Search

Recently, image inpainting techniques tend to be more concerned with how to enhance the quality of restoration than with how to function on various platforms with limited processing power. In this paper, we propose a lightweight method that combines group convolution and attention mechanism to improve or replace the traditional convolution module. Group convolution was used to achieve multi-level image inpainting, and the authors proposed the rotating attention mechanism for allocation to deal with the issue of information mobility between channels in traditional convolution processing. The parallel discriminator structure was utilized throughout the network's overall design phase to guarantee both local and global consistency of the image inpainting process. The experimental results can demonstrate that, while the quality of image inpainting has been ensured, the proposed image inpainting network's inference time and resource usage are significantly lower than those of comparable lightweight approaches. [ABSTRACT FROM AUTHOR]

The rate of wind energy shows a sharp increase in recent years. Due to the instability, it is necessary to accurately predict the value of wind power over a period of time. Wind turbine cluster power prediction is an emerging form of wind power prediction in recent years, which can simultaneously predict the wind power value in a region. Recently, many researchers have utilized convolutional neural networks to process spatio-temporal representation constructed by multi-turbines' data. However, the research to date has tended to focus on local spatial scale rather than global spatio-temporal scale. The wind has instability, drastic changes, and wake effects, but local features cannot cover the affected area in the entire wind propagation process. To address this problem, this paper proposes a Spatio-temporal Graph Cross-correlation Auto-encoding Network (STGCAN) for wind power prediction, which can extract global wind power features and improve the problem that convolution and its derived methods only extract local features. The STGCAN performs hierarchical feature extraction on wind turbine clusters through the hierarchical structure of the shallow spatio-temporal feature layer, the feature cohesion layer, as well as the spatio-temporal graph cross-correlation auto-encoding layer altogether. It can extract both the local and global features of wind power. Experiments show that the mean square error of wind power prediction of the new approach is on average reduced by 20.87% over all the baseline methods. [ABSTRACT FROM AUTHOR]

Accurate power load forecasting is crucial for the sustainable operation of smart grids. However, the complexity and uncertainty of load, along with the large-scale and high-dimensional energy information, present challenges in handling intricate dynamic features and long-term dependencies. This paper proposes a computational approach to address these challenges in short-term power load forecasting and energy information management, with the goal of accurately predicting future load demand. The study introduces a hybrid method that combines multiple deep learning models, the Gated Recurrent Unit (GRU) is employed to capture long-term dependencies in time series data, while the Temporal Convolutional Network (TCN) efficiently learns patterns and features in load data. Additionally, the attention mechanism is incorporated to automatically focus on the input components most relevant to the load prediction task, further enhancing model performance. According to the experimental evaluation conducted on four public datasets, including GEFCom2014, the proposed algorithm outperforms the baseline models on various metrics such as prediction accuracy, efficiency, and stability. Notably, on the GEFCom2014 dataset, FLOP is reduced by over 48.8%, inference time is shortened by more than 46.7%, and MAPE is improved by 39%. The proposed method significantly enhances the reliability, stability, and cost-effectiveness of smart grids, which facilitates risk assessment optimization and operational planning under the context of information management for smart grid systems. [ABSTRACT FROM AUTHOR]

Online chatter detection is crucial to ensure the quality and safety of the high-speed milling process. The rapid development of the deep learning community provides a promising tool for chatter detection. However, most previous chatter detection studies rely on complex signal processing techniques, leading to the separation of feature extraction and chatter detection and reducing detection efficiency. Additionally, these studies are developed for a limited range of machining conditions because the development of their model relies on experimental data, while performing experiments with numerous combinations of machining parameters is expensive and time-consuming. To tackle these drawbacks, this paper proposes a transfer learning chatter detection framework that doesn't rely on any experimental data. The proposed framework is composed of the dynamic milling process model, the Double Attention-driven One-Dimension Convolutional Neural Networks (DAOCNN), and the ensemble prediction strategy. Firstly, a dynamic milling process model is established to generate simulated cutting force signals over a wide range of machining parameters, providing abundant training data and saving experimental costs. Then, without any complex signal processing method, the detection results are directly predicted by the proposed DAOCNN from the time-domain signals, eliminating the separation of feature extraction and chatter detection. Finally, a novel ensemble prediction strategy is proposed to ensure an accurate and robust prediction. To validate the effectiveness of the proposed framework, actual milling experiments are carried out under different cutting conditions. Moreover, contrastive studies with other detection approaches and ensemble methods are also performed. The results demonstrate that the milling stability is correctly predicted by the proposed method in an accurate and efficient manner, which indicates the proposed framework can be a promising tool for online chatter detection. [ABSTRACT FROM AUTHOR]

