Search

Currently, most image caption generation models use object features output from pre-trained detectors as input for multimodal learning of images and text. However, such downscaled object features are unable to capture the global contextual relationships of an image, leading to information loss and limitations in semantic understanding. In addition, relying on a single object feature extraction framework limits the accuracy of feature extraction, resulting in the inaccurate descriptive sentences. In order to solve the above-mentioned problems, we propose the feature-augmented (FA) module, which forms the front-end part of the encoder architecture. This module uses a multimodal pre-trained model CLIP (contrastive language-image pre-training) as an auxiliary feature extractor, which enhances the image descriptive capability by utilizing the extracted rich semantic features as a complement to the missing information such as scene and object relations, so as to better encode the information necessary for the captioning task. In addition, we add channel attention to the self-attention in the encoder structure, compressing the spatial dimensions of the feature map to allow the model to concentrate on more valuable information. We validate the effectiveness of our proposed method on datasets such as MS-COCO, analyze the components and importance of our model, and demonstrate a marked increase in performance compared to other state-of-the-art models. [ABSTRACT FROM AUTHOR]

Unmanned aerial vehicles (UAVs) have gained significant popularity in recent years due to their ability to capture high-resolution aerial imagery for various applications, including traffic monitoring, urban planning, and disaster management. Accurate road and vehicle segmentation from UAV imagery plays a crucial role in these applications. In this paper, we propose a novel approach combining dual attention mechanisms and efficient multilayer feature aggregation to enhance the performance of road and vehicle segmentation from UAV imagery. Our approach integrates a spatial attention mechanism and a channel-wise attention mechanism to enable the model to selectively focus on relevant features for segmentation tasks. In conjunction with these attention mechanisms, we introduce an efficient multi-layer feature aggregation method that synthesizes and integrates multi-scale features at different levels of the network, resulting in a more robust and informative feature representation. Our proposed method is evaluated on the UAVid semantic segmentation dataset, showcasing its exceptional performance in comparison to renowned approaches such as U-Net, DeepLabv3+, and SegNet. The experimental results affirm that our approach surpasses these state-of-the-art methods in terms of segmentation accuracy. [ABSTRACT FROM AUTHOR]

The vast volumes of videos produced daily require highly efficient measures to ensure that key information is reported for effective review and storage, which leads to the popularity of video summarization techniques. Deep learning has shown its advantages in video summarization, especially convolutional neural network, which are effective in extracting features for video summarization. However, the deep network layers and the limited range of temporal dependence make it challenging to deploy the network and thus affect the accuracy of identifying important video frames. To tackle these issues, we present a knowledge distillation-based attentive network (KDAN) for supervised video summarization in this paper. The proposed method separates the full convolutional network from the attention mechanism based on the idea of education and learning processes in biology and uses a full convolutional network as a teacher network to guide the learning of the student network consisting of an attention mechanism. The obtained lightweight network considers the knowledge learned from both networks, thus solving the problems of explosion in the number of participants and slow training. We have conducted experiments on two widely used benchmarks SumMe and TVSum. DANtea achieves F-scores 53.09 and 60.30, and DAN achieves F-scores 51.26 and 61.55 in Canonical settings on the SumMe and TVSum datasets, respectively. Experiments on two public benchmarks SumMe and TVSum demonstrate the effectiveness and superiority of the proposed network over existing state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Objective. Laparoscopic renal unit-preserving resection is a routine and effective means of treating renal tumors. Image segmentation is an essential part before tumor resection. The current segmentation method mainly relies on doctors manual delineation, which is time-consuming, labor-intensive, and influenced by their personal experience and ability. And the image quality of segmentation is low, with problems such as blurred edges, unclear size and shape, which are not conducive to clinical diagnosis. Approach. To address these problems, we propose an automated segmentation method, i.e. the UNet++ algorithm fusing multiscale residuals and dual attention (MRDA_UNet++). It replaces two consecutive 3 × 3 convolutions in UNet++ with the 'MultiRes block' module, which incorporates coordinate attention to fuse features from different scales and suppress the impact of background noise. Furthermore, an attention gate is also added at the short connections to enhance the ability of the network to extract features from the target area. Main results. The experimental results show that MRDA_UNet++ achieves 93.18%, 92.87%, 93.66%, and 92.09% on the real-world dataset for MIoU, Dice, Precision, and Recall, respectively. Compared to the baseline model UNet++ on three public datasets, the MIoU, Dice, and Recall metrics improved by 6.00%, 7.90% and 18.09% respectively for BUSI, 0.39%, 0.27% and 1.03% for Dataset C, and 1.37%, 1.75% and 1.30% for DDTI. Significance. The proposed MRDA_UNet++ exhibits obvious advantages in feature extraction, which can not only significantly reduce the workload of doctors, but also further decrease the risk of misdiagnosis. It is of great value to assist doctors diagnosis in the clinic. [ABSTRACT FROM AUTHOR]

In the real world, the data distribution often presents a long tail distribution, and the imbalance of data will lead to the model learning bias to the head class. To address the influence of long tail distribution on image classification, this paper proposes a feature channel interactive long tail image classification model based on dual attention. Firstly, the dual attention module is used to capture the autocorrelation and spatial dimension information of the feature map, and the enhanced image is obtained by transformation and class activation map. After that, image preprocessing is performed on the enhanced data set to reduce the over-fitting of the model to the head, and the features that are more conducive to tail classification are obtained through learning. Finally, by interacting with the local channels adjacent to the features, the correlation between the channels is extracted to obtain more robust features. The method achieves good performance on CIFAR10-LT, CIFAR100-LT and ImageNet datasets, which proves the effectiveness of the model. [ABSTRACT FROM AUTHOR]

