Attention Fusion
注意力融合

Attention Fusion(注意力融合)研究综述

Attention Fusion 注意力融合 - Unlike present works, this article presents a hierarchical self-attention fusion (H-SATF) model for capturing contextual information better among utterances, a contextual self-attention temporal convolutional network (CSAT-TCN) for sentiment recognition in the social Internet of Things, and a multibranch memory (MBM) network that stores self-speaker and interspeaker sentimental states into global memories. ^[1] Besides, we introduce a double-attention fusion (DAF) block to fuse the low-level and high-level features efficiently. ^[2] Finally, the two networks are optimized jointly through attention fusion. ^[3] A video multimodal emotion recognition method based on Bi-GRU and attention fusion is proposed in this paper. ^[4] In this study, we introduce a general framework to automatically diagnose different types of WSIs via unit stochastic selection and attention fusion. ^[5]
与目前的工作不同，本文提出了一种分层自注意力融合 (H-SATF) 模型，用于更好地捕获话语之间的上下文信息，一种用于社交物联网中情感识别的上下文自注意力时间卷积网络 (CSAT-TCN)，和一个多分支记忆（MBM）网络，将自我说话者和说话者之间的情感状态存储到全局记忆中。 ^[1] 此外，我们引入了双注意力融合（DAF）块来有效地融合低级和高级特征。 ^[2] 最后，通过注意力融合对两个网络进行联合优化。 ^[3] 本文提出了一种基于Bi-GRU和注意力融合的视频多模态情感识别方法。 ^[4] 在这项研究中，我们引入了一个通用框架，通过单元随机选择和注意力融合自动诊断不同类型的 WSI。 ^[5]

Channel Attention Fusion 渠道注意力融合

For the generator, a set of recursive modules including a physical degradation model and a multiscale residual channel attention fusion module integrate the spectral-spatial difference information between input images and estimated degradation images to restore the details of the fused image. ^[1] Secondly, a novel channel attention fusion strategy is proposed to improve the feature fusion and network ability. ^[2] To address the above problem, we propose a novel dual-channel scale-aware segmentation network with position and channel attentions (DSPCANet) for high-resolution aerial images, which contains an Xception branch and a digital surface model-based position and channel attention fusion (DSMPCF) branch to process the near-infrared, red, and green (IRRG) spectral images and DSM images, respectively. ^[3] In this paper, we proposed an effective deep convolutional neural network model of multi-scale spatial and channel attention fusion (MS-SCANet) for underwater image enhancement. ^[4]
对于生成器，一组递归模块包括物理退化模型和多尺度残差通道注意力融合模块，整合输入图像和估计退化图像之间的光谱空间差异信息，以恢复融合图像的细节。 ^[1] 其次，提出了一种新颖的通道注意力融合策略，以提高特征融合和网络能力。 ^[2] 为了解决上述问题，我们提出了一种新颖的具有位置和通道注意的双通道尺度感知分割网络 (DSPCANet)，用于高分辨率航空图像，其中包含一个 Xception 分支和一个基于数字表面模型的位置和通道注意融合(DSMPCF) 分支分别处理近红外、红色和绿色 (IRRG) 光谱图像和 DSM 图像。 ^[3] 在本文中，我们提出了一种用于水下图像增强的有效的多尺度空间和通道注意力融合深度卷积神经网络模型（MS-SCANet）。 ^[4]

Residual Attention Fusion 残余注意力融合

Therefore, we proposed a residual attention fusion network (RAFN), which is an improved residual fusion (RF) framework, to effectively extract hierarchical features for use in single-image super-resolution. ^[1] Furthermore, we propose a residual attention fusion block (RAFB), whose purpose is to simultaneously focus on meaningful low-level feature maps and spatial location information. ^[2] The LFI module is a cascade of several dual residual attention fusion (DRAF) blocks with a dense connection structure. ^[3]
因此，我们提出了一种残差注意力融合网络（RAFN），它是一种改进的残差融合（RF）框架，可以有效地提取分层特征以用于单图像超分辨率。 ^[1] 此外，我们提出了一种残差注意力融合块（RAFB），其目的是同时关注有意义的低级特征图和空间位置信息。 ^[2] LFI 模块是几个具有密集连接结构的对偶残差注意力融合 (DRAF) 块的级联。 ^[3]