Real-time semantic segmentation poses a significant challenge in scene parsing. Despite traditional semantic segmentation networks have made remarkable leap-forwards in semantic accuracy, the performance of inference speed remains unsatisfactory. This paper introduces the Cross-CBAM network, a novel lightweight architecture designed for real-time semantic segmentation. Specifically, a Squeeze-and-Excitation Atrous Spatial Pyramid Pooling Module (SE-ASPP) is proposed to obtain variable field-of-view and multiscale information. Additionally, we propose a Cross Convolutional Block Attention Module (CCBAM), wherein a cross-multiply operation guides low-level detail information with high-level semantic information. Unlike previous approaches that leverage attention to concentrate on the relevant information in the backbone, CCBAM utilizes cross-attention for feature fusion within the Feature Pyramid Network (FPN) structure. Extensive experiments on the Cityscapes dataset and Camvid dataset demonstrate the effectiveness of the proposed Cross-CBAM model by achieving a promising trade-off between segmentation accuracy and inference speed. On the Cityscapes test set, we achieve 73.4% mIoU with a speed of 240.9 FPS and 77.2% mIoU with a speed of 88.6 FPS on NVIDIA GTX 1080Ti. [ABSTRACT FROM AUTHOR]

Heat load prediction is essential to discover blast furnace (BF) anomalies in time and take measures in advance to reduce erosion in the ironmaking process. However, owing to the redundancy of the high dimensional data and the multi-granularity features of the state monitoring data, the general prediction model is hard to accurately predict the heat load, especially the rapid change caused by physical and chemical reactions. Therefore, this paper puts forward an attention-based one-dimension convolution neural network (1DCNN) combined with a bidirectional long short-term memory (BiLSTM) network for heat load prediction. Firstly, the two-stage data pre-processing realizes dimension reduction and key variable selection. Secondly, fine-grained features are extracted by 1DCNN, and the BiLSTM extracts the coarse-grained features for prediction output. Moreover, an attention branch is added to the 1DCNN to extract the critical fine-grained features when the heat load changes rapidly. Finally, experiments are carried out with the actual industrial data from a BF ironmaking process. The efforts show that the proposed prediction model presents better performances in the result of different metrics and has higher accuracy than the traditional prediction algorithms. [ABSTRACT FROM AUTHOR]

The prediction of power equipment activity plays a vital role in optimizing power resource dispatch, ensuring supply and demand balance, and guiding network planning and management. However, due to the complex nonlinear, multi-scale, and multivariate characteristics of power grid communication networks service activity (PCNSA) data, it is often challenging to capture its intrinsic patterns and dynamic changes through a simple model. To address this issue, this paper proposes a power grid communication network service activity prediction method based on convolutional neural network, bidirectional gated recurrent unit, and attention mechanism, referred to as CBGA. CNN is used to extract features from time-series data, BiGRU is used to capture long-term feature changes in the data, and the attention mechanism is used to enhance the extraction of important information. To validate the performance of the proposed method, experiments and comparisons were conducted on three real power grid communication network datasets, demonstrating that our method shows better performance and robustness. [ABSTRACT FROM AUTHOR]

In aerial image detection, difficulties in feature extraction and low detection accuracy arise due to background interference, occlusion, and the presence of multiple small objects. This paper proposes a DM-YOLOX aerial object target detection method with intensive attention mechanism. Firstly, the proposed approach incorporates coordinate attention (CA) and a dense connection method into the backbone network architecture, enabling adaptive channel weighting throughout the feature extraction process. This facilitates the enhancement of significant features while suppressing less relevant ones, thereby augmenting the network's capacity to represent object features and ensuring retention and reinforcement of key features. Secondly, the multibranch extraction module (MBE) is incorporated into the feature fusion network to enhance the network's ability in extracting multi-scale feature information from images with extensive coverage, thereby enhancing the detection accuracy and efficiency of small- and medium-sized objects in complex scenes. Finally, the utilization of SIoU instead of IoU as the bounding box loss function effectively addresses the issue of mismatch between real and predicted boxes, leading to accelerated network convergence and improved performance during model training. After training and testing on the VisDrone 2019 dataset, this method effectively detects small objects in complex environments. The DM-YOLOX model shows a significant improvement of 2.7% in mAP compared to the baseline network, while achieving an 8% increase in frames per second (FPS). [ABSTRACT FROM AUTHOR]