The inductive relation prediction of knowledge graphs, as an important research topic, aims at predicting the missing relation between unknown entities with many real-world applications. Existing approaches toward this problem mostly use enclosing subgraphs to extract the features of target nodes to make predictions; however, there is a tendency to ignore the neighboring relations outside the enclosing subgraph, thus leading to inaccurate predictions. In addition, they also neglect the rich commonsense information that can help filter out less convincing results. In order to address the above issues, this paper proposes a commonsense-guided inductive relation prediction method with a dual attention mechanism called CNIA. Specifically, in addition to the enclosing subgraph, we added the multi-hop neighboring relations of target nodes, thereby forming a neighbor-enriched subgraph where the initial embeddings are generated. Next, we obtained the subgraph representations with a dual attention (i.e., edge-aware and relation-aware) mechanism, as well as the neighboring relational path embeddings. Then, we concatenated the two embeddings before feeding them into the supervised learning model. A commonsense re-ranking mechanism was introduced to filter the results that conformed to commonsense. Extensive experiments on WN18RR, FB15k-237, and NELL995 showed that CNIA achieves better prediction results when compared to the state-of-the-art models. The results suggested that our proposed model can be considered as an effective and state-of-the-art solution for inductive relation prediction. [ABSTRACT FROM AUTHOR]

Hyperspectral image (HSI) reconstruction plays a crucial role in compressive spectral imaging with coded aperture snapshot spectrometry. Although HSI reconstruction has attracted much attention in recent years, it remains a challenging problem. Existing deep learning-based methods leverage all the spectral information to reconstruct the HSI images without considering the spectral redundancy of HSI images, leading to high computational costs. In this article, we present an efficient method named dynamic spectral guided spatial sparse transformer (DGST). Specifically, DGST consists of three core modules as follows. 1) spectral sparse multihead self-attention hybrid spatial feature enhancement (SSHE) module, which employs a top-k spectral sparsity method to filter noise and redundant spectral information while extracting spectral information from HSI. 2) Spatial information compensation module, which utilizes a multiscale approach to extract spatial information and compensates for the spatial information neglected by SSHE. 3) Mask-guided spatial sparse multihead self-attention hybrid spectral enhancement module, which dynamically generates masks to guide the filtering of irrelevant regions, reducing computational costs while focusing on spatial information reconstruction. Our DGST improves the quality of HSI reconstruction by integrating spatial–spectral details and global information. Extensive experiments on public HSI reconstruction benchmark datasets demonstrate that our approach achieves state-of-the-art performance in end-to-end hyperspectral reconstruction. The superior performance of the proposed DGST is showcased on real and simulated hyperspectral imaging datasets.

One of the challenges in image processing involves single image deraining (SID). Existing methods do not exploit images at multi-scales and often overlook spatial and channel information. These methods fail to account for different rain conditions, leading to challenges in effectively removing a wide range of rain streak patterns, such as diverse directional or dense streaks. Moreover, they often result in the loss of fine texture details or a blurred background in the process of eliminating rain streaks from images captured in heavy rain. Mostly state-of-the-art (SOTA) deraining models achieve higher performance in removing rain from rainy images but at the expense of a high number of parameters, which results in computational complexity and memory requirements. Furthermore, they also do not consider high-level visioned evaluation metrics to perform deep evaluations of the proposed models. In this paper, a simple lightweight network with few parameters and relatively shallow depth is proposed by fusing the traditional Gaussian-Laplacian pyramid technique with the attention module. We propose a novel attention-based lightweight pyramid network for image deraining and detection (LPN-IDD) to achieve better deraining performance. The proposed model includes a dual attention module integrated with the Gaussian-Laplacian pyramids network. In LPN-IDD, different levels of Laplacian pyramids can extract multi-scale features to adapt to different shapes and types of rain streaks. Residual and recursive blocks are used in each subnetwork along with dual attention blocks to resist the occlusion or texture feature while suppressing unnecessary features. Extensive experimentation performed on the SID synthetic and real-world datasets demonstrate the effectiveness of the proposed model in image deraining tasks. For deep evaluation of the proposed methodology, object detection models including faster-RCNN, YOLO-V3, YOLO-V7 are used along with full reference evaluation metrics.

Radar Automatic Target Recognition (RATR) based on high-resolution range profile (HRRP) has received intensive attention in recent years. In practice, RATR usually needs not only to recognize in-library samples but also to reject out-of-library samples. However, most rejection methods lack a specific and accurate description of the underlying distribution of HRRP, which limits the effectiveness of the rejection task. Therefore, this paper proposes a novel rejection method for HRRP, named Deep Multi-modal Support Vector Data Description (DMMSVDD). On the one hand, it forms a more compact rejection boundary with the Gaussian mixture model in consideration of the high-dimensional and multi-modal structure of HRRP. On the other hand, it captures the global temporal information and channel-dependent information with a dual attention module to gain more discriminative structured features, which are optimized jointly with the rejection boundary. In addition, a semi-supervised extension is proposed to refine the boundary with available out-of-library samples. Experimental results based on measured data show that the proposed methods demonstrate significant improvement in the HRRP rejection performance. [ABSTRACT FROM AUTHOR]

