A lightweight adaptive image deblurring framework using dynamic convolutional neural networks

Inverse identification of cohesive zone parameters for sintered nano-silver joints based on dynamic convolution neural network

To solve the problems of image detail blurring and insufficient utilization of global information in the existing medical image super-resolution reconstruction, this paper proposes a dual-branch fusion network based on residual Transformer network and dynamic convolutional neural network (CTGFSR). The network consists of two branches, one is the global branch based on residual Transformer network, and the other is the local branch based on dynamic convolutional neural network. The global branch uses the self-attention mechanism of Transformer network, which can effectively mine the large-scale global information in the image and improve the overall quality of the image. The local branch uses the characteristic of dynamic convolution to adaptively adjust the convolution kernel parameters, which can enhance the feature extraction ability of convolutional neural network for multi-scale information and improve the detail restoration ability of the image without significantly increasing the network model size. The network uses residual skip connections to preserve the detail information in medical image super-resolution reconstruction. Finally, through the bidirectional gated attention mechanism, the two branches are fused to obtain the final super-resolution reconstruction image. This paper evaluates the performance of the network on two medical image datasets, namely ACDC abdominal MR related to medical image segmentation and L2R2022 lung CT related to registration. The experimental results show that compared with the mainstream super-resolution algorithms, CTGFSR has better overall performance. When the magnification factor is 2 or 4, compared with the convolutional neural network based CFIPC, PDCNCF, ESPCN, FSRCNN, VDSR and the Transformer network based ESRT, SwinIR, the structural similarity SSIM and peak signal-to-noise ratio PSNR have a certain improvement.

Due to its prominent applications, time series classification is one of the most important fields of machine learning. Although there are various approaches for time series classification, dynamic time warping (DTW) is generally considered to be a well-suited distance measure for time series. Therefore, in the early 2000s, techniques based on DTW dominated this field. On the other hand, deep learning techniques, especially convolutional neural networks (CNN) were shown to be able to solve time series classification tasks accurately. Although CNNs are extraordinarily popular, the scalar product in convolution only allows for rigid pattern matching. In this paper, we aim at combining the advantages of DTW and CNN by proposing the dynamic convolution operation and dynamic convolutional neural networks (DCNNs). The main idea behind dynamic convolution is to replace the dot product in convolution by DTW. We perform experiments on 10 publicly available real-world time-series datasets and demonstrate that our proposal leads to statistically significant improvement in terms of classification accuracy in various applications. In order to promote the use of DCNN, we made our implementation publicly available at https://github.com/kr7/DCNN.

CNN+RNN models have become the mainstream approach for semi-supervised sound event detection, and the CNN part is mainly a stack of several 2D convolutional layers to capture the representations of the time-frequency features. However, conventional 2D convolution is of limited ability in capturing detailed information about acoustic events. In this paper, to enhance the representation ability of CNN, we propose NAS-DYMC, a NAS-based dynamic multi-scale convolutional neural network to extract a more effective acoustic representation. Specifically, multi-scale convolution can capture the characteristics of sound events with different time-frequency distributions and dynamic convolution enhances the representation capability of conventional convolution by adapting attention weights onto basis kernels. Furthermore, a neural architecture search (NAS) method is adopted to find the optimal network architecture from the search space consisting of various dynamic multi-scale convolutions for the DCASE 2021 Task4 dataset. Experimental results demonstrate the superiority of our proposed method.

In this work, an improved dynamic convolutional neural network (DCNN) model to accurately identify the risk of a forest fire was established based on the traditional DCNN model. First, the DCNN network model was trained in combination with transfer learning, and multiple pre-trained DCNN models were used to extract features from forest fire images. Second, principal component analysis (PCA) reconstruction technology was used in the appropriate subspace. The constructed 15-layer forest fire risk identification DCNN model named “DCN_Fire” could accurately identify core fire insurance areas. Moreover, the original and enhanced image data sets were used to evaluate the impact of data enhancement on the model’s accuracy. The traditional DCNN model was improved and the recognition speed and accuracy were compared and analyzed with the other three DCNN model algorithms with different architectures. The difficulty of using DCNN to monitor forest fire risk was solved, and the model’s detection accuracy was further improved. The true positive rate was 7.41% and the false positive rate was 4.8%. When verifying the impact of different batch sizes and loss rates on verification accuracy, the loss rate of the DCN_Fire model of 0.5 and the batch size of 50 provided the optimal value for verification accuracy (0.983). The analysis results showed that the improved DCNN model had excellent recognition speed and accuracy and could accurately recognize and classify the risk of a forest fire under natural light conditions, thereby providing a technical reference for preventing and tackling forest fires.

