Dual Network Research Articles

SCANeXt: Enhancing 3D medical image segmentation with dual attention network and depth-wise convolution

Existing approaches to 3D medical image segmentation can be generally categorized into convolution-based or transformer-based methods. While convolutional neural networks (CNNs) demonstrate proficiency in extracting local features, they encounter challenges in capturing global representations. In contrast, the consecutive self-attention modules present in vision transformers excel at capturing long-range dependencies and achieving an expanded receptive field. In this paper, we propose a novel approach, termed SCANeXt, for 3D medical image segmentation. Our method combines the strengths of dual attention (Spatial and Channel Attention) and ConvNeXt to enhance representation learning for 3D medical images. In particular, we propose a novel self-attention mechanism crafted to encompass spatial and channel relationships throughout the entire feature dimension. To further extract multiscale features, we introduce a depth-wise convolution block inspired by ConvNeXt after the dual attention block. Extensive evaluations on three benchmark datasets, namely Synapse, BraTS, and ACDC, demonstrate the effectiveness of our proposed method in terms of accuracy. Our SCANeXt model achieves a state-of-the-art result with a Dice Similarity Score of 95.18% on the ACDC dataset, significantly outperforming current methods.

Deep learning assisted solar forecasting for battery swapping stations

ABSTRACT For the development of renewable energy generation as a viable alternative to fossil fuel-based generation, solar power has received widespread acceptance. An increase in solar power usage can mitigate the impact of climate change, energy demand, and cost-effective dispatch. To use solar power efficiently and ensure grid consistency, reliable and accurate forecasting of information becomes very crucial. Hence, the main purpose of this work is to get the best suitable solar irradiance forecasting model and to implement the forecasted solar power to the designed hybrid battery swapping stations. Each year, new techniques and approaches are used to improve the model accuracy with the critical aim of lowering the variability of predictions. This paper takes a close look at different deep-learning methods to predict solar irradiance for a certain period of time. The real-time time series data is used, and for analyzing the time series data, the statistical ARIMA model, LSTM-based RNN technique, and Dual Attention-based Recurrent Neural network have been used. These models are implemented using Jupyter Notebook with Python programming language. To analyze the efficiency and evaluate the performance of real-time data of the models, a comparative study has been done on the different error metrics. This paper reveals that the LSTM model is providing the best solar power forecasting with the least error metrics like MSE of .0091, MAE of .0525, RMSE of .0953, and MSLE of .0047.

Dual Cross-Linked Stimuli-Responsive Alginate-Based Hydrogels.

Sodium alginate with different molecular weights (55, 170, and 320 kg mol-1) were chemically modified by grafting methacrylic moieties onto the hydroxyl groups of the alginate backbone. The methacrylation was optimized to obtain different degrees of modification. Chemically cross-linked hydrogels were obtained following UV-light irradiation in the presence of a photoinitiator. The swelling behavior and the mechanical properties were observed to depend on both the degree of methacrylation and the alginate molecular weight. Due to the chain entanglement present in high-viscosity sodium alginate, lower degrees of modification were required to tune the hydrogel properties. Moreover, in the presence of Ca2+, secondary cross-linking was introduced by the coordination of the alginate guluronate moieties with the Ca2+ ions. The addition of this secondary cross-linking caused fast volume shrinkage and a reinforcement of the mechanical properties. The secondary cross-linking was reversible, and the hydrogels regained their original shape for at least three cycles. Additionally, the dual cross-linked network can be used to induce adhesion between hydrogels and serve as a building block for self-folding actuators.

