MSRConvNet: Classification of railway track defects using multi-scale residual convolutional neural network

Data interpolation is a very important step in seismic data processing. At present, there are many methods based on deep learning to solve the problem of seismic data interpolation, but few studies adopt the idea of multi-scale residuals combined with attention mechanism. In this paper, a seismic data interpolation algorithm based on multi-scale residual attention network is proposed. Multi-scale convolution is applied to residual structure for the first time, and the information of seismic data is extracted adaptively by cascade of multi-scale residual blocks. Then, the output of each residual block is used as hierarchical feature for global feature fusion. In the stage of residual mapping, the residual channel attention block is introduced to promote the flow of information features and make the channel attention mechanism focus on the high frequency and edge information of seismic data, thus enhancing the ability of network representation. A lot of numerical experiments show that the algorithm have short computation time and can achieve good reconstruction results.

Convolutional neural networks (CNNs) are becoming increasingly popular in modern remote sensing image processing tasks and exhibit outstanding capability for hyperspectral image (HSI) classification. However, for the existing CNN-based HSI-classification methods, most of them only consider single-scale feature extraction, which may neglect some important fine information and cannot guarantee to capture optimal spatial features. Moreover, many state-of-the-art methods have a huge number of network parameters needed to be tuned, which will cause high computational cost. To address the aforementioned two issues, a novel multiscale residual network (MSRN) is proposed for HSI classification. Specifically, the proposed MSRN introduces depthwise separable convolution (DSC) and replaces the ordinary depthwise convolution in DSC with mixed depthwise convolution (MDConv), which mixes up multiple kernel sizes in a single depthwise convolution operation. The DSC with mixed depthwise convolution (MDSConv) can not only explore features at different scales from each feature map but also greatly reduce learnable parameters in the network. In addition, a multiscale residual block (MRB) is designed by replacing the convolutional layer in an ordinary residual block with the MDSConv layer. The MRB is used as the major unit of the proposed MSRN. Furthermore, to enhance further the feature representation ability, the proposed network adds a high-level shortcut connection (HSC) on the cascaded two MRBs to aggregate lower level features and higher level features. Experimental results on three benchmark HSIs demonstrate the superiority of the proposed MSRN method over several state-of-the-art methods.

To address the problems of insufficient feature information and performance degradation with the deepening of network layers in traditional hyperspectral image classification methods based on convolutional neural networks, this paper proposes a hyperspectral image classification method based on multiscale residual networks. This paper constructs a multiscale residual network model by introducing dilated convolution and residual learning to extract deep-level features from the original hyperspectral image data and achieve the fusion of local features and global features at different scales to enhance the feature expression capability of the model; finally, a Softmax classifier is used to classify the extracted deep-level features. The experimental comparison between this paper and several methods on two hyperspectral images from Indian Pines and Pavia University shows that the proposed method obtains optimal classification results and can improve the classification accuracy with limited training samples.

The study presents the results of the development and testing of deep learning neural network architectures, which demonstrate high accuracy rates in classifying neurophysiological data, in particular hemodynamic brain activation patterns obtained by functional near-infrared spectroscopy, during solving mathematical problems on spatial-numerical associations. The analyzed signal represents a multidimensional time series of oxyhemoglobin and deoxyhemoglobin dynamics. Taking the specificity of the fNIRS signal into account, a comparative analysis of 2 types of neural network architectures was carried out: (1) architectures based on recurrent neural networks: recurrent neural network with long short-term memory, recurrent neural network with long short-term memory with fully connected layers, bidirectional recurrent neural network with long short-term memory, convolutional recurrent neural network with long short-term memory; (2) architectures based on convolutional neural networks with 1D convolutions: convolutional neural network, fully convolutional neural network, residual neural network. Trained long short-term memory recurrent neural network architectures showed worse results in accuracy in comparison with 1D convolutional neural network architectures. Residual neural network (model_Resnet) demonstrated the highest accuracy rates in three experimental conditions more than 88% in detecting age-related differences in brain activation during spatial-numerical association tasks considering the individual characteristics of the respondents’ signal.