The surge in malware threats propelled by the rapid evolution of the internet and smart device technology necessitates effective automatic malware classification for robust system security. While existing research has primarily relied on some feature extraction techniques, issues such as information loss and computational overhead persist, especially in instruction-level tracking. To address these issues, this paper focuses on the nuanced analysis of API (Application Programming Interface) call sequences between the malware and system and introduces TTDAT (Two-step Training Dual Attention Transformer) for malware classification. TTDAT utilizes Transformer architecture with original multi-head attention and an integrated local attention module, streamlining the encoding of API sequences and extracting both global and local patterns. To expedite detection, we introduce a two-step training strategy: ensemble Transformer models to generate class representation vectors, thereby bolstering efficiency and adaptability. Our extensive experiments demonstrate TTDAT's effectiveness, showcasing state-of-the-art results with an average F1 score of 0.90 and an accuracy of 0.96. [ABSTRACT FROM AUTHOR]

To address the problems of slow detection speed and low accuracy commonly found in cigarette factory warehouse workshops when detecting tobacco moth, a visual detection method for tobacco moth in cigarette factories based on improved lightweight YOLOv5s was developed. The method utilizes the correlation and redundancy between feature maps to design the EESP-Ghost module, and uses this module as the basis for designing a double-attention Ghost-bneck block incorporating an efficient spatial pyramid, which is introduced into the YOLOv5s model to achieve lightweighting of the deep neural network model while improving the detection accuracy. The method is used for validation experiments on the tobacco moth dataset. The results showed that the method improved the average accuracy by 4.37% with only 49.88% of the original YOLOv5s parameter count. When the tobacco moth adhering to the sticky board was detected in a real detection scenario, the method has high detection confidence and correct detection number, which could realize the high-precision real-time detection of the tobacco moth in the cigarette factory, and provide a guarantee for the effective control of the tobacco moth. [ABSTRACT FROM AUTHOR]

Recently, hyperspectral image classification has made great progress with the development of convolutional neural networks. However, due to the challenges of distribution shifts and data redundancies, the classification accuracy is low. Some existing domain adaptation methods try to mitigate the distribution shifts by training source samples and some labeled target samples. However, in practice, labeled target domain samples are difficult or even impossible to obtain. To solve the above challenges, we propose a novel dual-attention deep discriminative domain generalization framework (DAD3GM) for cross-scene hyperspectral image classification without training the labeled target samples. In DAD3GM, we mainly design two blocks: dual-attention feature learning (DAFL) and deep discriminative feature learning (DDFL). DAFL is designed to extract spatial features by multi-scale self-attention and extract spectral features by multi-head external attention. DDFL is further designed to extract deep discriminative features by contrastive regularization and class discrimination regularization. The combination of DAFL and DDFL can effectively reduce the computational time and improve the generalization performance of DAD3GM. The proposed model achieves 84.25%, 83.53%, and 80.63% overall accuracy on the public Houston, Pavia, and GID benchmarks, respectively. Compared with some classical and state-of-the-art methods, the proposed model achieves optimal results, which reveals its effectiveness and feasibility. [ABSTRACT FROM AUTHOR]

Accurate medical image segmentation is critical for disease quantification and treatment evaluation. While traditional U-Net architectures and their transformer-integrated variants excel in automated segmentation tasks. Existing models also struggle with parameter efficiency and computational complexity, often due to the extensive use of Transformers. However, they lack the ability to harness the image’s intrinsic position and channel features. Research employing Dual Attention mechanisms of position and channel have not been specifically optimized for the high-detail demands of medical images. To address these issues, this study proposes a novel deep medical image segmentation framework, called DA-TransUNet, aiming to integrate the Transformer and dual attention block (DA-Block) into the traditional U-shaped architecture. Also, DA-TransUNet tailored for the high-detail requirements of medical images, optimizes the intermittent channels of Dual Attention (DA) and employs DA in each skip-connection to effectively filter out irrelevant information. This integration significantly enhances the model’s capability to extract features, thereby improving the performance of medical image segmentation. DA-TransUNet is validated in medical image segmentation tasks, consistently outperforming state-of-the-art techniques across 5 datasets. In summary, DA-TransUNet has made significant strides in medical image segmentation, offering new insights into existing techniques. It strengthens model performance from the perspective of image features, thereby advancing the development of high-precision automated medical image diagnosis. The codes and parameters of our model will be publicly available at https://github.com/SUN-1024/DA-TransUnet.

Oil spill detection is an important task for protecting and minimizing the harmful effects of oil on the marine ecosystem. Currently, the application of images from unmanned aerial vehicles, along with deep learning, is widely employed. Although these methods have yielded good results, the issue of oil spill classification based on these methods has not received much attention at present. In this research, a deep learning model with a dual attention mechanism consisting of two modules was utilized. The first module focuses on capturing the spatial relationships between each pixel and the entire image, the second module identifies the characteristics between channels in the image, thereby enhancing the ability to detect and classify oil. Additionally, a data augmentation technique based on the Generative Adversarial Networks model was refined and employed to improve the model's accuracy. Experimental results, obtained through comparisons between dataset construction methods, the use of different encoders and decoders, and adjustments hyperparameters, reveal that the best model achieves a mean Intersection over Union by 72.49%. Data augmentation techniques also contribute to a 2.56% increase in mean Intersection over Union. The findings of this research provide a feasible solution not only for detecting but also for classifying oil spills, thereby assisting marine environmental managers in making timely decisions to respond to oil spill accidents.

