Narrow Results

Due to the limitations of imaging equipment and image transmission conditions on daily image acquisition, the images acquired are usually low-resolution images, and it will cost a lot of time and economic costs to increase image resolution by upgrading hardware equipment. In this paper, we propose an image super-resolution reconstruction algorithm based on spatio-temporal-dependent residual network MSRN, which fuses multiple features. The algorithm uses the surface feature extraction module to extract the input features of the image, and then uses the deep residual aggregation module to adaptively learn the deep features, and then fuses multiple features and learns the global residual. Finally, the high-resolution image is obtained through the up-sampling module and the reconstruction module. In the model structure, different convolution kernels and jump connections are used to extract more high-frequency information, and spatio-temporal attention mechanism is introduced to focus on more image details. The experimental results show that compared with SRGAN, VDSR and Laplacian Pyramid SRN, the proposed algorithm finally achieves better reconstruction effect, and the image texture details are clearer under different scaling factors. In objective evaluation, the peak signal-to-noise ratio (PSNR) and structure similarity (SSIM) of the proposed algorithm are improved compared with SRGAN.

With the rapid advancement of the digital economy, the financial market confronts unprecedented complexity and interconnectivity, rendering the precise prediction of credit bond default risk particularly crucial. This paper introduces a novel credit bond default risk measurement model (GST-GRU) predicated on a spatio-temporal attention network and genetic algorithm, designed to enhance the accuracy and robustness of risk prediction. Initially, data preprocessing is undertaken, encompassing time series data cleaning, completion of missing values for historical financial information and bond default statuses, and extraction of spatial discrete information through independent vector coding. Subsequently, a spatio-temporal attention mechanism is employed to amplify feature information in both domains, while the GRU network captures the long-term dependencies within the time series data. Thereafter, model parameters are refined using a genetic algorithm to ensure global optimality. Experimental results demonstrate that the GST–GRU model markedly improves prediction accuracy across multiple public and self-constructed datasets, surpassing traditional models. This research furnishes robust technical support for risk management in the financial market, fosters the evolution of credit bond default risk prediction technology, and lays the groundwork for the intelligence and automation of future financial systems.

Accurately predicting and controlling the flue gas pressure at the induced draft fan inlet is crucial for environmental protection and boiler efficiency. This study designed a model of Attention-enhanced Deep Belief Network (AT-DBN) to predict the flue gas pressure parameters at the induced draft fan inlet, based on key improvements to the Deep Belief Network (DBN). The main improvement involves introducing an attention mechanism to enhance the prediction accuracy of the model for the flue gas pressure parameters at the induced draft fan inlet. The study begins with meticulous preprocessing of historical operational data, including identifying and removing outliers to ensure data quality for subsequent analysis. Next, Pearson correlation analysis is employed to systematically evaluate the correlation between various variables and the flue gas pressure, thereby identifying key variables that significantly affect prediction performance. Finally, an attention mechanism is integrated into the traditional DBN architecture, aiming to enhance the network’s ability to perceive important information within the input features, thereby improving prediction accuracy and model generalization. The results demonstrate that the proposed AT-DBN model can accurately predict the flue gas pressure at the induced draft fan inlet across multiple test sets, providing valuable guidance for power plant production operations.

In order to accurately and quickly understand the characteristics and skills of athletes’ traditional sports actions, an unmarked AI recognition algorithm for athletes’ traditional sports actions based on attention mechanism is designed. The lightweight space-time map convolution neural network (ST-GCN) based on attention mechanism in AI technology is used to complete the traditional sports action recognition of athletes. The traditional sports action skeleton map of athletes is constructed as unlabeled samples and input into the ST-GCN network. The time and space characteristics of the input skeleton map are extracted through the time convolution network (TCN) and graph convolution neural network (GCN), respectively. Add graph attention mechanism and channel attention mechanism in the network layer and channel, improve the feature expression ability and action recognition accuracy, and introduce Ghost module to replace the original image convolution work, complete the network lightweight processing, and improve the efficiency of ST-GCN network recognition of sports actions. Complete the classification and recognition of athletes’ traditional sports actions through the standard SoftMax. Experiments show that when the size of the attention mechanism window added in ST-GCN is 1000, the network can have the best performance, and the identified resource cost is relatively minimal. After adding attention mechanism, the F1 score of ST-GCN network in unmarked training is close to 1. The skeleton extraction results of athletes are very accurate, and the algorithm can accurately identify different actions generated by different kinds of movements.