Imputation of missing time-activity data with long-term gaps: A multi-scale residual CNN-LSTM network model

Diabetic retinopathy grading based on multi-scale residual network and cross-attention module

Recent studies on video super-resolution (SR) has shown that convolutional neural network (CNN) combined with motion compensation (MC) is able to merge information from multiple low-resolution (LR) frames to generate high quality images. However, Most SR and MC modules based on deep CNN simply increase the depth of the network, which cannot make full use of hierarchical and multi-scale features, thereby achieving relatively low performance. To address the above problem, a novel multi-scale residual dense Network (MSRDN) is proposed in this paper. A multi-scale residual density block (MSRDB) is first designed to extract abundant local features through dense convolution layer, which helps to adaptively detect image features of different scales with convolution kernels of different scales. Then, we redesign SR module and MC module with MSRDB, which adaptively learns more effective features from local features and uses global feature fusion to jointly and adaptively learn global hierarchical features. Comparative results on Vid4 dataset demonstrate that MSRDB can make more full use of feature information, which helps to effectively improve the reconstruction performance of video SR.

Energy disaggregation technology is a key technology to realise real-time non-intrusive load monitoring (NILM). Current energy disaggregation methods use the same scale to extract features from the sequence, which makes part of the local features lost, resulting in low recognition accuracy of electrical appliances. Aiming at the low accuracy of non-intrusive energy disaggregation with low-frequency sampling, a sequential multi-scale residual network is proposed. Multi-scale residual network extracts multi-scale feature information through multi-scale convolution, and uses residual learning to deepen the network structure to further improve the performance of the algorithm. Different scales features are captured by multi-scale convolution to improve the recognition accuracy of electrical appliances with low using frequency. Sequence-to-sequence energy disaggregation mechanism can improve the disaggregation efficiency of the algorithm. The experimental comparison results show that the model can get better disaggregation effect, and can effectively identify the start-stop state of electrical appliances.

A multiscale neural network model for the prediction on the equivalent permeability of discrete fracture network

Brain tumor is generally structured by the subtle summation of anomalous cells because it categorized as a mass of tissue, and it is very crucial to categorize brain tumors from Magnetic Resonance Imaging (MRI) for diagnosis purpose. The investigation of brain tumors by humans is a routine process for detecting and classifying MRI brain tumors. Due to the low contrast of the images, poor boundaries and noises, the image classification techniques provide poor performance in medical images. Hence, a new classification model for detecting brain tumors is needed to solve the critical challenges in existing approaches. This decreases human intervention while making decisions about brain tumors. Therefore, a new brain tumor classification model is investigated to solve the challenges in existing approaches with deep learning. The proposed framework consists of two main different phases (i) classifying normal and abnormal images and (b) Segmentation and severity classification of abnormal images. The first step is comprised of pre-processing and classification. Firstly, the required images are gathered from the benchmark databases. These collected images are undergone pre-processing, which is done by filtering methods and Contrast Limited Adaptive Histogram Equalization (CLAHE). Subsequently, the pre-processed image is fed into the new classification model as a Transformer-based Multi-scale Residual Attention Network (TMRAN), where the normal and abnormal images are obtained. While in the second case, the classified abnormal image is considered as the input for the segmentation process. The segmented image is obtained by Convolutional Neural Network (CNN), Modified UNet and UNet, in turn the parameters in UNet is optimized by using the Modified Controlling Parameters-based African Vultures Optimization Algorithm (MCP-AVOA), termed as Modified UNet. At the final stage, these segmented images are applied as input to the novel TMRAN model, where the hyper parameters are tuned optimally using improved MCP-AVOA in order to acquire the optimal severity classified results. The performance is validated using diverse measures and compared over other conventional methodologies. Thus, the findings elucidate that the proposed model attains higher classification results.