PDMS/functionalized h-BN/liquid metal flexible composite with dual-network structure for achieving superior thermal conductivity and low thermal resistance via interfacial grafting

Thermal interface materials (TIMs) play a crucial role in addressing the heat dissipation challenges associated with high-power electrical equipment. Apart from possessing excellent thermal conductivity, and they must also exhibit low contact thermal resistance to facilitate efficient heat transfer at the interface. In this study, a novel composite with a dual-network structure was prepared by interfacial grafting to achieve superior thermal conductivity. The composite comprises polydimethylsiloxane (PDMS), functionalized boron nitride (f-BN), and liquid metal (LM). This dual network structure effectively enhances phonon transportation and significantly improves the thermal transport performance of the composite. The thermal conductivity of the obtained composite can achieve 4.4 W/m·K, accompanied by a low contact thermal resistance of 0.32 °C·cm2/W. Furthermore, the dual network structure ensures the flexibility of the composite, as evidenced by the maximum elongation reaching 267.43% at a tensile strength of 0.35 MPa. This exceptional mechanical performance is expected to alleviate stress concentration issues arising from thermal expansion mismatch, thereby enhancing the practical applicability of the prepared TIMs. Overall, this study introduces a novel approach for fabricating flexible composites with high thermal conductivity and low contact thermal resistance, offering new insights into the preparation of TIMs.

Dual contrastive attention-guided deformable convolutional network for single image super-resolution

With its powerful ability to model geometric transformations, the deformable convolutional network brings great improvements for single image super-resolution (SISR). Nevertheless, its location-variant sampling method leads to an escalation in spatial variance as the deformable convolutional layers are stacked, consequently resulting in limited performance. Hence, we propose a novel and effective approach called dual contrastive attention-guided deformable convolutional network (DCADCN) for SISR modeling. Specifically, we propose an attention-guided deformable convolutional module with joint inner and external attention mechanisms to fully exploit the correspondences between input and deformation features and preserve spatial characteristics to the extent possible. Additionally, we propose a dual mixed feature extractor consisting of two parallel sub-paths. This design allows for the learning of diverse and complementary spatial features. Furthermore, contrastive learning is applied to further amplify the role of key features and mitigate the interference of noisy features. Extensive experimental results demonstrate that DCADCN is capable of effectively handling classic SISR, SISR with blind noise, and real-world SISR tasks. Moreover, our method achieves comparable or even better performance with lower computational cost compared to state-of-the-art methods.

Dual branch deep learning network for detection and stage grading of diabetic retinopathy

Diabetic retinopathy is a severe complication of diabetes that can lead to permanent blindness if not treated promptly. Early and accurate diagnosis of the disease is essential for successful treatment. This paper introduces a deep learning method for the detection and stage grading of diabetic retinopathy, using a single fundus retinal image. Our model utilizes transfer learning, employing two state-of-the-art pre-trained models as feature extractors and fine-tuning them on a new dataset. The proposed model is trained on a large multi-center dataset, including the APTOS 2019 dataset, obtained from publicly available sources. It achieves remarkable performance in diabetic retinopathy detection and stage classification on the APTOS 2019, outperforming the established literature. For binary classification, the proposed approach achieves an accuracy of 98.50%, a sensitivity of 99.46%, and a specificity of 97.51%. In stage grading, it achieves a quadratic weighted kappa of 93.00%, an accuracy of 89.60%, a sensitivity of 89.60%, and a specificity of 97.72%. The proposed approach serves as a reliable screening and stage grading tool for diabetic retinopathy, offering significant potential to enhance clinical decision-making and patient care.

Effect of polydiborosiloxane on the mechanical properties of silicone rubber foam with dual network structure

AbstractA silicone rubber foam is prepared by introducing polydiborosiloxane (PDBS) into silicone rubber, and using expandable microspheres as blowing agents. The influences of the dual network structure with dynamic cross‐linking network on the silicone rubber foam is characterized by the cell size, mechanical properties and impact resistance. The cell size increases from 77 to 93 μm due to the introduction of a dynamic cross‐linking network which reduces foaming resistance. The elongation at break elevates from 277% to 574%, and the yield strength gradually decreases. The dynamic cross‐linking network also enhances the internal damping characteristic and energy absorption capacity of the foam. The impact resistance of silicone rubber foam is improved because of the synergistic effect of the dual network structure and the bubble structure. The research improves the performance of silicone rubber foam.

