检索结果-内蒙古大学图书馆

Spatial convolutional self-attention-based transformer module for strawberry disease identification under complex background

引用

COMPUTERS AND ELECTRONICS IN AGRICULTURE 2023年第1期212卷

作者： Li, Gaoqiang Jiao, Lin Chen, Peng Liu, Kang Wang, Rujing Dong, Shifeng Kang, Chenrui Anhui Univ Sch Internet Hefei 2300391 Anhui Peoples R China Chinese Acad Sci Hefei Inst Phys Sci Inst Intelligent Machines Hefei 230031 Peoples R China Univ Sheffield Dept Comp Sci Sheffield S10 2TN England Univ Sci & Technol China Hefei 23002631 Anhui Peoples R China

The occurrence of strawberry diseases has a huge impact on the yield and quality of strawberry fruits, resulting in huge economic losses. Real-time and effective identification and diagnosis of strawberry disease is an essential step for strawberry disease prevention. Machine learning-based methods are widely used in strawberry disease identification tasks, but these methods require expertise to design proper strawberry disease feature descriptors. Deep-learning methods have remarkably improved the capability of feature extraction. However, the strawberry disease with complex backgrounds brings great challenges for accurate feature extraction, which leads to poor recognition results of strawberry disease under complex backgrounds. In this paper, an improved transformerbased strawberry disease identification method is proposed to achieve precise and fast recognition of multiple classes of strawberry diseases. First, a multi-classes strawberry disease dataset has been constructed with 5369 images and 12 types of common strawberry disease. To increase the diversity of samples under complex backgrounds, various data augmentation strategies are introduced into the strawberry disease recognition method. Then, Multi-Head Self-Attention (MSA) is used to capture feature dependencies over long distances of strawberry disease images by leveraging the self-attention mechanism. To improve the recognition efficiency, the spatial convolutional self-attention-based transformer (SCSA-transformer) is proposed to reduce the parameters of the transformer network. The experimental results validated on the constructed strawberry disease dataset demonstrate that the recognition accuracy of the proposed method can achieve 99.10%, which outperforms other methods. Besides, we also observe that the parameters of the classification model are reduced compared with other methods, which effectively improves the recognition efficiency of strawberry diseases.

关键词： Strawberry disease Image recognition Multi-head self-attention transformer module

来源：评论

学校读者我要写书评

暂无评论

Improved GAN: Using a transformer module generator approach for material decomposition

引用

COMPUTERS IN BIOLOGY AND MEDICINE 2022年 149卷 105952-105952页

作者： Wang, Guoshuai Liu, Zhou Huang, Zhengyong Zhang, Na Luo, Honghong Liu, Lijian Shen, Hao Che, Canwen Niu, Tianye Liang, Dong Luo, Dehong Hu, Zhanli Chinese Acad Sci Shenzhen Inst Adv Technol Lauterbur Res Ctr Biomed Imaging Shenzhen 518055 Peoples R China Univ Chinese Acad Sci Beijing 100049 Peoples R China Chinese Acad Med Sci & Peking Union Med Coll Canc Hosp Natl Canc Ctr Natl Clin Res Ctr CancDept Radiol Shenzhen 518116 Peoples R China Chinese Acad Med Sci & Peking Union Med Coll Shenzhen Hosp Shenzhen 518116 Peoples R China United Imaging Res Inst Innovat Med Equipment Shenzhen 518045 Peoples R China Shenzhen Bay Lab Inst Biomed Engn Shenzhen 518118 Peoples R China Chinese Acad Sci Key Lab Hlth Informat Shenzhen 518055 Peoples R China