Several efforts have been conducted to handle sparse product rating in e-commerce recommender system. One of them is the inclusion of texts such as product review, abstract, product description, and synopsis. Later, it converted to become rating value. Previous researches have tried to extract these texts based on bag of word and word order. However, this approach was given misunderstanding of text description of products. This research proposes a novel Dynamic Convolutional Neural Network (DCNN) to improve meaning accuracy of product review on a collaborative filtering recommender system. DCNN was used to eliminate item sparse data on text product review while the accuracy level was measured by Root Mean Squared Error (RMSE). The result shows that DCNN has outperformed the other previous methods.

In modern industrial production (such as wind power generation, automobile manufacturing, heavy-duty mining, etc.), bearings often operate for long periods of time in environments with variable loads and strong mechanical vibration noise. Once a fault is not easily detected early, it often leads to equipment shutdown, significantly increased maintenance costs, and even safety accidents. This paper proposes a bearing fault diagnosis algorithm based on Residual Network Multi-Scale Dynamic Convolutional Residual Network-Bidirectional Long Short-Term Memory + Multi head Self Attention (ResDARN-BiLSTM + Attention) to address the problem of insufficient feature extraction ability and low diagnostic accuracy of traditional fault diagnosis methods in such working conditions. The algorithm first uses a residual network with dynamic convolution to extract deep features from the raw vibration signals, followed by multi-scale dynamic convolution layers to capture multi-level information in the signals, thereby enhancing the feature representation capability. Additionally, an improved multi-head self-attention mechanism is introduced to strengthen the focus on key features. A bidirectional long short-term memory network is then used for temporal feature modeling, enabling precise classification of fault patterns. Finally, a classifier is employed for bearing fault diagnosis. The experimental results show that the proposed algorithm can achieve an accuracy of 100% in diagnosing performance on standard datasets without noise, significantly improving the accuracy and robustness of fault detection compared to traditional methods. The average accuracy of diagnostic performance in the presence of noise interference can also reach 92%. This model can extract and identify fault features online, providing feasible technical means for real-time monitoring and preventive maintenance of industrial production lines, helping to reduce equipment downtime losses and improve operation and maintenance efficiency. Compared with other fault diagnosis models, it has strong universality and provides a new technical direction for bearing fault detection and preventive maintenance.

Pan-sharpening is an effective method to obtain high-resolution multispectral images by fusing panchromatic (PAN) images with fine spatial structure and low-resolution multispectral images with rich spectral information. In this article, a multiscale pan-sharpening method based on dynamic convolutional neural network is proposed. The filters in dynamic convolution are generated dynamically and locally by the filter generation network which is different from the standard convolution and strengthens the adaptivity of the network. The dynamic filters are adaptively changed according to the input images. The proposed multiscale dynamic convolutions extract detail feature of PAN image at different scales. Multiscale network structure is beneficial to obtain effective detail features. The weights obtained by the weight generation network are used to adjust the relationship among the detail features in each scale. The GeoEye-1, QuickBird, and WorldView-3 data are used to evaluate the performance of the proposed method. Compared with the widely used state-of-the-art pan-sharpening approaches, the experimental results demonstrate the superiority of the proposed method in terms of both objective quality indexes and visual performance.