Transportation mode recognition is an important and challenging problem in intelligent transportation systems. For decades, many data mining and deep learning methods have been proposed, but all these methods have some deficiencies and do not fully use temporal and spatial information hidden in raw GPS trajectory data. In this paper, we presented a new deep learning model called TSANET, which designs an attention mechanism to combine global and local spatiotemporal features. The input of the model is seven kinematic features, derived from the GPS trajectories and covers all information hidden in the GPS points. The model adopts TCN and ST-Block to extract global and local spatiotemporal features, respectively. In ST-Block, BiGRU is used to capture temporal features and DenseNet is applied to capture spatial features. In TCN, convolution operations are applied to obtain temporal and spatial information. Moreover, a dual-layer attention mechanism is developed to integrate and reconstruct these features by assigning different weights to global and local spatiotemporal features caught by TCN and ST-Block. By comprehensively considering global and local features, the model can greatly improve classification performance and recognition granularity, and with the further introduction of the attention mechanism, it can identify all transportation modes in a fine-grained manner. At last, the experiments have been conducted on the Geo1 and Geo2 datasets used by many researchers. The experimental results show that our model cannot only achieve the highest accuracy of 94.34 % among all recent methods but also identify all seven transportation modes clearly, thus verifying the advantages and effectiveness of the model. [ABSTRACT FROM AUTHOR]

Semantic relatedness is one of the most significant aspects of natural language processing. It has been identified as a critical technology for developing intelligent systems like Siri, Microsoft Ice, Cortana, and Xiaoai. In 2014, SemEval ranked SR as the top task. While many existing studies have focused on analyzing the entailment of single phrases, advancements in deep learning have made it possible to analyze complete sentences or texts. While the natural parallelism of membrane computing has shown promise for data processing, harnessing this potential to advance semantic relatedness remains an open problem yet to be tackled. This paper proposes a novel Siamese Manhattan LSTM-SNP approach (SiMaLSTM-SNP) for the SR problem. The approach uses a collaborative Word2vec and 10-Layer Attention strategy to represent and extract sentence pairs and a Siamese LSTM-SNP structure to calculate the hidden states of sentences. The multi-head self-attention layer identifies text associations and redistributes hidden state weights. The last hidden state is extracted, and the relatedness score is calculated using the Manhattan distance. The experiments demonstrate that SiMaLSTM-SNP outperforms 17 classical SR baselines and 7 novel approaches on the standard datasets SICK and STS in terms of mean square error performance. This indicates that SiMaLSTM-SNP can accurately capture the semantic distinction between two sentences and effectively preserve their semantic information. [ABSTRACT FROM AUTHOR]

Parameter identification of transmission lines plays a crucial role in power systems, and many deep learning methods have been continuously applied to this domain. However, these methods are highly sensitive to data corruption, and as the scale of the power grid continues to expand, the model’s solving accuracy deteriorates. In response to these challenges, this paper introduces a multi-scale graph attention network (UMGAN) that leverages the spatial structural characteristics of the power grid. In order to enable the model to fully exploit the correlations between power grid branches and learn more precise node feature representations, we have designed a simplified graph neural network convolution kernel, creating a lighter U-shaped structure for multi-scale sampling of power grid data. Additionally, an attention layer has been incorporated at skip connections to reduce the impact of data corruption. Furthermore, for the multi-task learning module, we have devised a multi-task loss function tailored for power grid parameter identification tasks. This loss function effectively balances multiple objectives, allowing simultaneous identification of multiple parameters. Experimental results demonstrate that the multi-scale graph attention network proposed in this paper outperforms other machine learning and deep learning methods, providing more accurate predictions of power grid branch parameters. [ABSTRACT FROM AUTHOR]

Floor plan is an important communication tool between architects, construction engineers, and clients for a building project. Estimation of building features from a floor plan image is often a time-consuming task. Automatic analysis of floor plan images can significantly improve work efficiency and accuracy. A few research works have been reported in the literature on automated floor image analysis. However, the scope and performance of the existing techniques are limited. In this paper, a CNN-based technique, referred to as FloorNet, is proposed for the multiclass semantic segmentation of a floor plan. The proposed FloorNet has five modules: Encoder, Room type decoder, Room boundary decoder, Multiscale room boundary attention model and Floor classification. The proposed technique is evaluated using simple brochure type and complex architectural type floor plan images. Experimental results show that the proposed technique provides an improvement of 5–11% mIoU for semantic segmentation (for 9–11 classes) compared to the state-of-the-art techniques. [ABSTRACT FROM AUTHOR]