The accurate identification of errors in machine-translated English textual content is both necessary and tough, with massive implications for natural language processing packages. current device translation structures, while an increasing number of sophisticated, nevertheless fall prey to quite a number of errors that may compromise which means and fluency. This paper addresses these shortcomings through offering a greater system mastering method for the adaptive recognition of translation errors. We gift a singular framework that integrates the Transformer version, famed for its efficacy in shooting contextual relationships within text sequences, with a robust attention mechanism that prioritizes salient data at some stage in translation. The innovation of this studies lies in the incorporation of meta-getting to know techniques, allowing the model to self-modify in response to various classes of mistakes, therefore refining the precision of errors recognition. We outline the technique in detail, emphasizing the systematic steps taken to enrich the model’s adaptability. The efficacy of our approach is substantiated through comparative experiments, which demonstrate extremely good upgrades in errors identification over existing techniques. The results suggest the capacity of our adaptive mechanism to decorate the excellent of machine translation, paving the method for greater, reliable and nuanced language translation tools.

As the core component of the wind turbine, blades are susceptible to deterioration caused by natural environmental factors. This can manifest as erosion, fissures, and gel coat detachment, which collectively impair the efficiency of wind power generation and the safe operation of the turbine. In order to address the issue of low detection accuracy for wind turbine blade defects in complex environments, an enhanced YOLOv8n wind turbine blade defect detection algorithm has been proposed. First, the Large Separable Kernel Attention (LSKA) attention mechanism is introduced into the Spatial Pyramid Pooling Fast (SPPF) module of the backbone network, thereby enhancing the network’s attention and improving the model’s feature extraction capability. Second, the neck employs a weighted bidirectional feature pyramid (Bi-FPN) structure and integrates a P2 small target detection layer, thus enhancing the model’s multi-scale feature fusion capability and improving its small target detection accuracy. Finally, the loss function of the original model is optimized using WIoU, thereby improving the model’s defect detection accuracy. The results of the defect detection experiments on wind turbine blade images demonstrate that the accuracy of the proposed method has been enhanced by 7.9%, the mAP 50 has been improved by 2.6%, and the number of parameters has been reduced by 23%.

Image detection is to accurately locate the object in the image and classify the object. Graph neural network is a kind of deep neural network based on graph data structure. In recent years, with the rapid development of artificial intelligence technology, image object detection methods based on deep learning have emerged in an endless stream. Although some studies have improved the association expression of images, there is no effective mining of the association between objects and categories, resulting in limited detection accuracy. Therefore, this project intends to study a new object detection method based on graph-relation inference. On this basis, an attention-based approach is proposed to model the dependencies between candidate regions and tags. First, the extracted candidate regions are classified. On this basis, the present study proposes an autonomous attention mechanism to model and analyze the interdependencies among diverse data sets, and to learn the interdependence between the two sets through the interactive attention model. Finally, based on the similarity between the two new sets, the classification results are predicted and used in edge regression analysis to further improve the performance of the image object detection algorithm. Experimental results show that compared with the baseline model, the proposed method has a 1.6% increase in mean accuracy (mAP) on PascalVOC dataset and a 1.7% increase in mean accuracy (AP) on MS-COCO dataset, demonstrating the effectiveness and superiority of the proposed method in target detection tasks.

As society advances, computer vision will play an increasingly crucial role in digital and intelligent transformations. Known as deep learning models, Convolutional Neural Networks (CNNs) have emerged as a key component of computer vision due to their superior performance in automatically detecting image features, handling high-dimensional data and performing large-scale classification tasks. This paper examines the development of CNNs, leveraging the strengths of current mainstream image recognition methods, and proposes a Self-Distillation and Attention-based Convolutional Neural Network (SDACNN) model to further enhance CNN accuracy. Experimental results demonstrate that the proposed model effectively accomplishes image recognition tasks.