Spatial Attention Fusion 空间注意力融合

To address the aforementioned issues, a spectral–spatial attention fusion with a deformable convolution residual network (SSAF-DCR) is proposed for hyperspectral image classification. ^[1] Then we introduce spatial attention fusion mechanism for the adaptive fusion of features derived from edge maps and blurry images, and those representing shallow fine-grained details and in-depth abstract features. ^[2] To promote diversity in the contextual information, a spatial attention fusion mechanism is introduced to capture many local regions of interest in the attention fusion module. ^[3]
为了解决上述问题，提出了一种具有可变形卷积残差网络（SSAF-DCR）的光谱-空间注意力融合方法，用于高光谱图像分类。 ^[1] 然后我们引入空间注意力融合机制，用于自适应融合边缘图和模糊图像的特征，以及表示浅层细粒度细节和深度抽象特征的特征。 ^[2] 为了促进上下文信息的多样性，引入了空间注意力融合机制来捕获注意力融合模块中的许多局部感兴趣区域。 ^[3]

Scale Attention Fusion

To address the problem, we proposed a multi-modality and multi-scale attention fusion network for land cover classification from VHR remote sensing images. ^[1] For capturing more spatial information with multi-scale semantic features, in the generator network, we insert a multi-scale attention fusion (MSAF) module between the encoder and decoder paths. ^[2]
为了解决这个问题，我们提出了一种多模态和多尺度的注意力融合网络，用于 VHR 遥感图像的土地覆盖分类。 ^[1] 为了捕捉更多具有多尺度语义特征的空间信息，在生成网络中，我们在编码器和解码器路径之间插入了一个多尺度注意力融合（MSAF）模块。 ^[2]

Temporal Attention Fusion

Particularly, we propose a novel Bi-directional Spatial Temporal Attention Fusion Generative Adversarial Network (STA-GAN) to learn both spatial and temporal information to generate egocentric video sequences from the exocentric view. ^[1] Finally, a temporal attention fusion is employed for dynamically integrating all these parts. ^[2]
特别是，我们提出了一种新颖的双向时空注意力融合生成对抗网络（STA-GAN）来学习空间和时间信息，以从外中心视图生成以自我为中心的视频序列。 ^[1] 最后，采用时间注意力融合来动态整合所有这些部分。 ^[2]

Modal Attention Fusion

IMAN introduces a cross-modal attention fusion module to capture cross-modal interactions of multimodal information, and employs a conversational modeling module to explore the context information and speaker dependency of the whole conversation. ^[1] Furthermore, a multi-modal attention fusion layer is applied to assign weights to different parts of each modality of source code and then integrate them into a single hybrid representation. ^[2]
nan ^[1] 此外，应用多模态注意力融合层为源代码的每个模态的不同部分分配权重，然后将它们集成到单个混合表示中。 ^[2]

Hierarchical Attention Fusion

To address this limitation, we introduce a hierarchical attention fusion network using multi-scale features for geo-localization. ^[1] The HRR module is proposed to adaptively integrate multi-level features through learning, acting as a hierarchical attention fusion mechanism. ^[2]
为了解决这个限制，我们引入了一个分层注意力融合网络，使用多尺度特征进行地理定位。 ^[1] HRR 模块被提出通过学习自适应地集成多级特征，充当分层注意力融合机制。 ^[2]