PCTC-Net: A Crack Segmentation Network with Parallel Dual Encoder Network Fusing Pre-Conv-Based Transformers and Convolutional Neural Networks.

Cracks are common defects that occur on the surfaces of objects and structures. Crack detection is a critical maintenance task that traditionally requires manual labor. Large-scale manual inspections are expensive. Research has been conducted to replace expensive human labor with cheaper computing resources. Recently, crack segmentation based on convolutional neural networks (CNNs) and transformers has been actively investigated for local and global information. However, the transformer is data-intensive owing to its weak inductive bias. Existing labeled datasets for crack segmentation are relatively small. Additionally, a limited amount of fine-grained crack data is available. To address this data-intensive problem, we propose a parallel dual encoder network fusing Pre-Conv-based Transformers and convolutional neural networks (PCTC-Net). The Pre-Conv module automatically optimizes each color channel with a small spatial kernel before the input of the transformer. The proposed model, PCTC-Net, was tested with the DeepCrack, Crack500, and Crackseg9k datasets. The experimental results showed that our model achieved higher generalization performance, stability, and F1 scores than the SOTA model DTrC-Net.

Human-robot collaborative disassembly enabled by brainwaves and improved generative adversarial network

Human-robot collaboration (HRC) can greatly facilitate the disassembling processes of end-of-life (EoL) products. For the robot in HRC, an intuitive control function enabled by a brain-machine interface (BMI) and human brainwaves will be useful to support dynamic decision-making under various disassembly conditions. However, a major challenge in developing a BMI-enabled HRC disassembly system is that there are usually a limited number of brainwave signals available for system training. To address the challenge, a novel model, namely TCN-D2GAN, was designed in this study. The model was used to augment acquired brainwave signals and support the establishment of the BMI-enabled HRC disassembly system. The innovations of the TCN-D2GAN model come from the following aspects: (i) the strengths of a temporal convolutional network (TCN) and a dual discrimination generative adversarial network (D2GAN) were combined to accomplish high-performance signal augmentation, and (ii) new loss functions were defined in the model to avoid the use of hyperparameters, so the training effect of the model can be optimised. Based on two public electroencephalography (EEG) datasets, experiments for signal augmentation and system training were carried out. The TCN-D2GAN model was compared with several mainstream generative adversarial network (GAN) models. Results indicated that the TCN-D2GAN model exhibited better performance in terms of signal classification accuracy and training efficiency. Finally, disassembly experiments were presented to validate the applicability of the system.

Complex CFRP damage detection with dual multiscale 1DCNNs and self-selective fusion technique of guided waves

This article presents an innovative damage detection method, named DMS1DCNNs, based on new dual multiscale one-dimensional convolutional neural networks (MS1DCNNs) and a novel self-selective fusion technique for damage localization and quantification in complex carbon fiber reinforced plastic (CFRP) structures. Firstly, a new multiscale piecewise aggregate approximation algorithm (MIPA) is proposed to compress the guided wave (GW) signals and extract damage features at various time scales. Considering that signals from low-information sensing paths may interfere with the diagnostic performance of the model, a new method for self-selective fusion of GW sensing paths is proposed. This method can automatically fuse paths with more damage information and remove paths with less damage information. Then, a new DMS1DCNNs model is presented, capable of obtaining multiscale damage features in GW signals from both data and network levels. The model takes GW data as input and employs regression methods for direct prediction of the damage location and size. Finally, the method is validated using the GW testbed, and the mean relative error of location on complex CFRP is 8.32 % and the mean absolute error of quantification is 1.20 mm, superior to other state-of-the-art methods.

