features from the two networks are always partial. In this paper, to make good use of the advantages of CNN and GCN, we propose a novel multiple feature fusion model termed attention multi-hop graph and multi-scale convolutional fusion network (AMGCFN), which includes two sub-networks of multi-scale fully CNN and multi-hop GCN to extract the multi-level information of HSI. Specifically, the multi-scale fully CNN aims to comprehensively capture pixel-level features with different kernel sizes, and a multi-head attention fusion module is used to fuse the multi-scale pixel-level features. The multi-hop GCN systematically aggregates the multi-hop contextual information by applying multi-hop graphs on different layers to transform the relationships between nodes, and a multi-head attention fusion module is adopted to combine the multi-hop features. Finally, we design a cross attention fusion module to adaptively fuse the features of two sub-networks. AMGCFN makes full use of multi-scale convolution and multi-hop graph features, which is conducive to the learning of multi-level contextual semantic features. Experimental results on three benchmark HSI datasets show that AMGCFN has better performance than a few state-of-the-art methods.

Multiscale Convolution-Based Probabilistic Classification for Detecting Bare PCB Defects

Defect detection is an essential part of quality management for bare printed circuit board (PCB) production. Existing vision-based methods are not effective in detecting PCB defects when uncertainty exists. This article proposes a multiscale convolution-based detection methodology to classify bare PCB defects under uncertainty. First, a novel window-based loss function is designed to tackle the inter-class imbalance and uncertainty. Then, a multiscale convolution network is constructed to process the defects with intra-class variance, and large scale extraction features are fused on the small scale to guide the extraction process. After that, the classification probability is extracted and assembled into a multiscale probability matrix, on which entropy-based probabilistic decisions are integrated for the final decision. Finally, experimental studies indicate that the proposed methodology can achieve satisfactory detection performance and demonstrate visual interpretability compared to baseline methods.

Computationally Lightweight Hyperspectral Image Classification Using a Multiscale Depthwise Convolutional Network With Channel Attention

Convolutional networks have been widely used for the classification of hyperspectral images; however, such networks are notorious for their large number of trainable parameters and high computational complexity. Additionally, traditional convolution-based methods are typically implemented as a simple cascade of a number of convolutions using a single-scale convolution kernel. In contrast, a lightweight multiscale convolutional network is proposed, capitalizing on feature extraction at multiple scales in parallel branches followed by feature fusion. In this approach, 2D depthwise convolution is used instead of conventional convolution in order to reduce network complexity without sacrificing classification accuracy. Furthermore, multiscale channel attention is also employed to selectively exploit discriminative capability across various channels. To do so, multiple 1D convolutions with varying kernel sizes provide channel attention at multiple scales, again with the goal of minimizing network complexity. Experimental results reveal that the proposed network not only outperforms other competing lightweight classifiers in terms of classification accuracy but also exhibits a lower number of parameters as well as significantly less computational cost.

Dense Multi-Scale Convolutional Network for Plant Segmentation

Dense Multi-Scale Convolutional Network for Plant Segmentation

Lightweight Multiscale Convolutional Networks With Adaptive Pruning for Intelligent Fault Diagnosis of Train Bogie Bearings in Edge Computing Scenarios

Intelligent fault diagnosis of train bogie bearings based on edge computing is a promising technology to ensure the safety and reliability of train operation, which can give fault diagnosis systems better real-time performance and lower communication costs. Lightweight diagnosis networks are the core of this technology. However, existing lightweight diagnosis networks have the following limitations: 1) they lack a lightweight learning mechanism for edge computing scenarios to deal with time-scale changes of bearing fault features when trains operate under variational conditions and 2) an adaptive neural network pruning technique synchronized with training is expected to avoid the repetitive and cumbersome processes of traditional pruning techniques. To overcome the above limitations, this article proposes a new multiscale lightweight network with adaptive pruning for the intelligent diagnosis fault of train bogie bearings in edge computing scenarios. First, weight-sharing multiscale convolutions (WSMSCs) are developed to capture multi-time scale features from vibration signals with extremely low computing costs. Then, inverse separable convolution (ISC) blocks are built to further capture high-level features, and an adaptive pruning technique that does not require artificial jobs is proposed to eliminate useless network structures concurrently with training. Finally, diagnostic results are obtained through a dense layer activated by Softmax. Experimental results demonstrate that the accuracy and complexity of the proposed network are superior to other state-of-the-art lightweight bearing fault diagnosis networks under varying operating conditions. Furthermore, the proposed networks can be deployed on more general edge devices.