Visual Attention Fusion 视觉注意力融合

Meanwhile, we present a visual attention fusion (VAF) framework to enhance faces and lines in the saliency map, which are observed to be highly sensitive in the human visual system (HVS). ^[1] In this paper, we propose a cross visual attention fusion dual-path neural network with dual-constrained marginal ranking(DCAF) to solve the problem. ^[2]
同时，我们提出了一个视觉注意融合（VAF）框架来增强显着图中的人脸和线条，这些在人类视觉系统（HVS）中被认为是高度敏感的。 ^[1] 在本文中，我们提出了一种具有双约束边际排序（DCAF）的交叉视觉注意融合双路径神经网络来解决该问题。 ^[2]

Adaptive Attention Fusion 自适应注意力融合

In the first stage, our multi-branch feature extraction network utilizes Adaptive Attention Fusion (AAF) modules to produce cross-modal fusion features from single-modal semantic features. ^[1] To address the above problems, we propose a special attention module called the Dual Crisscross Attention (DCCA) module for road extraction, which consists of the CCA module, Rotated Crisscross Attention (RCCA) module and Self-adaptive Attention Fusion (SAF) module. ^[2]
在第一阶段，我们的多分支特征提取网络利用自适应注意融合 (AAF) 模块从单模态语义特征中生成跨模态融合特征。 ^[1] 为了解决上述问题，我们提出了一种特殊的注意力模块，称为双交叉注意力（DCCA）模块，用于道路提取，它由CCA模块、旋转交叉注意力（RCCA）模块和自适应注意力融合（SAF）模块组成。 ^[2]