Segmentation of fetal left ventricle (LV) in echocardiographic sequences is important for further quantitative analysis of fetal cardiac function. However, image gross inhomogeneities and fetal random movements make the segmentation a challenging problem. In this paper, a dynamic convolutional neural networks (CNN) based on multiscale information and fine-tuning is proposed for fetal LV segmentation. The CNN is pretrained by amount of labeled training data. In the segmentation, the first frame of each echocardiographic sequence is delineated manually. The dynamic CNN is fine-tuned by deep tuning with the first frame and shallow tuning with the rest of frames, respectively, to adapt to the individual fetus. Additionally, to separate the connection region between LV and left atrium (LA), a matching approach, which consists of block matching and line matching, is used for mitral valve (MV) base points tracking. Advantages of our proposed method are compared with an active contour model (ACM), a dynamical appearance model (DAM), and a fixed multiscale CNN method. Experimental results in 51 echocardiographic sequences show that the segmentation results agree well with the ground truth, especially in the cases with leakage, blurry boundaries, and subject-to-subject variations. The CNN architecture can be simple, and the dynamic fine-tuning is efficient.

With the development of Internet technology, network security is under diverse threats. In particular, attackers can spread malicious uniform resource locators (URL) to carry out attacks such as phishing and spam. The research on malicious URL detection is significant for defending against these attacks. However, there are still some problems in the current research. For instance, malicious features cannot be extracted efficiently. Some existing detection methods are easy to evade by attackers. We design a malicious URL detection model based on a dynamic convolutional neural network (DCNN) to solve these problems. A new folding layer is added to the original multilayer convolution network. It replaces the pooling layer with the k-max-pooling layer. In the dynamic convolution algorithm, the width of feature mapping in the middle layer depends on the vector input dimension. Moreover, the pooling layer parameters are dynamically adjusted according to the length of the URL input and the depth of the current convolution layer, which is beneficial to extracting more in-depth features in a wider range. In this paper, we propose a new embedding method in which word embedding based on character embedding is leveraged to learn the vector representation of a URL. Meanwhile, we conduct two groups of comparative experiments. First, we conduct three contrast experiments, which adopt the same network structure and different embedding methods. The results prove that word embedding based on character embedding can achieve higher accuracy. We then conduct the other three experiences, which use the same embedding method proposed in this paper and use different network structures to determine which network is most suitable for our model. We verify that the model designed in this paper has the highest accuracy (98%) in detecting malicious URL through these experiences.

The mainstream methods for change detection in synthetic-aperture radar (SAR) images use difference images to define the initial change regions. However, methods can suffer from semantic collapse, which makes it difficult to determine semantic information about the changes. In this paper, we proposed a hierarchical fusion SAR image change-detection model based on hierarchical fusion conditional random field (HF-CRF). This model introduces multimodal difference images and constructs the fusion energy potential function using dynamic convolutional neural networks and sliding window entropy information. By using an iterative convergence process, the proposed method was able to accurately detect the change-detection regions. We designed a dynamic region convolutional semantic segmentation network with a two-branch structure (D-DRUNet) to accomplish feature fusion and the segmentation of multimodal difference images. The proposed network adopts a dual encoder–single decoder structure where the baseline is the UNet network that utilizes dynamic convolution kernels. D-DRUNet extracts multimodal difference features and completes semantic-level fusion. The Sobel operator is introduced to strengthen the multimodal difference-image boundary information and construct the dynamic fusion pairwise potential function, based on local boundary entropy. Finally, the final change result is stabilized by iterative convergence of the CRF energy potential function. Experimental results demonstrate that the proposed method outperforms existing methods in terms of the overall number of detection errors, and reduces the occurrence of false positives.

In this paper, a multipath convolutional neural network (MP-CNN) is proposed for rehabilitation exercise recognition using sensor data. It consists of two novel components: a dynamic convolutional neural network (D-CNN) and a state transition probability CNN (S-CNN). In the D-CNN, Gaussian mixture models (GMMs) are exploited to capture the distribution of sensor data for the body movements of the physical rehabilitation exercises. Then, the input signals and the GMMs are screened into different segments. These form multiple paths in the CNN. The S-CNN uses a modified Lempel–Ziv–Welch (LZW) algorithm to extract the transition probabilities of hidden states as discriminate features of different movements. Then, the D-CNN and the S-CNN are combined to build the MP-CNN. To evaluate the rehabilitation exercise, a special evaluation matrix is proposed along with the deep learning classifier to learn the general feature representation for each class of rehabilitation exercise at different levels. Then, for any rehabilitation exercise, it can be classified by the deep learning model and compared to the learned best features. The distance to the best feature is used as the score for the evaluation. We demonstrate our method with our collected dataset and several activity recognition datasets. The classification results are superior when compared to those obtained using other deep learning models, and the evaluation scores are effective for practical applications.