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.

Plant disease control has long been a critical issue in agricultural production and relies heavily on the identification of plant diseases, but traditional disease identification requires extensive experience. Most of the existing deep learning-based plant disease classification methods run on high-performance devices to meet the requirements for classification accuracy. However, agricultural applications have strict cost control and cannot be widely promoted. This paper presents a novel method for plant disease classification using a binary neural network with dual attention (DABNN), which can save computational resources and accelerate by using binary neural networks, and introduces a dual-attention mechanism to improve the accuracy of classification. To evaluate the effectiveness of our proposed approach, we conduct experiments on the PlantVillage dataset, which includes a range of diseases. The F 1 s c o r e and A c c u r a c y of our method reach 99.39% and 99.4%, respectively. Meanwhile, compared to AlexNet and VGG16, the C o m p u t a t i o n a l c o m p l e x i t y of our method is reduced by 72.3% and 98.7%, respectively. The P a r a m s s i z e of our algorithm is 5.4% of AlexNet and 2.3% of VGG16. The experimental results show that DABNN can identify various diseases effectively and accurately. [ABSTRACT FROM AUTHOR]

Copyright of Experimental Technology & Management is the property of Experimental Technology & Management Editorial Office and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Blood vessel segmentation in fundus images is necessary for the diagnosis of ophthalmic diseases. In recent years, deep learning has achieved eminent performance in blood vessel segmentation, and there still exist challenges to reduce misidentification and improve microvascular segmentation accuracy. One reason is that traditional Convolutional Neural Network (CNN) can not effectively extract multiscale information and discard the unnecessary information. Another reason is we can’t explain why some blood vessels fail to be identified. On the one hand, this paper proposes a Dual Attention Res2UNet (DA-Res2UNet) model. The DA-Res2UNet model uses Res2block rather than CNN to obtain more multiscale information and adds Dual Attention to help the model focus on important information and discard unnecessary information. On the other hand, the explainable method based on a pre-trained fundus image generator is adopted to explore how the model identifies blood vessels. We deduce several special situations that lead to the misidentification based on the model’s explanation and adjust the dataset for these special cases. The adjusted datasets significantly reduce the misidentification in the CHASE_DB1 dataset. Finally, the model trained by the adjusted datasets achieves the state-of-the-art F1-score of 81.88%, 82.77%, and 83.96% on the CHASE_DB1, DRIVE and STARE datasets, respectively.

Unmanned Aerial Vehicle (UAV) inspection of transmission channels in mountainous areas is susceptible to non-homogeneous fog, such as up-slope fog and advection fog, which causes crucial portions of transmission lines or towers to become fuzzy or even wholly concealed. This paper presents a Dual Attention Level Feature Fusion Multi-Patch Hierarchical Network (DAMPHN) for single image defogging to address the bad quality of cross-level feature fusion in Fast Deep Multi-Patch Hierarchical Networks (FDMPHN). Compared with FDMPHN before improvement, the Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM) of DAMPHN are increased by 0.3 dB and 0.011 on average, and the Average Processing Time (APT) of a single picture is shortened by 11%. Additionally, compared with the other three excellent defogging methods, the PSNR and SSIM values DAMPHN are increased by 1.75 dB and 0.022 on average. Then, to mimic non-homogeneous fog, we combine the single picture depth information with 3D Berlin noise to create the UAV-HAZE dataset, which is used in the field of UAV power assessment. The experiment demonstrates that DAMPHN offers excellent defogging results and is competitive in no-reference and full-reference assessment indices. [ABSTRACT FROM AUTHOR]

BACKGROUND: Chest CT scan is an effective way to detect and diagnose COVID-19 infection. However, features of COVID-19 infection in chest CT images are very complex and heterogeneous, which make segmentation of COVID-19 lesions from CT images quite challenging. OBJECTIVE: To overcome this challenge, this study proposes and tests an end-to-end deep learning method called dual attention fusion UNet (DAF-UNet). METHODS: The proposed DAF-UNet improves the typical UNet into an advanced architecture. The dense-connected convolution is adopted to replace the convolution operation. The mixture of average-pooling and max-pooling acts as the down-sampling in the encoder. Bridge-connected layers, including convolution, batch normalization, and leaky rectified linear unit (leaky ReLU) activation, serve as the skip connections between the encoder and decoder to bridge the semantic gap differences. A multiscale pyramid pooling module acts as the bottleneck to fit the features of COVID-19 lesion with complexity. Furthermore, dual attention feature (DAF) fusion containing channel and position attentions followed the improved UNet to learn the long-dependency contextual features of COVID-19 and further enhance the capacity of the proposed DAF-UNet. The proposed model is first pre-trained on the pseudo label dataset (generated by Inf-Net) containing many samples, then fine-tuned on the standard annotation dataset (provided by the Italian Society of Medical and Interventional Radiology) with high-quality but limited samples to improve performance of COVID-19 lesion segmentation on chest CT images. RESULTS: The Dice coefficient and Sensitivity are 0.778 and 0.798 respectively. The proposed DAF-UNet has higher scores than the popular models (Att-UNet, Dense-UNet, Inf-Net, COPLE-Net) tested using the same dataset as our model. CONCLUSION: The study demonstrates that the proposed DAF-UNet achieves superior performance for precisely segmenting COVID-19 lesions from chest CT scans compared with the state-of-the-art approaches. Thus, the DAF-UNet has promising potential for assisting COVID-19 disease screening and detection. [ABSTRACT FROM AUTHOR]