Since the impressive superior performance demonstrated by deep learning methods is widely used in histopathological image analysis and diagnosis, existing work cannot fully extract the information in the breast cancer images due to the limited high resolution of histopathological images. In this study, we construct a novel intermediate layer structure that fully extracts feature information and name it DMBANet, which can extract as much feature information as possible from the input image by up-dimensioning the intermediate convolutional layers to improve the performance of the network. Furthermore, we employ the depth-separable convolution method on the Spindle Structure by decoupling the intermediate convolutional layers and convolving them separately, to significantly reduce the number of parameters and computation of the Spindle Structure and improve the overall network operation speed. We also design the Spindle Structure as a multi-branch model and add different attention mechanisms to different branches. Spindle Structure can effectively improve the performance of the network, the branches with added attention can extract richer and more focused feature information, and the branch with residual connections can minimize the degradation phenomenon in our network and speed up network optimization. The comprehensive experiment shows the superior performance of DMBANet compared to the state-of-the-art method, achieving about 98% classification accuracy, which is better than existing methods. The code is available at https://github.com/Nagi-Dr/DMBANet-main. [ABSTRACT FROM AUTHOR]

Road extraction from remote-sensing images is of great significance for vehicle navigation and emergency insurance. However, the road information extracted in the remote-sensing image is discontinuous because the road in the image is often obscured by the shadows of trees or buildings. Moreover, due to the scale difference of roads in remote-sensing images, it remains a computational challenge to extract small-size roads from remote-sensing images. To address those problems, we propose a road extraction method based on adaptive global feature fusion (AGF-Net). First, a dilated convolution strip attention (DCSA) module is designed from the encoder–decoder structure. It consists of the dilated convolution and the strip attention module, which adaptively emphasizes relevant features in vertical and horizontal directions. Then, multiple global feature fusion modules (GFFM) in the skip connection are designed to supplement the decoder with road detail features, and we design a multi-scale strip convolution module (MSCM) to implement the GFFM module to obtain multi-scale road information. We compare AGF-Net to state-of-the-art methods and report their performance using standard evaluation metrics, including Intersection over Union (IoU), F1-score, precision, and recall. Our proposed AGF-Net achieves higher accuracy compared to other existing methods on the Massachusetts Road Dataset, DeepGlobe Road Dataset, CHN6-CUG Road Dataset, and BJRoad Dataset. The IoU obtained on these datasets are 0.679, 0.673, 0.567, and 0.637, respectively. [ABSTRACT FROM AUTHOR]

As a practical and challenging task, deep learning-based methods have achieved effective results for fabric defect detection, however, most of them mainly target detection accuracy at the expense of detection speed. Therefore, we propose a fabric defect detection method called PEI-YOLOv5. First, Particle Depthwise Convolution (PDConv) is proposed to extract spatial features more efficiently while reducing redundant computations and memory access, reducing model computation and improving detection speed. Second, Enhance-BiFPN(EB) is proposed based on the structure of BiFPN to enhance the attention of spatial and channel feature maps and the fusion of information at different scales. Third, we improve the loss function and propose IN loss, which improves the problem that the original IOU loss is weak in detecting small targets while speeding up the convergence of the model. Finally, five more common types of defects were selected for training in the GuangDong TianChi fabric defect dataset, and using our proposed PEI-YOLOv5 with only 0.2 Giga Floating Point Operations (GFLOPs) increase, the mAP improved by 3.61%, reaching 87.89%. To demonstrate the versatility of PEI-YOLOv5, we additionally evaluated this in the NEU surface defect database, with the mAP of 79.37%. The performance of PEI-YOLOv 5 in these two datasets surpasses the most advanced fabric defect detection methods at present. We deployed the model to the NVIDIA Jetson TX2 embedded development board, and the detection speed reached 31 frames per second (Fps), which can fully meet the speed requirements of real-time detection. [ABSTRACT FROM AUTHOR]