attention fusion module 注意力融合模块

For the generator, a set of recursive modules including a physical degradation model and a multiscale residual channel attention fusion module integrate the spectral-spatial difference information between input images and estimated degradation images to restore the details of the fused image. ^[1] To address the first problem, we design a non-local fusion module (NFM) and an attention fusion module (AFM), and construct the multi-level pyramids' architecture to explore the local and global correlations of rain information from the rain image pyramid. ^[2] Subsequently, the gated attention fusion module (GAFM) is applied to the RGB depth (RGB-D) information to obtain advantageous structural and spatial fusion features. ^[3] An attention fusion module is constructed to combine the features extracted by the artifact correction and SR modules. ^[4] It gathers multi-level features from two feature extractions and fuses them with the Multi-perspective Attention Fusion Module (MPAFM) we propose. ^[5] This paper proposes MBA which consists of pre-trained feature extractors, text encoder, image encoder, multimodal bilinear attention fusion module, and summary decoder to complete abstractive multimodal summarization task. ^[6] The model introduces combined attention fusion module and multiscale residual fusion module in the backbone network ResNet 50 to enhance the feature flow between residual blocks and better fuse multiscale features. ^[7] Besides, an attention fusion module is investigated to further improve the NDI and NGIA features. ^[8] To combine features from two paths, we propose a novel fusion strategy named Attention Fusion Module (AFM). ^[9] In this paper, we proposed a self-attention fusion module named as SAF module which combines spatial attention and channel attention in parallel to handle this problem. ^[10] Secondly, a mutual-perception attention fusion module is designed for simulating binocular fusion. ^[11] we further propose a multi-scale decoder with a novel dual attention fusion module (DAF), which can combine both the restored and known regions in a smooth way. ^[12] Besides, we construct our network with low-high pass block and edge attention fusion module, which extract spatial and semantic information effectively to improve the power line detection result along the boundary. ^[13] In addition, to further reconstruct more details of moire patterns, this paper proposes an efficient attention fusion module (EAFM). ^[14] IMAN introduces a cross-modal attention fusion module to capture cross-modal interactions of multimodal information, and employs a conversational modeling module to explore the context information and speaker dependency of the whole conversation. ^[15] We introduce the Self-Multi-Attention Fusion module, Multi-Attention fusion module, and Video Fusion module, which are attention based multimodal fusion mechanisms using the recently proposed transformer architecture. ^[16] In this paper, we aim to utilize the features from LR and HR space more efficiently and propose the novel network, which applies a frequency-slicing mechanism to divide features into LR and HR space, a direction-aware fusion residual group to extract distinctive features in LR space and an attention fusion module to recalibrate features in HR space. ^[17] Furthermore, we designed a dual-attention fusion module that can extract and enhance the spatial relationship of magnetic resonance imaging and applied the strategy of deep supervision in different parts of the proposed network. ^[18]
对于生成器，一组递归模块包括物理退化模型和多尺度残差通道注意力融合模块，整合输入图像和估计退化图像之间的光谱空间差异信息，以恢复融合图像的细节。 ^[1] 为了解决第一个问题，我们设计了一个非局部融合模块（NFM）和一个注意力融合模块（AFM），并构建了多层金字塔的架构，以探索来自雨图像金字塔的雨信息的局部和全局相关性. ^[2] 随后，将门控注意力融合模块（GAFM）应用于RGB深度（RGB-D）信息，以获得有利的结构和空间融合特征。 ^[3] 构建了一个注意力融合模块，将伪影校正和SR模块提取的特征结合起来。 ^[4] 它从两个特征提取中收集多级特征，并将它们与我们提出的多视角注意力融合模块 (MPAFM) 融合。 ^[5] 本文提出了由预训练特征提取器、文本编码器、图像编码器、多模态双线性注意力融合模块和摘要解码器组成的 MBA，以完成抽象的多模态摘要任务。 ^[6] 该模型在骨干网络 ResNet 50 中引入了组合注意力融合模块和多尺度残差融合模块，以增强残差块之间的特征流，更好地融合多尺度特征。 ^[7] 此外，还研究了注意力融合模块以进一步改进 NDI 和 NGIA 特征。 ^[8] 为了结合来自两条路径的特征，我们提出了一种新的融合策略，称为注意力融合模块（AFM）。 ^[9] 在本文中，我们提出了一种名为 SAF 模块的自注意力融合模块，它同时结合了空间注意力和通道注意力来处理这个问题。 ^[10] 其次，设计了一个相互感知注意力融合模块，用于模拟双目融合。 ^[11] 我们进一步提出了一个多尺度解码器，它带有一个新颖的双注意力融合模块（DAF），它可以平滑地结合恢复的区域和已知区域。 ^[12] nan ^[13] nan ^[14] nan ^[15] nan ^[16] nan ^[17] nan ^[18]