Three-dimensional (3D) point cloud classification methods based on deep learning have good classification performance; however, they adapt poorly to diverse datasets and their classification accuracy must be improved. Therefore, FSDCNet, a neural network model based on the fusion of static and dynamic convolution, is proposed and applied for multiview 3D point cloud classification in this paper. FSDCNet devises a view selection method with fixed and random viewpoints, which effectively avoids the overfitting caused by the traditional fixed viewpoint. A local feature extraction operator of dynamic and static convolution adaptive weight fusion was designed to improve the model’s adaptability to different types of datasets. To address the problems of large parameters and high computational complexity associated with the current methods of dynamic convolution, a lightweight and adaptive dynamic convolution operator was developed. In addition, FSDCNet builds a global attention pooling, integrating the most crucial information on different view features to the greatest extent. Due to these characteristics, FSDCNet is more adaptable, can extract more fine-grained detailed information, and can improve the classification accuracy of point cloud data. The proposed method was applied to the ModelNet40 and Sydney Urban Objects datasets. In these experiments, FSDCNet outperformed its counterparts, achieving state-of-the-art point cloud classification accuracy. For the ModelNet40 dataset, the overall accuracy (OA) and average accuracy (AA) of FSDCNet in a single view reached 93.8% and 91.2%, respectively, which were superior to those values for many other methods using 6 and 12 views. FSDCNet obtained the best results for 6 and 12 views, achieving 94.6%, 93.3%, 95.3%, and 93.6% in OA and AA metrics, respectively. For the Sydney Urban Objects dataset, FSDCNet achieved an OA and F1 score of 81.2% and 80.1% in a single view, respectively, which were higher than most of the compared methods. In 6 and 12 views, FSDCNet reached an OA of 85.3% and 83.6% and an F1 score of 85.5% and 83.7%, respectively.

To maintain a high-quality machining process, it is necessary to develop an online cutting temperature prediction model accurately. However, using the conventional cutting temperature modeling methods, it is difficult to capture the continuous and multivariate time-variant characteristic of cutting temperature. Recently, local learning and switching dynamics have been employed to identify the time-variant characteristic. However, it requires quantitative and precise analysis with certain switching patterns, which are hard to realize in the machining process. In this article, a self-adaptive local learning method integrated with dynamic convolutional neural networks (DCNN) is proposed. The moving window strategy and t-test analysis of performance maps are first to give the local domains. Then, a DCNN is designed to capture spatial cross-correlations and temporal autocorrelations in each local domain. Finally, the local models are ensemble using Bayesian ensemble learning. The effectiveness of the proposed method is demonstrated through different process information and tool conditions.

The driving speed of bearing in rotating machines is usually variable rather than constant, so methods for accurate fault diagnosis under varying speed condition is required. In this paper, we propose the fault diagnosis model using dynamic convolutional neural network (DY-CNN) that considers variation of fault frequency characteristics by utilizing content-adaptive kernels for fault diagnosis of bearing under varying speed condition. As the input of model, 1-second intervals of vibration data with varying speed condition were used. These kernels adapt to short interval vibration data with varying speed condition by applying weighted sum of trained basis kernels. DY-CNN-based fault diagnosis model improved diagnosis accuracy by 7.07% compared to CNN-based fault diagnosis model. In addition, we showed that the adaptive kernels changed depending on fault types. DY-CNN-based fault diagnosis model adapted itself to fault types, and it performed accurate and robust fault diagnosis of ball bearing under varying speed condition.