Incorporating syntax and semantics with dual graph neural networks for aspect-level sentiment analysis

Aspect-level sentiment analysis is a more fine-grained task that aims to determine the sentiment polarity of specific aspects. Recent studies have employed graph attention networks and graph convolutional networks to model dependency trees, effectively establishing explicit associations between aspects and opinions, yielding promising performance. However, these methods have limitations in capturing complex linguistic features and the intricate dependencies between aspects and their contexts, resulting in suboptimal performance. In this paper, we propose a dual graph neural network that incorporates syntax and semantics, called IDGNN. Specifically, we utilize the relational graph attention network (RGAT) to encode the syntactic dependency tree and obtain syntactic information, while incorporating dependency labels to enhance aspect representation. Additionally, the semantic graph convolutional network (SemGCN) is employed to encode the self-attention matrix and capture semantic information, with the inclusion of orthogonal regularization to enhance semantic association. Furthermore, we introduce two fusion strategies based on gate mechanisms: the syntax fusion module (SYF) and the semantic fusion module (SEF). SYF combines contextual and syntactic representations to obtain global syntactic features, while SEF fuses semantic information with global syntactic features to obtain the final feature representation. Experimental results demonstrate that our proposed model achieves state-of-the-art performance on several benchmark datasets.

A temperature‐responsive dual network hydrogel for reversible smart actuator

AbstractNowadays, soft actuators have received extensive attention in many application fields, among which hydrogels have become an important choice for constructing soft actuators due to their unique properties. However, the actuating behaviors of hydrogel‐based actuators are usually monotonous due to their unchangeable shapes and structures. Herein, we report a temperature‐responsive hydrogel actuator with a bilayer structure. Based on the dual network structure (polyvinyl alcohol/poly acrylamide and polyvinyl alcohol/poly (N‐isopropylacrylamide), the actuators can realize the reinforcement compared with the single network. Because of the intrinsic lower critical solution temperature of poly (N‐isopropylacrylamide, both sides of actuators have different swelling rates, enabling them to achieve the thermal‐responsive actuation and shape programming. Therefore, this work is promising to provide a new strategy for designing temperature switches and thermally driven soft robots.

Dual fusion network for semantic segmentation of point clouds

The current networks for point cloud segmentation fail to fully consider the connections between points in each neighborhood, which makes the final segmentation capability underperform. To obtain richer local relationships between points, this paper proposes a network called GDFNet (Graph Convolution-based Dual Fusion Network), which dynamically mines the correlations between points in each layer of the network. GDFNet is divided into two parts: a density-sensitive assignment module and a spatial dynamic convolution module. The density-sensitive assignment module treats the traditional convolution kernel as a nonlinear function consisting of a weight and a density function applicable to the local coordinates of 3D points. For a given point, the weight function is learned by a multilayer perceptron network, and the density function is learned by kernel density estimation. In this way, the density features of each point are fitted, thus improving the inherent defect of the original point cloud with non-uniform density. The spatial dynamic convolution module groups the original feature channels and then uses gated processing units to enrich the feature representation of the input points. At the same time, dynamic convolution is embedded in the Euclidean space to obtain the structural information of each point and further supplement the local detail features. Extensive experiments demonstrate the effectiveness and superiority of our method on S3DIS and ScanNetv2 datasets.

Adaptive Dual Aggregation Network with Normalizing Flows for Low-Light Image Enhancement.

Low-light image enhancement (LLIE) aims to improve the visual quality of images taken under complex low-light conditions. Recent works focus on carefully designing Retinex-based methods or end-to-end networks based on deep learning for LLIE. However, these works usually utilize pixel-level error functions to optimize models and have difficulty effectively modeling the real visual errors between the enhanced images and the normally exposed images. In this paper, we propose an adaptive dual aggregation network with normalizing flows (ADANF) for LLIE. First, an adaptive dual aggregation encoder is built to fully explore the global properties and local details of the low-light images for extracting illumination-robust features. Next, a reversible normalizing flow decoder is utilized to model real visual errors between enhanced and normally exposed images by mapping images into underlying data distributions. Finally, to further improve the quality of the enhanced images, a gated multi-scale information transmitting module is leveraged to introduce the multi-scale information from the adaptive dual aggregation encoder into the normalizing flow decoder. Extensive experiments on paired and unpaired datasets have verified the effectiveness of the proposed ADANF.