For visually impaired individuals, the ability to receive real-time textual descriptions of image content can significantly enhance their quality of life and learning. In daily life, automatically generated image descriptions can help them quickly acquire key information, improving their responsiveness and communication experience. In education, particularly in subjects involving complex images or spatial understanding, descriptive content can aid in better comprehension of study materials, thereby narrowing the learning gap between visually impaired students and their peers, and promoting educational equity. This study has developed an automated image description generation system for visually impaired individuals, leveraging deep learning models to accurately diagnose and describe image content. The system employs a combination of Convolutional Neural Networks (CNNs) and Long Short-Term Memory networks (LSTM), integrating attention mechanisms to significantly enhance the quality and performance of the generated descriptions. Trained on large-scale image datasets such as CIFAR-100, Flickr8k and MS COCO, the system can accurately recognize image content and produce corresponding descriptions. Experimental results demonstrate that the proposed system outperforms traditional methods across multiple datasets, generating descriptions with high accuracy and fluency. This paper offers an in-depth description of the system’s design and execution, covering the design of the system architecture, dataset selection and preprocessing, model training and optimization and system testing and evaluation. The research results showcase the broad application prospects of automated image description technology in visual aid devices and automated news generation systems, offering valuable references for future technical optimization and application expansion.

Accurately and rapidly detecting traffic object has been attracted intensive attention due to its potential applications in the fields of autonomous driving, traffic flow monitoring, augmented reality (AR) and so on. However, there are many difficulties in the process of traffic object detection indeed, such as occlusion and aggregation between objects, insufficient feature extraction of objects, in particular the presence of a large number of small objects, which bring great challenges to these traffic objects detection. In this paper, an improved traffic object detection model based on You-Only-Look-Once version 5 small (YOLOv5s) is proposed to address the issues. By utilizing spatial pyramids to extract multi-scale spatial features and applying Squeeze-and-Excitation (SE) channel attention to capture more global and local semantic features, especially by designing a sub-network in the neck to fuse high-resolution information in shallow layers with more accurately semantic information in deep layers, the detection sensitivity of object features is enhanced. More importantly, by explanting decoupled-head into the network, outstanding performance of the model with high detection accuracy and rapid detection speed is realized. The experimental results on the Karlsruhe Institute of Technology and Toyota Technological Institute (KITTI) and Laboratory for Intelligent and Safe Automobiles (LISA) traffic signs datasets both show that the modified model significantly improves the detection accuracy. Meanwhile, the high real-time performance is still maintained. Undoubtedly, the modified model proposed in this paper can effectively address many difficulties in traffic object detection under various complex scenes, which would be greatly helpful for its potential applications in the future.

The attention mechanism has been successfully used in a sequence consisted of a series of word embeddings to improve the representation of the sequence. Inspired by this, we leverage the attention mechanism on a set of tasks to implement a multi-task transfer learning method called AMP (Attentions between Multiple Prompts). First, we encode a task into a prompt as task representation called task embedding. Second, we learn an attention component on all task embeddings to generate a combined prompt for each task, which is an attention-weighted sum of task embeddings. Each combined prompt incorporates the knowledge of all tasks. The word embedding is a vector, but the task embedding is a 2D matrix. The attention mechanism can be exploited on a set of vectors rather than on a set of matrices. The prior methods employ pooling or flattened method to transform the matrix to the vector for computing the attentions between matrices. We propose a method called DAM (Direct Attention Mechanism) which can compute attentions between matrices directly without transforming. DAM method can more exactly compute the attentions between matrices. Wide experiments demonstrate that AMP outperforms prompt-tuning method and prior prompt transfer methods.

Liver metastases are a common complication of liver cancer, and accurate image segmentation is crucial for diagnosis and treatment. Manual segmentation of the liver and tumors from CT images is time-consuming and subjective. While computer-aided segmentation has been widely adopted, segmenting liver metastases remains challenging due to the variability in shape, size and contrast. In this study, we propose a 2D network model (RL-RCUNet) that enhances the UNet architecture by incorporating an LSK module to address the issue of incomplete receptive fields. Additionally, an improved RF-CBAM module is added to the skip connections to optimize parameter sharing. Trained and tested on a dataset from cooperative hospitals, our model demonstrates accurate segmentation of liver and liver metastases from CT images.