Dual-energy computed tomography (CT) can be used for material decomposition, allowing for the precise quantitative mapping of body substances;this has a wide range of clinical applications, including disease diagnosis, treatment response evaluation and prognosis prediction. However, dual-energy CT has not yet become the mainstream technique in most clinical settings due to its limited accessibility. To fully take advantage of material quantification, researchers have attempted to use deep learning to generate material decomposition maps from conventional single-energy CT images, mainly by synthesizing another single-energy CT image from a conventional single-energy CT image to form a dual-energy CT image first and then generate material decom-position maps. This is not a straightforward process, and it potentially introduces many inaccuracies after multiple steps. In this work, we proposed a generative adversarial network (GAN) framework as the base and improved its generator;this approach combines convolutional neural networks (CNNs) and a transformer module to directly generate material decomposition maps from conventional single-energy CT images. Our model pays attention to both local and global information. Then, we compared our method with 6 competitive deep learning methods on water (calcium) and calcium (water) substrate density image datasets. The average PSNR, SSIM, MAE, and RMSE of the generated and ground truth of the water (calcium) substrate density images were 32.7207, 0.9685, 0.0323, and 0.0555, respectively. Furthermore, the average PSNR, SSIM, MAE, and RMSE of the generated and ground truth of the calcium (water) substrate density images were 30.2823, 0.9449, 0.0652, and 0.0715, respectively. Our model achieved better performance and stronger stability than competing approaches.

关键词： Deep learning Generative adversarial network transformer module Material decomposition Dual -energy CT

来源：评论

学校读者我要写书评

暂无评论

FutuTP: Future-based trajectory prediction for autonomous driving

引用

APPLIED INTELLIGENCE 2025年第7期55卷 1-12页

作者： Xu, Qingchao Liu, Yandong Wen, Shixi Yang, Xin Zhou, Dongsheng Dalian Univ Sch Software Engn Key Lab Adv Design & Intelligent Comp Minist Educ Dalian 116622 Peoples R China Dalian Univ Technol Sch Comp Sci & Technol Dalian 116024 Peoples R China

Trajectory prediction is an essential aspect of autonomous driving technology. Based on the historical trajectories and envi-ronmental information, trajectory prediction methods predict the future trajectory of a vehicle. Goal-based methods have beensuccessful because of their excellent interpretability. However, these methods ignore future lane information and interactionsin future trajectories, which leads to prediction failures in some scenes. In this paper, we propose an encoder-decoder modelcalled future-based trajectory prediction (FutuTP). The encoder fuses the interactions of future trajectories through a trans-former module. The decoder predicts the future lane area and applies the results to generate a trajectory. The experimentalresults show that FutuTP achieves more accurate predictions than does the SOTA method on Argoverse 1. Especially in termsof the minFDE(6) metric, FutuTP outperforms the SOTA method by approximately 6%. The code can be accessed via thefollowing link https://***/Qingchao-Xu/FutuTP.

关键词： Trajectory prediction Autonomous driving Vehicle interaction modelling transformer module

来源：评论

学校读者我要写书评

暂无评论

A dual-branch infrared and visible image fusion network using progressive image-wise feature transfer

引用

JOURNAL OF VISUAL COMMUNICATION AND IMAGE REPRESENTATION 2024年 102卷

作者： Xu, Shaoping Zhou, Changfei Xiao, Jian Tao, Wuyong Dai, Tianyu Nanchang Univ Sch Math & Comp Sci Nanchang 330031 Jiangxi Peoples R China

To achieve a fused image that contains rich texture details and prominent targets, we present a progressive dual-branch infrared and visible image fusion network called PDFusion, which incorporates the transformer module. Initially, the proposed network is divided into two branches to extract infrared and visible features independently. Subsequently, the image-wise transfer block (ITB) is introduced to fuse the infrared and visible features at different layers, facilitating the exchange of information between features. The fused features are then fed back into both pathways to contribute to the subsequent feature extraction process. Moreover, in addition to conventional pixel-level and structured loss functions, the contrastive language- image pretraining (CLIP) loss is introduced to guide the network training. Experimental results on publicly available datasets demonstrate the promising performance of PDFusion in the task of infrared and visible image fusion. The exceptional fusion performance of the proposed fusion network can be attributed to the following reasons: (1) The ITB block, particularly with the integration of the transformer, enhances the capability of representation learning. The transformer module captures long-range dependencies among image features, enabling a global receptive field that integrates contextual information from the entire image. This leads to a more comprehensive fusion of features. (2) The feature loss based on the CLIP image encoder minimizes the discrepancy between the generated and target images. Consequently, it promotes the generation of semantically coherent and visually appealing fused images. The source code of our method can be found at https://***/Changfei-Zhou/PDFusion.