Dual cerebella model neural networks based robust adaptive output feedback control for electromechanical actuator with anti‐parameter perturbation and anti‐disturbance

SummaryTo realize a high‐accuracy tracking control of an electromechanical actuator in which only position signal is available, a new robust adaptive output feedback control strategy based on dual CMAC neural networks is proposed in this article. A high‐gain observer and a neural network are combined to estimate the unmeasured system states, in which a neural network is designed to estimate and compensate the system parameter estimation error and other uncertain nonlinearity to improve the performance of the traditional observer. A robust adaptive controller and another neural network are combined to decrease the impact of parameter perturbation and external disturbance and strengthen the system robustness. Lyapunov theorem is used to derive an on‐line weight adaption law for the two neural networks and an on‐line parameter adaptation law for the uncertain parameters to stabilize the system. Thus the parameter uncertainty and disturbance can be compensated with feedforward compensation technique. A nonlinear robust feedback term is designed to suppress the model compensation deviation. Considering the stronger approximation ability of CMAC basis function against RBF basis function, an improved CMAC neural network is introduced in the proposed controller. A large number of simulations and experiments show that the proposed controller has better tracking performance than many main‐stream output feedback controllers in present.

A Shear-Stiffening Mouthguard with Excellent Shock Absorption Capability and Remoldability via a Dynamic Dual Network.

Mouthguards are used to reduce injuries and the probability of them to orofacial tissues when impacted during sports. However, the usage of a mouthguard is low due to the discomfort caused by the thickness of the mouthguard. Herein, we have constructed a dynamic dual network to fabricate a shear-stiffening mouthguard with remoldability, which are called remoldable shear-stiffening mouthguards (RSSMs). Based on diboron/oxygen dative bonds, RSSMs show a shear-stiffening effect and excellent shock absorption ability, which can absorb more than 90% of the energy of a blank. Even reducing the thickness to half, RSSMs can reduce approximately 25% of the transmitted force and elongate by about 1.6-fold the buffer time compared to commercial mouthguard materials (Erkoflex and Erkoloc-pro). What is more, owing to the dynamic dual network, RSSMs show good remolding performance with unchanged shear-stiffening behavior and impact resistance, which conforms to the existing vacuum thermoforming mode. In addition, RSSMs exhibit stability in artificial saliva and biocompatibility. In conclusion, this work will broaden the range of mouthguard materials and offer a platform to apply shear-stiffening materials to biomedical applications and soft safeguarding devices.

Dual–network carboxymethyl chitosan conductive hydrogels for multifunctional sensors and high–performance triboelectric nanogenerators

With the development of technology, there is a growing demand for wearable electronics that can fulfill different application scenarios. Hydrogel–based sensors are considered ideal candidates for realizing multifunctional wearable flexible devices. However, there are great challenges in preparing hydrogel–based sensors with both superior mechanical and electrical properties. Herein, we report a composite conductive hydrogel prepared by using a dynamically crosslinked carboxymethyl chitosan network and a covalently crosslinked polymer network, and carboxylated carbon nanotubes as conductive filler. The carboxymethyl chitosan–based hydrogels had excellent mechanical properties and strength (tensile strength of 475.4 kPa, and compressive strength of 1.9 MPa) and ultra–high conductivity (0.19 S·cm−1). Based on the above characteristics, the hydrogel could accurately identify the movement signals of the human body and different writing signals, and achieve encrypted transmission of signals, broadening the application scenarios. In addition, a triboelectric nanogenerator (TENG) was fabricated based on the hydrogel, which had an outstanding output performance with open–circuit voltage of 336 V, short–circuit current of 18 μA, transferred charge of 52 nC and maximum power density of 340 mW·m−2, and could power small devices. This work is expected to provide new ideas for the development of self–powered, multi–functional wearable, and flexible polysaccharide–based devices.