With the rapid development of the smart grid, the fast and accurate fault location of low-voltage distribution networks has become the key to ensuring the stability and reliability of the power supply. This paper aims to explore and construct a fault location model of low-voltage distribution network based on an attention diagram neural network. First, this paper analyzes the current situation and challenges of fault location in low-voltage distribution network, and points out that traditional methods have limitations when processing large-scale and high-dimensional power system data. Subsequently, a graph neural network (GNN) is introduced for processing graph-structured network data, and combined with attention mechanisms. Thus, an innovative attention-graph neural network model (named as A-GNN) is proposed for the purpose. The model can make full use of the topology structure and node feature information in the power grid, and dynamically adjust the information aggregation weight between different nodes through the attention mechanism. This is expected to achieve efficient and accurate fault location. In the experimental part, we trained and tested the A-GNN model based on the real low-voltage distribution network dataset, and compared it with several prediction models. The experimental results show that the A-GNN model has higher accuracy and recall rate in fault location tasks, especially in complex fault scenarios.

In an era of big data, accurately detection towards cyber attacks is crucial to retain healthy operation of cyberspace. Traditional detection methods often existed locality when detecting new types of attacks. To deal with this issue, this paper proposes a robust cyber attack detection method through attention-based graph neural network. It introduces channel attention to achieve linear pooling of data, and then fuses graph convolutional network to construct a graph representation model for cyberspace data. The nodes and edges in the network are represented as the structure of the graph, and the attention mechanism is combined to dynamically calculate the importance weight of nodes in each graph convolution layer. The technical roadmap is to utilize the strong feature representation ability of attention-based graph learning to establish a robust cyber attack detection approach. At last, the proposed approach is assessed using the CSIC2010 dataset by comparing with several typical methods. The experimental results show that the proposed method can better identify and distinguish cyber attack types, and has higher accuracy and robustness. This proposal is expected to help cyberspace and operators avoid increasingly complex cyber attacks in practical applications.

With the advent of the information explosion era, how to obtain the core content of long text quickly and accurately has become a research hotspot. With traditional automatic summarization methods, it is often difficult to capture the deep semantic relationships and long-distance dependencies in the text, resulting in poor quality of the generated summarization. In this paper, based on Graph Neural Network (GNN) and Bidirectional Encoder Representations from Transformers (BERT), an automatic summarization model for long text is constructed. First, long text is divided into text embedding representation to obtain rich context information. The hierarchical decomposition position embedding and Convolutional Neural Network (CNN) are adopted to pre-train BERT model. Furthermore, in order to capture the global and local semantic perception information of text, the GNN model based on topic algorithm and BiLSTM is integrated to construct a new model that can accurately identify the overall structure and context of text and realize more accurate automatic summarization of long text. Extensive evaluation was performed simultaneously on multiple long text datasets and compared with other summarization methods. The experimental results show that the proposed method can significantly improve the performance of the long text automatic summarization task, which not only improves the accuracy of the summarization, but also better preserves the key information in the original text. The research results not only provide a new technical path for the field of automatic summarization, but also provide a new idea for long text processing and analysis.

The monitoring and maintenance of grid equipment have become increasingly crucial due to the continual progress in smart grid technology. Efficient identification technology for grid equipment is crucial for enabling equipment status monitoring and fault diagnosis, directly influencing the operational stability of the grid concerning precision and timely functionality. Nevertheless, the reliance of current image recognition methods on intricate models and extensive computational resources poses implementation challenges in resource-limited field environments, thereby restricting their use in operations such as drone-based power line inspections. In response to this obstacle, the paper introduces a streamlined identification approach for grid equipment through model compression. This method aims to uphold recognition precision while minimizing the computational workload and storage demands of the model, making it well-suited for integration into drone-based power line inspections. Introducing a target recognition network, this method integrates tailored multi-scale information for grid equipment and embeds an attention mechanism within the network to enhance the model’s capacity for identifying crucial features. Expanding on this approach, model compression techniques are utilized to condense the trained model. This process maintains accuracy by removing redundant weights, thereby shrinking the model’s size and computational complexity, ultimately achieving a lightweight network.