The classification of heartbeats is important for the diagnosis of arrhythmia, and more and more attention has been paid to deep learning methods to avoid manual feature design and extraction. Many methods are proposed based on the two-dimensional image representation of heartbeat and achieve good results, but these methods usually have a large number of parameters and high computational cost, which are not conducive to the real-world applications (e.g. wearable health monitoring). To address the problem, we propose a lightweight and high-performance inter-patient heartbeat classification method with dual attention mechanism. Specifically, we design a lightweight residual block, which is mainly composed of residual and depthwise separable convolution, for reducing the number of parameters. The dual attention mechanism is implemented by combining the "Squeeze-and-Excitation" (SE) module and a new "Convolutional Squeeze-and-Excitation" (Conv-SE) module. Two different kinds of channel attention are fused to improve the performance of the model. In addition, an effective image-based data augmentation procedure for 2-D representation of heartbeats is developed to overcome the data imbalance problem. We evaluated the proposed method on the MIT-BIH Arrhythmia Database and the Supraventricular Arrhythmia Database. Our method achieved an average classification accuracy of 98.68% and an average F1-score of 0.9202 on the MIT-BIH Arrhythmia database, and achieved an average classification accuracy of 97.46% and an average F1-score of 0.8984 on the Supraventricular Arrhythmia Database. Compared with other methods, the proposed method achieved best classification performance, and the number of parameters and the computational complexity were reduced by more than 70%. [ABSTRACT FROM AUTHOR]

Image super-resolution (SR) is an important image processing technique in computer vision. Although the convolutional neural network has developed rapidly and made some breakthroughs in the field of super-division, there are still some problems when images are magnified at large upscaling factors. Recently, generative adversarial network is popular, but the structural similarity (SSIM) between the super-resolution (SR) image generated by GAN network and high-resolution (HR) image is always unsatisfactory. In this paper, we propose a pixel-level self-paced adversarial network with multiple attention (PSPA) method to reduce the noise of SR image and increase its structural similarity with HR image. The combination of multiple attentions makes the model grasp the global information and restore the detail texture more accurately. The PSPA network can make the model notice the position with a large difference between the pixel values of SR and HR images and speed up the gradient descent speed. Our method shows excellent performance on Set5, Set14 and BSD100 datasets and overcomes many popular algorithms. [ABSTRACT FROM AUTHOR]

Anomaly detection is a crucial research field in computer vision with diverse applications in practical scenarios. The common anomaly detection methods employed currently consist of autoencoders, generative adversarial networks, and knowledge distillation (KD) models. However, the teacher and student models in KD might not always yield distinct representations to signify anomalies due to their similar model structure and data flow. This study proposes a novel autoencoder-like KD model based on the attention mechanism for anomaly detection. The pre-trained teacher model incorporates a dual attention module as the encoder, while the student model integrates the same dual attention module as the decoder. The teacher guides the student to learn the feature knowledge of the input image. To connect the teacher-student model, a BottleNeck module is employed, converting the features extracted from the teacher model into more compact latent codes for precise restoration by the student model, thereby achieving anomaly detection. In general, the proposed model exhibits superior performance compared to other existing anomaly detection models on specific datasets. Experimental results demonstrate that the proposed model attains the state-of-the-art (SOTA) performance in anomaly detection on the public dataset MVTec. It achieves an average AUC of 98.2% and 98.0% at sample and pixel levels, respectively.

Accurate tropical cyclone (TC) intensity and size estimation are key in disaster management and prevention. While great breakthroughs have been made in TC intensity estimation research, there is currently a lack of research on TC size reflecting TC influence radius. Therefore, we propose a lightweight multi-task learning model (TC-MTLNet) with adaptive loss balance to simultaneously estimate TC intensity and size. Adaptive loss balance is utilized to solve the problem of inconsistent convergence speed of TC intensity and size estimation tasks. The model based on four 2-D convolutions, four 3-D convolutions and three fully connected layers takes up less computational and storage space and improves the accuracy of TC intensity and size estimation by sharing knowledge among multiple tasks. In addition, due to the imbalanced distribution of TC samples, with significantly few low-intensity and high-intensity TC satellite data, this phenomenon poses a great challenge to TC intensity and size estimation. So, we utilize the influence of nearby samples to calibrate the sample density to weight the loss function to enable the model to be generalized to all samples. The result shows that the root-mean-square error (RMSE) of TC intensity estimation is $\text{8.40}\,\text{kts}$, which is 33.5% lower than that of the Advanced Dvorak Technique (ADT) and 11.4% lower than that of the deep learning method (3DAttentionTCNet). The mean absolute error (MAE) of the TC size estimation is $\text{20.89}\,\text{nmi}$, which is a 16% reduction compared to the Multi-Platform Tropical Cyclone Surface Winds Analysis (MTCSWA).