attention fusion network 注意力融合网络

To solve above problem, we propose a Multi-style Attention Fusion Network (MAFNet). ^[1] We design a novel multi-attention fusion network (MAFNet) based on the self-attention mechanism to extract the spatial features related to the sound source in the video frames and fuse them into audio features well. ^[2] To address the problem, we proposed a multi-modality and multi-scale attention fusion network for land cover classification from VHR remote sensing images. ^[3] In other words, Ordered Memory Attention Fusion Network is a comprehensive captioning model with the support of residual attention network and ordered memory module. ^[4] In this paper, we propose a cross-layer attention fusion network (CLAF-CNN) to solve the problem of accurately segmenting OARs. ^[5] To address this limitation, we introduce a hierarchical attention fusion network using multi-scale features for geo-localization. ^[6] In this article, a novel classification method called the deep group spatial-spectral attention fusion network is proposed for PAN and MS images. ^[7] In addition, the MLDNN also has a novel backbone, which is made up of three blocks to extract discriminative features, and they are CNN block, bidirectional gated recurrent unit (BiGRU) block and a step attention fusion network (SAFN) block. ^[8] Therefore, we proposed a residual attention fusion network (RAFN), which is an improved residual fusion (RF) framework, to effectively extract hierarchical features for use in single-image super-resolution. ^[9] To solve this problem, we propose a Dual-way Feature attention Fusion Network (DFFNet), which consists of two branches, optical remote sensing image branch and elevation feature branch. ^[10] Objective: To explore the performance of the attention-multiple instance learning (MIL) framework, an attention fusion network-based MIL, in the automated diagnosis of chronic gastritis with multiple indicators. ^[11] To improve the accuracy of multi-spectral semantic segmentation, an attention fusion network (AFNet) based on deep learning is proposed. ^[12] In this paper, we propose an Attention Fusion Network (AFN) and Food-Ingredient Joint Learning module for fine-grained food and ingredients recognition. ^[13] In the proposed framework, an attention fusion network is utilized to amalgamate the predictions of the individual models. ^[14] Therefore, in this letter, a multiattention fusion network (MAFN) for HSI classification is proposed. ^[15] In this paper, we propose a multiple attention fusion network (MAFN) with the goal of improving emotion recognition performance by modeling human emotion recognition mechanisms. ^[16] In this paper, we propose a novel Multi-head Attention Fusion Network (MAFN) to exploit aspect-level semantic information from texts to enhance prediction performance. ^[17]
为了解决上述问题，我们提出了一种多风格注意力融合网络（MAFNet）。 ^[1] 我们基于自注意力机制设计了一种新颖的多注意力融合网络（MAFNet），以提取视频帧中与声源相关的空间特征，并将其很好地融合为音频特征。 ^[2] 为了解决这个问题，我们提出了一种多模态和多尺度的注意力融合网络，用于 VHR 遥感图像的土地覆盖分类。 ^[3] 换句话说，Ordered Memory Attention Fusion Network 是一个在残差注意力网络和有序记忆模块的支持下的综合字幕模型。 ^[4] 在本文中，我们提出了一种跨层注意力融合网络 (CLAF-CNN) 来解决准确分割 OAR 的问题。 ^[5] 为了解决这个限制，我们引入了一个分层注意力融合网络，使用多尺度特征进行地理定位。 ^[6] 在本文中，针对 PAN 和 MS 图像提出了一种称为深度组空间-光谱注意力融合网络的新分类方法。 ^[7] 此外，MLDNN 还具有一个新颖的主干，它由三个块组成，用于提取判别特征，它们是 CNN 块、双向门控循环单元 (BiGRU) 块和步进注意融合网络 (SAFN) 块。 ^[8] 因此，我们提出了一种残差注意力融合网络（RAFN），它是一种改进的残差融合（RF）框架，可以有效地提取分层特征以用于单图像超分辨率。 ^[9] 为了解决这个问题，我们提出了一种双向特征注意力融合网络（DFFNet），它由两个分支组成，光学遥感图像分支和高程特征分支。 ^[10] 目的：探讨注意力多实例学习 (MIL) 框架（一种基于注意力融合网络的 MIL）在多指标慢性胃炎自动诊断中的性能。 ^[11] nan ^[12] nan ^[13] nan ^[14] nan ^[15] 在本文中，我们提出了一种多注意力融合网络（MAFN），旨在通过对人类情感识别机制进行建模来提高情感识别性能。 ^[16] 在本文中，我们提出了一种新颖的多头注意力融合网络（MAFN）来利用文本中的方面级语义信息来提高预测性能。 ^[17]