Durable and sustainable CoFe2O4@MXene-silver nanowires/cellulose nanofibers composite films with controllable electric–magnetic gradient towards high-efficiency electromagnetic interference shielding and Joule heating capacity

Exploring electromagnetic waves (EMW) absorption-dominated high-performance electromagnetic interference (EMI) shielding materials with robust mechanical properties and remarkable Joule heating capability is highly desired but remains a daunting challenge for the development of modern electronic equipment. Herein, durable and multifunctional cellulose nanofiber-based composite films with elaborate electric–magnetic dual gradient structure were successfully fabricated through highly controllable vacuum-assisted filtration technology. The electric–magnetic dual gradient network consists of CoFe2O4@MXene hybrids with electric–magnetic cooperative loss concentrated on the top region as impedance matching layer and silver nanowires (AgNWs) with high conductivity placed on the bottom as high-efficiency shielding layer. Benefiting from the construction of dual gradient network with rational arrangement of the impedance matching layer and conductive shielding layer, the synergy of electric–magnetic loss, polarization loss induced by the numerous heterointerfaces and “absorption-reflection-reabsorption” shielding mechanism, the resultant composite films simultaneously reveal ultraefficient EMI shielding effectiveness (SE) and low EMW reflectivity. The prepared composite films (0.1 mm in thickness) achieve an extraordinary EMI SE of 84.3 dB and a satisfactory reflection coefficient of 0.42, realizing EMW absorption-dominated shielding feature. Moreover, the mechanical properties and Joule heating capacity of the prepared composite films are outstanding, which mainly benefits from the construction of hierarchically layered structure. This work paves a novel route for designing and fabricating absorption-dominated high-performance EMI shielding composite films with durable mechanical performance and outstanding Joule heating capability, which are potentially suitable for new-generation advanced electronics.

Multi-knowledge enhanced graph convolution for learning resource recommendation

In recent years, E-learning has gained immense popularity as a prominent mode of education. However, accurately recommending learning resources from a vast amount of data remains a significant challenge. This study addresses two primary challenges impacting recommendation performance. Firstly, an imbalance exists between the abundance of available learning resources and the limited interaction behavior of learners. Secondly, the existing algorithms often overlook dynamic preference information, focusing primarily on learners’ short-term, static preferences only by learning interactive behavior but disregarding the multi-correlation between learning resources and learners. To tackle these challenges, we propose MkEGC (Multi-knowledge Enhanced Graph Convolution), a novel framework for learning resource recommendation. We approach the recommendation process as a Markov decision process. Initially, we construct a dual knowledge graph convolutional network, operating in learning resource-knowledge and learner-knowledge domains. This network facilitates the extraction of vector features from learning resources, enhances the learner vector representation, and captures higher-order preferences. Subsequently, we design hierarchical and attention weighting strategies to effectively extract latent hierarchical information from the knowledge graph. Finally, we integrate the learning resource state, the learning interaction state, and the sequence state to represent a multi-dimensional learner-state within the Markov decision framework, enabling prise learning resource recommendations. To validate the effectiveness of MkEGC, we conduct extensive experiments, comparing multiple sets of metrics with six state-of-the-art recommendation algorithms, utilizing two real-world datasets.

Robust Classification and 6D Pose Estimation by Sensor Dual Fusion of Image and Point Cloud Data

It is an important aspect to fully leverage complementary sensors of images and point clouds for objects classification and six-dimensional (6D) pose estimation tasks. Prior works extract objects category from a single sensor such as RGB camera or LiDAR, limiting their robustness in the event that a key sensor is severely blocked or fails. In this work, we present a robust objects classification and 6D object pose estimation strategy by dual fusion of image and point cloud data. Instead of solely relying on 3D proposals or mature 2D object detectors, our model deeply integrates 2D and 3D information of heterogeneous data sources by a robustness dual fusion network and an attention-based nonlinear fusion function Attn-fun(.), achieving efficiency as well as high accuracy classification for even missed some data sources. Then, our method is also able to precisely estimate the transformation matrix between two input objects by minimizing the feature difference to achieve 6D object pose estimation, even under strong noise or with outliers. We deploy our proposed method not only to ModelNet40 datasets but also to a real fusion vision rotating platform for tracking objects in outer space based on the estimated pose.