With the growing demand for video content analysis, sports video activity recognition has wide application prospects and commercial value, such as computer-assisted highlight extraction, tactic statistics and strategic analysis. Volleyball group activity recognition focuses on understanding the action performed by a group of players in volleyball matches. However, due to the cluttered backgrounds and the complex relationships between individuals in volleyball video, group activity recognition for sports video has become a significant and challenging issue. Therefore, we propose a dual attention based on a spatial-temporal inference network for volleyball group activity recognition. First, a dual attention model composed of spatial attention and mixture channel attention is proposed to assign attention weight dynamically to each feature and concern on the interdependencies of group members. It can improve the capacity of the model to distinguish features representation with intra-class variation by obtaining rich contextual relationships. Next, to focus on individual spatial-temporal information, an individual spatial-temporal inference network (ISTIN) is designed to capture person-group interactions for emphasizing the variability of these information. Finally, these features are fed into a recurrent neural network to capture temporal dependencies and make the classification. Experimental results show that this approach can be effective in group activity recognition, with our model improving recognition rates over baseline method on the benchmark datasets: Volleyball dataset and Collective Activity dataset. [ABSTRACT FROM AUTHOR]

Blood vessel segmentation in fundus images is necessary for the diagnosis of ophthalmic diseases. In recent years, deep learning has achieved eminent performance in blood vessel segmentation, and there still exist challenges to reduce misidentification and improve microvascular segmentation accuracy. One reason is that traditional Convolutional Neural Network (CNN) can not effectively extract multiscale information and discard the unnecessary information. Another reason is we can't explain why some blood vessels fail to be identified. On the one hand, this paper proposes a Dual Attention Res2UNet (DA-Res2UNet) model. The DA-Res2UNet model uses Res2block rather than CNN to obtain more multiscale information and adds Dual Attention to help the model focus on important information and discard unnecessary information. On the other hand, the explainable method based on a pre-trained fundus image generator is adopted to explore how the model identifies blood vessels. We deduce several special situations that lead to the misidentification based on the model's explanation and adjust the dataset for these special cases. The adjusted datasets significantly reduce the misidentification in the CHASE_DB1 dataset. Finally, the model trained by the adjusted datasets achieves the state-of-the-art F1-score of 81.88%, 82.77%, and 83.96% on the CHASE_DB1, DRIVE and STARE datasets, respectively. [ABSTRACT FROM AUTHOR]

In natural language processing, event detection is a critical step in event extraction, aiming to detect the occurrences of events and categorize them. Currently, the defects of Chinese event detection based on triggers include polysemous triggers and trigger-word mismatches, which reduce the accuracy of event detection models. Therefore, event detection without triggers based on dual attention (EDWTDA), a trigger-free model that can skip the trigger identification process and determine event types directly, is proposed to fix the problems mentioned above. EDWTDA adopts a dual attention mechanism, integrating local and global attention. Local attention captures key semantic information in sentences and simulates hidden event trigger words to solve the problem of trigger-word mismatch, while global attention digs for the context of documents, fixing the problem of polysemous triggers. Besides, event detection is transformed into a binary classification task to avoid problems caused by multiple tags. Meanwhile, the sample imbalance brought about by the transformation is settled with the application of the focal loss function. The experimental results on the ACE 2005 Chinese corpus show that, compared with the best baseline model, JMCEE, the accuracy rate, recall rate, and F1-score of the proposed model increased by 3.40%, 3.90%, and 3.67%, respectively. [ABSTRACT FROM AUTHOR]

Copyright of Journal of Computer Engineering & Applications is the property of Beijing Journal of Computer Engineering & Applications Journal Co Ltd. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Copyright of Journal of Computer Engineering & Applications is the property of Beijing Journal of Computer Engineering & Applications Journal Co Ltd. and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

The inductive relation prediction of knowledge graphs, as an important research topic, aims at predicting the missing relation between unknown entities with many real-world applications. Existing approaches toward this problem mostly use enclosing subgraphs to extract the features of target nodes to make predictions; however, there is a tendency to ignore the neighboring relations outside the enclosing subgraph, thus leading to inaccurate predictions. In addition, they also neglect the rich commonsense information that can help filter out less convincing results. In order to address the above issues, this paper proposes a commonsense-guided inductive relation prediction method with a dual attention mechanism called CNIA. Specifically, in addition to the enclosing subgraph, we added the multi-hop neighboring relations of target nodes, thereby forming a neighbor-enriched subgraph where the initial embeddings are generated. Next, we obtained the subgraph representations with a dual attention (i.e., edge-aware and relation-aware) mechanism, as well as the neighboring relational path embeddings. Then, we concatenated the two embeddings before feeding them into the supervised learning model. A commonsense re-ranking mechanism was introduced to filter the results that conformed to commonsense. Extensive experiments on WN18RR, FB15k-237, and NELL995 showed that CNIA achieves better prediction results when compared to the state-of-the-art models. The results suggested that our proposed model can be considered as an effective and state-of-the-art solution for inductive relation prediction.

Radar Automatic Target Recognition (RATR) based on high-resolution range profile (HRRP) has received intensive attention in recent years. In practice, RATR usually needs not only to recognize in-library samples but also to reject out-of-library samples. However, most rejection methods lack a specific and accurate description of the underlying distribution of HRRP, which limits the effectiveness of the rejection task. Therefore, this paper proposes a novel rejection method for HRRP, named Deep Multi-modal Support Vector Data Description (DMMSVDD). On the one hand, it forms a more compact rejection boundary with the Gaussian mixture model in consideration of the high-dimensional and multi-modal structure of HRRP. On the other hand, it captures the global temporal information and channel-dependent information with a dual attention module to gain more discriminative structured features, which are optimized jointly with the rejection boundary. In addition, a semi-supervised extension is proposed to refine the boundary with available out-of-library samples. Experimental results based on measured data show that the proposed methods demonstrate significant improvement in the HRRP rejection performance.