Pyramidal Multiscale Convolutional Network With Polarized Self-Attention for Pixel-Wise Hyperspectral Image Classification

In recent years, pixel-wise hyperspectral image (HSI) classification has received growing attention in the field of remote sensing. Plenty of spectral-spatial convolutional neural network (CNN) methods with diverse attention mechanisms have been proposed for HSI classification due to the attention mechanisms being able to provide more flexibility over standard convolutional blocks. However, it remains a challenge to effectively extract multi-scale features of high-resolution HSI in a real-world complex environment. In this paper, we propose a pyramidal multi-scale spectral-spatial convolutional network with polarized self-attention for pixel-wise HSI classification. It contains three stages: channel-wise feature extraction network, spatial-wise feature extraction network, and classification network, which are used to extract spectral features, extract spatial features, and generate classification results, respectively. Pyramidal convolutional blocks and polarized attention blocks are combined to extract spectral and spatial features of HSI. Furthermore, residual aggregation and one-shot aggregation are employed to better converge the network. Experimental results on several public HSI datasets demonstrate that the proposed network outperforms other related methods.

Multi-scale convolutional networks for traffic forecasting with spatial-temporal attention

• Capturing complex spatial-temporal correlations for both short and long term traffic prediction. • Applying soft thresholding fusion to eliminate unimportant spatial-temporal features. • Designing a trainable module to automatically determine the spatial and temporal thresholds. • Constructing multiple convolutions networks and forming groups of units to extract multi-scale features. Traffic forecasting is a classical spatial-temporal data prediction problem with extensive applications, such as dynamic traffic management, route planning and traffic congestion alleviation. However, accurate traffic forecasting is challenging due to complex correlations of traffic data. Few approaches are satisfied with both long-term and short-term prediction tasks. In this paper, we propose a novel Multi-Scale Convolutional Networks (MSCN), an end-to-end solution to solve traffic forecasting problem. MSCN first employs an encoder with spatial-temporal attention mechanism to model both spatial and temporal correlations. Then, the decoder utilizes multiple convolutions and form groups of units to to extract spatial-temporal features for different resolutions. In particular, we propose a soft thresholding fusion mechanism to adaptively adjust the conditions of spatial and temporal correlations. Experiments on two real traffic datasets demonstrate that the proposed MSCN obtains improvements over the state-of-the-art baselines.

Multiscale Backcast Convolution Neural Network for Traffic Flow Prediction in The Frequency Domain

With the construction of intelligent transportation systems in recent years, intelligent methods for the prediction of traffic flow are becoming more and more important, and accurate prediction plays a key role in enabling downstream scheduling algorithms. However, the accuracy of most current forecasting algorithms remains unsatisfactory. Because traffic depends on the time of the day and varies throughout the week, such as during peak commuting periods as opposed to other times, traffic flow data show evident cyclical patterns. We capitalize on this notion and propose a multiscale convolutional feedback network for frequency prediction based on frequency angle. We combine multiscale convolution (MSC) with dilated convolution, and increase the convolutional receptive field by expanding cavity size while retaining similar parameterization costs, and achieve multiscale convolution with kernels referring to different receptive fields. At the same time, we incorporate an autoencoding module by assigning the same set of hidden features to input reconstruction and output prediction, which results in enhanced stability of features within the hidden layers. When we tested our approach on the Traffic dataset, our model achieved the best performance as assessed via the three indicators measured using mean squared error (MSE), mean absolute error (MAE), and correlation coefficient (CORR), with improvements of 3.818%, 2.472% and, 0.1515%, respectively.

MSANet: an improved semantic segmentation method using multi-scale attention for remote sensing images

ABSTRACT With the development of deep learning technology in the field of computer vision, the land cover classification of remote sensing images has gradually developed. In this paper, a deep learning algorithm for land cover classification of remote sensing images is proposed. Two dual-attention mechanism modules with multi-scale spatial attention and channel attention are designed and integrated into the DeepLabv3+ network, called multi-scale convolutional attention network (MCAN) and multi-scale atrous convolutional attention network (MACAN). Besides, Inception-Squeeze-and-Excitation network (ISNet) is added. So, the network can focus on the parts of remote sensing images that need attention at different scales, and it can improve the accuracy of remote sensing semantic segmentation. The experimental results show that the optimized network can achieve a Mean Intersection over Union (mIoU) of 74.03% on the DeepGlobe datasets, which is 14.55%, 13.47%, and 5.73% higher than that of UNet, SegNet, and DeepLabv3+. It also achieved good results on Vaihingen datasets. This result indicates that the optimized network can effectively improve the accuracy of remote sensing land cover classification.

Tool wear prediction based on domain adversarial adaptation and channel attention multiscale convolutional long short-term memory network

Intelligent real-time monitoring of tool wear is significant to ensure the quality of workpieces and the efficiency of machining. However, various factors in the machining process can cause large variations in the monitoring signals, making it difficult to accurately predict tool wear values. To solve this, a tool wear prediction method based on domain adversarial adaptation and squeeze-and-excitation channel attention multiscale convolutional long short-term memory network (SE-DAAMSCLSTM) is proposed. A feature extractor combining multiscale convolution and channel attention with the introduction of domain adversarial mechanism was constructed to extract domain-independent multiscale spatiotemporal features that characterize tool wear, thus enabling accurate prediction of tool wear values. By validating the model on milling datasets and comparing it with conventional prediction methods, the results show that the model enables accurate prediction with variation in tool monitoring signals, demonstrating the superiority of the method in predicting tool wear.