attention fusion mechanism 注意力融合机制

To address the problems of complex water scenes with scale changes, much background information and inability to focus on key features, this paper proposes a multi-scale surface target recognition algorithm based on attention fusion mechanism. ^[1] Then we introduce spatial attention fusion mechanism for the adaptive fusion of features derived from edge maps and blurry images, and those representing shallow fine-grained details and in-depth abstract features. ^[2] To promote diversity in the contextual information, a spatial attention fusion mechanism is introduced to capture many local regions of interest in the attention fusion module. ^[3] Then, a simple co-attention fusion mechanism is used to dynamically combine information from the CSE and NSE. ^[4] In this paper, we propose a semantic attention fusion mechanism (SAF) to increase the discriminability of detector. ^[5] The HRR module is proposed to adaptively integrate multi-level features through learning, acting as a hierarchical attention fusion mechanism. ^[6]
针对复杂水景尺度变化、背景信息多、无法聚焦关键特征等问题，提出一种基于注意力融合机制的多尺度水面目标识别算法。 ^[1] 然后我们引入空间注意力融合机制，用于自适应融合边缘图和模糊图像的特征，以及表示浅层细粒度细节和深度抽象特征的特征。 ^[2] 为了促进上下文信息的多样性，引入了空间注意力融合机制来捕获注意力融合模块中的许多局部感兴趣区域。 ^[3] 然后，使用简单的共同注意融合机制来动态组合来自 CSE 和 NSE 的信息。 ^[4] 在本文中，我们提出了一种语义注意融合机制（SAF）来增加检测器的可辨别性。 ^[5] HRR 模块被提出通过学习自适应地集成多级特征，充当分层注意力融合机制。 ^[6]

attention fusion block 注意力融合块

The two feature pyramids in DoubleHigherNet consists of 1/4 resolution feature and higher-resolution (1/2) maps generated by attention fusion blocks and transposed convolutions. ^[1] For adaptively enhancing different patch with spatial and temporal information, a channel and spatial-wise attention fusion block is proposed to achieve patch-wise recalibration and fusion of spatial and temporal features. ^[2] Furthermore, we propose a residual attention fusion block (RAFB), whose purpose is to simultaneously focus on meaningful low-level feature maps and spatial location information. ^[3]
DoubleHigherNet 中的两个特征金字塔由注意力融合块和转置卷积生成的 1/4 分辨率特征和更高分辨率 (1/2) 的映射组成。 ^[1] 为了利用时空信息自适应地增强不同的patch，提出了一种通道和spatial-wise attention fusion block，以实现patch-wise recalibration和时空特征的融合。 ^[2] 此外，我们提出了一种残差注意力融合块（RAFB），其目的是同时关注有意义的低级特征图和空间位置信息。 ^[3]

attention fusion method

We show that augmenting text with image embedding information immediately leads to a boost in performance, while applying additional attention fusion methods brings further improvement. ^[1] The primary contributions of this study are the propositions of (a) a fusion attention mechanism, and (b) a multiheaded attention fusion method. ^[2]
我们表明，使用图像嵌入信息增强文本会立即提高性能，而应用额外的注意力融合方法会带来进一步的改进。 ^[1] 本研究的主要贡献是（a）融合注意力机制和（b）多头注意力融合方法的命题。 ^[2]

attention fusion layer

We attend on technique- and recipe-level representations of a user’s previously consumed recipes, fusing these ‘user-aware’ representations in an attention fusion layer to control recipe text generation. ^[1] Furthermore, a multi-modal attention fusion layer is applied to assign weights to different parts of each modality of source code and then integrate them into a single hybrid representation. ^[2]
我们关注用户先前使用的食谱的技术和食谱级别的表示，将这些“用户感知”表示融合到注意力融合层中以控制食谱文本的生成。 ^[1] 此外，应用多模态注意力融合层为源代码的每个模态的不同部分分配权重，然后将它们集成到单个混合表示中。 ^[2]

attention fusion model 注意力融合模型

In this paper, an efficient linear self-attention fusion model is proposed for the task of hyperspectral image (HSI) and LiDAR data joint classification. ^[1] In order to tackle this limitation, we propose a simple yet effective attention fusion model based on Disentangled Non-local (DNL) network for hyperspectral and LiDAR data joint classification task. ^[2]
在本文中，针对高光谱图像（HSI）和激光雷达数据联合分类的任务，提出了一种高效的线性自注意力融合模型。 ^[1] 为了解决这个限制，我们提出了一个简单而有效的基于解纠缠非局部（DNL）网络的注意力融合模型，用于高光谱和激光雷达数据联合分类任务。 ^[2]