The surge in malware threats propelled by the rapid evolution of the internet and smart device technology necessitates effective automatic malware classification for robust system security. While existing research has primarily relied on some feature extraction techniques, issues such as information loss and computational overhead persist, especially in instruction-level tracking. To address these issues, this paper focuses on the nuanced analysis of API (Application Programming Interface) call sequences between the malware and system and introduces TTDAT (Two-step Training Dual Attention Transformer) for malware classification. TTDAT utilizes Transformer architecture with original multi-head attention and an integrated local attention module, streamlining the encoding of API sequences and extracting both global and local patterns. To expedite detection, we introduce a two-step training strategy: ensemble Transformer models to generate class representation vectors, thereby bolstering efficiency and adaptability. Our extensive experiments demonstrate TTDAT’s effectiveness, showcasing state-of-the-art results with an average F1 score of 0.90 and an accuracy of 0.96.

Change detection (CD) in high-resolution remote sensing images (RSIs) can be regarded as a binary visual recognition problem. Metric learning (ML) is a reliable method to determine pixel class attributes based on discriminative distance function between learnable image features. However, most related works learn discriminative distance functions in single-scale feature pairs, which suffer from slow convergence and poor local optima, partially due to that the loss function employs only large-scale feature samples while not interacting with the other scale features. Furthermore, more effective features are a prerequisite for improving the performance of ML-based CD. Hence, we propose a novel hierarchical metric learning network (HMLNet) with dual attention for CD in RSIs, where the key point is that hierarchical metric learning is performed in an ensemble manner to improve detection accuracy and accelerate model convergence. Specifically, based on the features extracted from the encoder-decoder backbone, we construct a feature pyramid to handle the complex details of objects at various scales in RSIs, and then perform metric learning between the paired pyramid features at the same scale. In addition, the dual-attention module is proposed to enhance the internal consistency of changed objects and effectively obtain more detailed information by acting on multi-scale pyramid features. Extensive experiments on the two public RSIs CD datasets, and the results demonstrate that the proposed HMLNet can accurately locate changed objects, which consistently outperforms the state-of-the-art CD competitors. [ABSTRACT FROM AUTHOR]

Microblog is an important platform for mining public opinion, and it is of great value to conduct emotional analysis of microblog texts during the current epidemic. Aiming at the problem that most current emotional classification methods cannot effectively extract deep text features, and that traditional word vectors cannot dynamically obtain the semantics of words according to their context, which leads to classification bias, this research put forward a microblog text emotion classification algorithm based on TCN-BiGRU and dual attention (TCN-BiGRU-DATT). First, the vector representation of the text was obtained using ALBERT. Second, the TCN and BiGRU networks were used to extract the emotional information contained in the text through dual pathway feature extraction, to efficiently obtain the deep semantic features of the text. Then, the dual attention mechanism was introduced to allocate the global weight of the key information in the semantic features, and the emotional features were spliced and fused. Finally, the Softmax classifier was applied for emotion classification. The findings of a comparative experiment on a set of microblog text comments collected throughout the pandemic revealed that the accuracy, recall, and F1 value of the emotion classification method proposed in this paper reached 92.33%, 91.78%, and 91.52%, respectively, which was a significant improvement compared with other models. [ABSTRACT FROM AUTHOR]

To accurately detect multi-scale remote sensing objects in complex backgrounds, we propose a novel transformer-based adaptive object detection method. The backbone network of the method is a dual attention vision transformer network that utilizes spatial window attention and channel group attention to capture feature interactions between different objects in complex scenes. We further design an adaptive path aggregation network. In the designed network, CBAM (Convolutional Block Attention Module) is utilized to suppress background information in the fusion paths of different-level feature maps, and new paths are introduced to fuse same-scale feature maps to increase the feature information of the feature maps. The designed network can provide more effective feature information and improve the feature representation capability. Experiments conducted on the three datasets of RSOD, NWPU VHR-10, and DIOR show that the mAP of our method is 96.9%, 96.6%, and 81.7%, respectively, which outperforms compared object detection methods. The experimental results show that our method can detect remote-sensing objects better. [ABSTRACT FROM AUTHOR]

Accurate segmentation of the left atrial structure using magnetic resonance images provides an important basis for the diagnosis of atrial fibrillation (AF) and its treatment using robotic surgery. In this study, an image segmentation method based on sequence relationship learning and multi-scale feature fusion is proposed for 3D to 2D sequence conversion in cardiac magnetic resonance images and the varying scales of left atrial structures within different slices. Firstly, a convolutional neural network layer with an attention module was designed to extract and fuse contextual information at different scales in the image, to strengthen the target features using the correlation between features in different regions within the image, and to improve the network's ability to distinguish the left atrial structure. Secondly, a recurrent neural network layer oriented to two-dimensional images was designed to capture the correlation of left atrial structures in adjacent slices by simulating the continuous relationship between sequential image slices. Finally, a combined loss function was constructed to reduce the effect of positive and negative sample imbalance and improve model stability. The Dice, IoU, and Hausdorff distance values reached 90.73%, 89.37%, and 4.803 mm, respectively, based on the LASC2013 (left atrial segmentation challenge in 2013) dataset; the corresponding values reached 92.05%, 89.41% and 9.056 mm, respectively, based on the ASC2018 (atrial segmentation challenge at 2018) dataset. [ABSTRACT FROM AUTHOR]