Subject adaptation convolutional neural network for EEG-based motor imagery classification

Objective. Deep transfer learning has been widely used to address the nonstationarity of electroencephalogram (EEG) data during motor imagery (MI) classification. However, previous deep learning approaches suffer from limited classification accuracy because the temporal and spatial features cannot be effectively extracted. Approach. Here, we propose a novel end-to-end deep subject adaptation convolutional neural network (SACNN) to handle the problem of EEG-based MI classification. Our proposed model jointly optimizes three modules, i.e. a feature extractor, a classifier, and a subject adapter. Specifically, the feature extractor simultaneously extracts the temporal and spatial features from the raw EEG data using a parallel multiscale convolution network. In addition, we design a subject adapter to reduce the feature distribution shift between the source and target subjects by using the maximum mean discrepancy. By minimizing the classification loss and the distribution discrepancy, the model is able to extract the temporal-spatial features to the prediction of a new subject. Main results. Extensive experiments are carried out on three EEG-based MI datasets, i.e. brain–computer interface (BCI) competition IV dataset IIb, BCI competition III dataset IVa, and BCI competition IV dataset I, and the average accuracy reaches to 86.42%, 81.71% and 79.35% on the three datasets respectively. Furthermore, the statistical analysis also indicates the significant performance improvement of SACNN. Significance. This paper reveals the importance of the temporal-spatial features on EEG-based MI classification task. Our proposed SACNN model can make fully use of the temporal-spatial information to achieve the purpose.

Remote sensing image super-resolution using multi-scale convolutional sparse coding network.

With the development of convolutional neural networks, impressive success has been achieved in remote sensing image super-resolution. However, the performance of super-resolution reconstruction is unsatisfactory due to the lack of details in remote sensing images when compared to natural images. Therefore, this paper presents a novel multiscale convolutional sparse coding network (MCSCN) to carry out the remote sensing images SR reconstruction with rich details. The MCSCN, which consists of a multiscale convolutional sparse coding module (MCSCM) with dictionary convolution units, can improve the extraction of high frequency features. We can obtain more plentiful feature information by combining multiple sizes of sparse features. Finally, a layer based on sub-pixel convolution that combines global and local features takes as the reconstruction block. The experimental results show that the MCSCN gains an advantage over several existing state-of-the-art methods in terms of peak signal-to-noise ratio and structural similarity.

Multi-scale convolutional network with channel attention mechanism for rolling bearing fault diagnosis

In recent years, deep learning has achieved great success in bearing fault diagnosis due to its robust feature learning capabilities. However, in the actual industry, the diagnostic accuracy would be degraded under varying operation conditions or in noisy environments. To enhance the diagnostic performance in industrial applications, a Multi-scale Convolutional Neural Network with Channel Attention (CA-MCNN) is proposed in this paper. In CA-MCNN, the maximum pooling and average pooling layers are used to extract the multi-scale information of the bearing signals, which increases the dimensions of input. The channel attention mechanism is introduced to increase the convolutional layer feature learning ability by adaptively scoring and assigning weights to the learned features. Moreover, the feature parallel fusion mechanism based on 1-D convolution is applied to capture complementary multi-scale information and reduce network complexity. The performance of CA-MCNN is compared with other fault diagnosis models, and experimental results verify that the CA-MCNN achieves the highest diagnosis accuracy under noisy environments and varying working speeds.

Multi-Scale Convolutional Network for Space-Based ADS-B Signal Separation with Single Antenna

Automatic Dependent Surveillance-Broadcast (ADS-B) signals are very vital in air traffic control. However, the space-based ADS-B signals are easily overlapped and their message cannot be correctly received. It is challenge to separate overlapped signals especially for a single antenna. The existing methods have a low decoding accuracy for small power difference, carrier frequency difference and relative time delay between overlapped signals. In order to solve these problems, we apply the deep learning method to single antenna ADS-B signal separation. A multi-scale Conv-TasNet (MConv-TasNet) is proposed to capture long temporal information of the ADS-B signal. In MConv-TasNet, a multi-scale convolutional separation (MCS) network is proposed to fuse different scale temporal features extracted from overlapping ADS-B signals and generate an effective separation mask to separate signals. Moreover, a large dataset is created by using the real ADS-B data. In addition, the proposed method has been evaluated on the dataset. The average decoding accuracy on the test set is 90.34%. It has achieved the state-of-the-art results.