The growth of multimedia applications poses new challenges to software facilities in edge computing. Developers must effectively develop edge computing software to accommodate the rapid expansion of multimedia applications. Code search has become a prevalent practice to enhance the efficiency of the construction of edge software infrastructure. Researchers have proposed lots of approaches for code search, and employed deep learning technology to extract features from program representations, such as token, AST, graphs, method name, and API. Nevertheless, two prominent issues remain: 1) there are only a few studies on the effective use of graph representation for code search (especially in Java language), and 2) there is a lack of empirical study on the contributions of different program representations. To address these issues, we conduct an empirical study to explore program representations, especially program graphs. To the best of our knowledge, this is the first attempt to conduct code search with mixed graphs representation for Java language, containing the control flow graph and the program dependence graph. We also present a hybrid approach to capture and fuse the features of a program with representations of Token, AST, and Mixed Graphs (TAMG). The results of our experiment show that our approach possesses the best ability (R@1 with 37% and R@10 with 67.1%). Our graph representation exhibits a positive effect, and the token and AST also have a significant contribution to the code search. Our findings can aid developers in efficiently searching for the desired code while constructing the software infrastructure for edge computing, which is crucial for the rapid expansion of multimedia applications. [ABSTRACT FROM AUTHOR]

Educational institutions today are embracing technology to enhance education quality through intelligent systems. This study introduces an innovative strategy to boost the performance of such procedures by seamlessly integrating machine learning on edge devices and cloud infrastructure. The proposed framework harnesses the capabilities of a Hybrid 1D Convolutional Neural Network (CNN) and Long Short-Term Memory Network (LSTM) architecture, offering profound insights into intelligent education. Operating at the crossroads of localised and centralised analyses, the Hybrid 1D CNN-LSTM architecture signifies a significant advancement. It directly engages edge devices used by students and educators, laying the groundwork for personalised learning experiences. This architecture adeptly captures the intricacies of various modalities, including text, images, and videos, by harmonising 1D CNN layers and LSTM modules. This approach facilitates the extraction of tailored features from each modality and the exploration of temporal intricacies. Consequently, the architecture provides a holistic comprehension of student engagement and comprehension dynamics, unveiling individual learning preferences. Moreover, the framework seamlessly integrates data from edge devices into the cloud infrastructure, allowing insights from both domains to merge. Educators benefit from attention-enhanced feature maps that encapsulate personalised insights, empowering them to customise content and strategies according to student learning preferences. The approach bridges real-time, localised analysis with comprehensive cloud-mediated insights, paving the path for transformative educational experiences. Empirical validation reinforces the effectiveness of the Hybrid 1D CNN-LSTM architecture, cementing its potential to revolutionise intelligent education within academic institutions. This fusion of machine learning across edge devices and cloud architecture can reshape the educational landscape, ushering in a more innovative and more responsive learning environment that caters to the diverse needs of students and educators alike. [ABSTRACT FROM AUTHOR]

For humans, taste is essential for perceiving the nutrient content or harmful components of food. The current method of taste sensory evaluation relies on artificial sensory evaluation and an electronic tongue. The former has strong subjectivity and poor repeatability, and the latter is not sufficiently flexible. To decode people's objective taste perception, a strategy for acquiring and recognizing four classes (sour, sweet, bitter, and salty) in taste-related electroencephalograms (EEGs) was proposed. First, according to the proposed experimental paradigm, the taste-related EEGs of subjects under different taste stimulations were collected. Second, to avoid insufficient training of the model due to the small number of EEG samples, a temporal and spatial reconstruction data augmentation (TSRDA) method was proposed, effectively augmenting taste-related EEGs by reconstructing the important features in temporal and spatial dimensions. Third, a multiview channel attention (MVCA) module was introduced into a designed convolutional neural network to extract the important features of the augmented EEG. The proposed method had an accuracy of 99.56%, F1 score of 99.48%, and kappa value of 99.38%, showing the method's ability to successfully decoded sour, sweet, bitter, and salty EEG signals. In conclusion, combining TSRDA with EEG technology provides an objective and effective method for the sensory evaluation of food taste. [ABSTRACT FROM AUTHOR]

Long sequence information remains a challenging problem in deep learning nowadays for predicting remaining useful life (RUL). In this work, we propose a novel deep learning module called attention long short-term memory projected (ALSTMP) for RUL estimation to mitigate the inefficient information of long-term dependencies. The ALSTMP is designed to utilize attention mechanisms in traditional long short-term memory (LSTM) for effectively collecting key features of the dataset. Moreover, the time-window length method is implemented to generate a better feature extraction. The proposed model not only outperforms the traditional LSTM and its extension but also the latest existing approaches with a smaller quantity of parameters compared with recent deep learning approaches. [ABSTRACT FROM AUTHOR]