Deep learning algorithms are highly effective at handling complex and challenging tasks such as image classification and detection. Over the past few decades, there have been a variety of convolutional neural networks (CNNs) with varying architectures to improve accuracy in object detection. However, a variety of factors, including brightness, enclosures, viewing distances, and background components, can affect the appearance, size, and shape of objects making the object detection task even more challenging. A multi-head external attention mechanism-based CenterNet model has been proposed to enhance the accuracy in detection of objects. The feature extraction process is carried out using Hourglass-104 network and Adaptive Feature Pyramid Network (AFPN). High-level features are derived through contextual modeling using an adaptive feature fusion (AFF) module with multi-head external attention. A hybrid feature selection method called adaptive hybrid feature selection (AHFS) determines the best features followed by prediction of the objects by the improved CenterNet. In order to assess the object detection accuracy, the experiment was conducted using the MS-COCO dataset on average precision (AP), mean average precision (mAP), and average recall (AR) metrics. Our proffered method achieves 64.76% on the MS-COCO dataset, improving the accuracy by 2.5% compared to other state-of-the-art models.

Human action recognition in videos is an important task in computer vision with applications in fields such as surveillance, human–computer interaction, and sports analysis. However, it is a challenging task due to the complex background changes and redundancy of long-term video information. In this paper, we propose a novel bi-directional long short-term memory method with attention pooling based on joint motion and difference entropy (JEAP-BiLSTM) to address these challenges. To obtain discriminative features, we introduce a joint entropy map that measures both the entropy of motion and the entropy of change. The Bi-LSTM method is then applied to capture visual and temporal associations in both forward and backward directions, enabling efficient capture of long-term temporal correlation. Furthermore, attention pooling is used to highlight the region of interest and to mitigate the effects of background changes in video information. Experiments on the UCF101 and HMDB51 datasets demonstrate that the proposed JEAP-BiLSTM method achieves recognition rates of 96.4% and 75.2%, respectively, outperforming existing methods. Our proposed method makes significant contributions to the field of human action recognition by effectively capturing both spatial and temporal patterns in videos, addressing background changes, and achieving state-of-the-art performance.

With the development of convolutional neural networks (CNNs), deep learning-based face super-resolution (FSR) approaches have achieved remarkable results in recent years. However, existing FSR methods often rely on face prior knowledge, which greatly increases network complexity and computation. In this paper, we propose DAFNet, a dual attention fusion-based FSR network comprising numerous face attention fusion modules (FAFMs). FAFM is a residual structure containing attention mechanism, which is divided into feature branches and attention branches. The feature branch introduces an hourglass block to extract multi-scale information, while the attention branch incorporates channel attention and spatial attention in series. This design ensures that the network prioritizes important information and effectively recovers detailed facial features. Luminance-chrominance error loss and gradient loss are introduced to guide the training process. Additionally, adversarial loss and perceptual loss are incorporated to enhance the recovery of visually realistic face images. Notably, our method can produce clear faces at a high-scale factor of eight times without relying on any facial prior information, effectively reducing network complexity. Quantitative and qualitative experiments conducted on the CelebA and Helen datasets underscore the effectiveness of the proposed model.

Researchers across the globe have been investigating underwater photographs and a way to capture pictures with outstanding clarity for the past couple of decades. Also, improving the obtained photographs is an exhausting task. Usually, applied underwater image-capturing devices are unable to acquire high-resolution photos underwater, and their maintenance is extremely costly. The underwater images include multiple flaws because of biological processes like attenuation and scattering. These photographs experience color distortion, blurriness, and low contrast effects. Technologies that utilize deep learning have become more popular among several research studies and have gradually grown in impact on society. Several techniques demand sets of training photos, but gathering such expected sets can be challenging because of the complex nature of the underwater environment. Generating and restoring an image from water is a difficult task that has gained prominence in recent days. By lowering graininess, adjusting, and refining the photos using deep learning models, the major goal is to enhance underwater images. To accomplish this objective, an intelligent attention-based deep learning model is proposed. In the first stage, the unrefined images are gathered from typical data sources. Further, the collected underwater images are fed into the model of Attentive-based Trans-UNet-CycleGAN (ATUNet-CGAN), where the Transformer-based UNet model is integrated with the Cycle Generative Adversarial Networks (GANs). Also, the attention mechanism process is involved in Trans-UNet-CycleGAN for improving the superiority of submarine images. Finally, the performance of the model is validated using different metrics and correlated among baseline approaches. Therefore, the proposed methodology outperforms the exploitation of better enhancement of image quality.