Lightweight single image super-resolution with attentive residual refinement network

This paper highlights the ability of residual convolutional neural network (ResNet) at classifying railway shelling defect dataset without the requirement for handcrafted features. A 41-convolutional layers ResNet with the residual learning block is introduced. Bottleneck architecture of three convolutional layers is used in ResNet to decrease the computation cost. VGG convolutional neural network (VGGNet) classifier and Support Vector Machine (SVM) classifiers based on Histogram of Oriented Gradient (HOG), Local Binary Patterns (LBP) and Scale-invariant Feature Transform (SIFT) are compared with ResNet classifier at classifying our dataset. The performance of ResNet presented in this paper achieves 95% TOP-1 accuracy in testing dataset. In comparison, the result of ResNet is better than VGGNet with 92% TOP-1 accuracy and HOG (42.88%), LBP (52.26%), SIFT (60.69%). In the context of designing neural computing models for ResNet analysis, this paper shows that ResNet used in our experiment is able to not only achieve the high accuracy in railway shelling defect testing dataset, but also has a faster testing speed than the other classifiers.

ResNet and its application to medical image processing: Research progress and challenges

Sequence-based prediction of one dimensional structural properties of proteins has been a long-standing subproblem of protein structure prediction. Recently, prediction accuracy has been significantly improved due to the rapid expansion of protein sequence and structure libraries and advances in deep learning techniques, such as residual convolutional networks (ResNets) and Long-Short-Term Memory Cells in Bidirectional Recurrent Neural Networks (LSTM-BRNNs). Here we leverage an ensemble of LSTM-BRNN and ResNet models, together with predicted residue-residue contact maps, to continue the push towards the attainable limit of prediction for 3- and 8-state secondary structure, backbone angles (θ, τ, ϕ and ψ), half-sphere exposure, contact numbers and solvent accessible surface area (ASA). The new method, named SPOT-1D, achieves similar, high performance on a large validation set and test set (≈1000 proteins in each set), suggesting robust performance for unseen data. For the large test set, it achieves 87% and 77% in 3- and 8-state secondary structure prediction and 0.82 and 0.86 in correlation coefficients between predicted and measured ASA and contact numbers, respectively. Comparison to current state-of-the-art techniques reveals substantial improvement in secondary structure and backbone angle prediction. In particular, 44% of 40-residue fragment structures constructed from predicted backbone Cα-based θ and τ angles are less than 6 Å root-mean-squared-distance from their native conformations, nearly 20% better than the next best. The method is expected to be useful for advancing protein structure and function prediction. SPOT-1D and its data is available at: http://sparks-lab.org/. Supplementary data are available at Bioinformatics online.

Image dehazing is a useful technique which can eliminate the bad effect of haze on images and enhance the performances of image/video processing algorithms in hazy weather. In this chapter, we first propose a single-image dehazing method. We estimate the initial transmission properly based on latent region segmentation and refine the estimated initial transmission by an objective function with a novel weighted L1-norm regularization term. The half-quadratic splitting minimization method is employed to solve this optimization problem. We also define an evaluation function to estimate the reliable global atmospheric light. With the refined transmission map and atmospheric light, we recover the haze-free image by the haze imaging model. We then propose a retinex-based decomposition model for a hazy image and a novel end-to-end image dehazing network. In the model, the illumination of the hazy image is decomposed into natural illumination for the haze-free image and residual illumination caused by haze. Based on this model, we design a deep retinex dehazing network (RDN) to jointly estimate the residual illumination map and the haze-free image. Our RDN consists of a multiscale residual dense network and a U-Net with channel and spatial attention mechanisms. The multiscale residual dense network can simultaneously capture global contextual information from small-scale receptive fields and local detailed information from large-scale receptive fields to precisely estimate the residual illumination map caused by haze. And in the dehazing U-Net, we apply the channel and spatial attention mechanisms in the skip connection to achieve a trade-off between overdehazing and underdehazing by automatically adjusting the channel-wise and pixel-wise attention weights so as to obtain a refined haze-free image.