Copyright of Automobile Technology is the property of Automobile Technology Editorial Office and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

Remote sensing images can obtain broad geomorphic features and provide a strong basis for analysis and decision making. As 71% of the earth is covered by water, shipping has become an efficient means of international trade and transportation, and the development level of coastal cities will directly reflect the development level of a country. The coastline is the boundary line between seawater and land, so it is of great significance to accurately identify it to assist shipping traffic and docking, and this identification will also play a certain auxiliary role in environmental analysis. Currently, the main problems of coastline recognition conducted by remote sensing images include: (1) in the process of remote sensing, image transmission inevitably brings noise causing poor image quality and difficult image quality enhancement; (2) s single scale does not allow for the identification of coastlines at different scales; and (3) features are under-utilized, false detection is high and intuitive measurement is difficult. To address these issues, we used the following multispectral methods: (1) a PCA-based image enhancement algorithm was proposed to improve image quality; (2) a dual attention network and HRnet network were proposed to extract suspected coastlines from different levels; and (3) a decision set fusion approach was proposed to transform the coastline identification problem into a probabilistic problem for coastline extraction. Finally, we constructed a coastline straightening model to visualize and analyze the recognition effect. Experiments showed that the algorithm has an AOM greater than 0.88 and can achieve coastline extraction. [ABSTRACT FROM AUTHOR]

Infrared and visible image fusion aims at generating a fused image containing the intensity and detail information of source images, and the key issue is effectively measuring and integrating the complementary information of multi-modality images from the same scene. Existing methods mostly adopt a simple weight in the loss function to decide the information retention of each modality rather than adaptively measuring complementary information for different image pairs. In this study, we propose a multi-scale dual attention (MDA) framework for infrared and visible image fusion, which is designed to measure and integrate multi-scale complementary information in both structure and loss function at the image and patch level. In our method, the residual downsample block decomposes source images into three scales first. Then, dual attention fusion block integrates complementary information and generates a spatial and channel attention map at the same and adjacent scale for feature fusion. Finally, the output image is reconstructed by the residual reconstruction block. Loss function consists of image-level, feature-level and patch-level three parts, of which the calculation of the image-level and patch-level two parts are based on the weights generated by the complementary information measurement. Indeed, to constrain the pixel intensity distribution between the output and infrared image, a style loss is added. Our fusion results perform robust and informative across different scenarios. Qualitative and quantitative results on three datasets illustrate that our method is able to preserve both thermal radiation and detailed information from two modalities and achieve comparable results compared with the other state-of-the-art methods. Ablation experiments show the effectiveness of our information integration architecture and adaptively measure complementary information retention in the loss function. Our code is available at https://github.com/SSyangguang/MDA. [ABSTRACT FROM AUTHOR]

Accurately predicting blood glucose levels is crucial in diabetes management to mitigate patients' risk of complications. However, blood glucose values exhibit instability, and existing prediction methods often struggle to capture their volatile nature, leading to inaccurate trend forecasts. To address these challenges, we propose a novel blood glucose level prediction model based on the Informer architecture: BGformer. Our model introduces a feature enhancement module and a microscale overlapping concerns mechanism. The feature enhancement module integrates periodic and trend feature extractors, enhancing the model's ability to capture relevant information from the data. By extending the feature extraction capacity of time series data, it provides richer feature representations for analysis. Meanwhile, the microscale overlapping concerns mechanism adopts a window-based strategy, computing attention scores only within specific windows. This approach reduces computational complexity while enhancing the model's capacity to capture local temporal dependencies. Furthermore, we introduce a dual attention enhancement module to augment the model's expressive capability. Through prediction experiments on blood glucose values from sixteen diabetic patients, our model outperformed eight benchmark models in terms of both MAE and RMSE metrics for future 60-minute and 90-minute predictions. Our proposed scheme significantly improves the model's dependency-capturing ability, resulting in more accurate blood glucose level predictions. [Display omitted] [ABSTRACT FROM AUTHOR]

Recently, hyperspectral image classification has made great progress with the development of convolutional neural networks. However, due to the challenges of distribution shifts and data redundancies, the classification accuracy is low. Some existing domain adaptation methods try to mitigate the distribution shifts by training source samples and some labeled target samples. However, in practice, labeled target domain samples are difficult or even impossible to obtain. To solve the above challenges, we propose a novel dual-attention deep discriminative domain generalization framework (DAD3GM) for cross-scene hyperspectral image classification without training the labeled target samples. In DAD3GM, we mainly design two blocks: dual-attention feature learning (DAFL) and deep discriminative feature learning (DDFL). DAFL is designed to extract spatial features by multi-scale self-attention and extract spectral features by multi-head external attention. DDFL is further designed to extract deep discriminative features by contrastive regularization and class discrimination regularization. The combination of DAFL and DDFL can effectively reduce the computational time and improve the generalization performance of DAD3GM. The proposed model achieves 84.25%, 83.53%, and 80.63% overall accuracy on the public Houston, Pavia, and GID benchmarks, respectively. Compared with some classical and state-of-the-art methods, the proposed model achieves optimal results, which reveals its effectiveness and feasibility.