关键词： Infrared and visible image fusion Dual-branch fusion network Progressive image-wise feature transfer transformer module CLIP loss

来源：评论

学校读者我要写书评

暂无评论

LGDNet: local feature coupling global representations network for pulmonary nodules detection

引用

MEDICAL & BIOLOGICAL ENGINEERING & COMPUTING 2024年第7期62卷 1991-2004页

作者： Chi, Jianning Zhao, Jin Wang, Siqi Yu, Xiaosheng Wu, Chengdong Northeastern Univ Fac Robot Sci & Engn Shenyang 110167 Liaoning Peoples R China Northeastern Univ Key Lab Intelligent Comp Med Image Minist Educ Shenyang 110167 Peoples R China

Detection of suspicious pulmonary nodules from lung CT scans is a crucial task in computer-aided diagnosis (CAD) systems. In recent years, various deep learning-based approaches have been proposed and demonstrated significant potential for addressing this task. However, existing deep convolutional neural networks exhibit limited long-range dependency capabilities and neglect crucial contextual information, resulting in reduced performance on detecting small-size nodules in CT scans. In this work, we propose a novel end-to-end framework called LGDNet for the detection of suspicious pulmonary nodules in lung CT scans by fusing local features and global representations. To overcome the limited long-range dependency capabilities inherent in convolutional operations, a dual-branch module is designed to integrate the convolutional neural network (CNN) branch that extracts local features with the transformer branch that captures global representations. To further address the issue of misalignment between local features and global representations, an attention gate module is proposed in the up-sampling stage to selectively combine misaligned semantic data from both branches, resulting in more accurate detection of small-size nodules. Our experiments on the large-scale LIDC dataset demonstrate that the proposed LGDNet with the dual-branch module and attention gate module could significantly improve the nodule detection sensitivity by achieving a final competition performance metric (CPM) score of 89.49%, outperforming the state-of-the-art nodule detection methods, indicating its potential for clinical applications in the early diagnosis of lung diseases.

关键词： Pulmonary nodule detection 3D medical image processing Convolutional neural network transformer module Feature alignment

来源：评论

学校读者我要写书评

暂无评论

SACA-fusion: a low-light fusion architecture of infrared and visible images based on self- and cross-attention

引用

VISUAL COMPUTER 2024年第5期40卷 3347-3356页

作者： Yu, Chongchong Li, Shunan Feng, Wenbin Zheng, Tong Liu, Shu Beijing Technol & Business Univ Sch Artificial Intelligence 11 Fucheng Rd Beijing 100048 Peoples R China Shenyang Res Inst China Coal Technol & Engn Grp Fushun 113122 Peoples R China China Acad Ind Internet Beijing 101103 Peoples R China

Visible-infrared image fusion cannot only reveal respective features of multiband imaging but also combine complementary information. It thus highlights salient information that cannot be directly obtained from a single waveband and enhances scene detection and perception. However, low-light condition for special scenarios, i.e., underground coal mine, impacts the performance of visible-infrared image fusion as they lead to lower contrast for visible light images and loss of local details. In this respect, we propose an infrared and visible image fusion architecture in low-light conditions based on self- and cross-attention (SACA-Fusion). This architecture replaces traditional fusion approaches with a transformer-based fusion network. It better extracts long-range dependencies of images and improves space recovery of fused images. The architecture has an attention mechanism composed of two modules. The self-attention module achieves global interaction and fusion of features and reduces loss in local details;the cross-attention module in nest connect enhances features in low-light conditions and achieves low-contrast space recovery. In the experiment part, through ablation, we confirm that the wonderful fusion strategy is transformer module, rather than RFN or directly connecting. Then, based on comparison experiments on TNO and LLVIP datasets, it is shown that the better fusion performance of the proposed one under some evaluation indicators. Especially in the actual low-light condition, the improvement of the fusion effect is commendable.

关键词： Visible-infrared image fusion Low light Self-attention module Cross-attention module transformer module

来源：评论

学校读者我要写书评

暂无评论

ViST: A Ubiquitous Model with Multimodal Fusion for Crop Growth Prediction

引用

ACM TRANSACTIONS ON SENSOR NETWORKS 2024年第1期20卷 1-23页

作者： Li, Junsheng Wang, Ling Liu, Jie Tang, Jinshan Harbin Inst Technol Dept Comp Sci & Technol 92 Xidazhi St Harbin 150000 Heilongjiang Peoples R China George Mason Univ Coll Publ Hlth Dept Hlth Adm & Policy Hlth Informat Fairfax VA 22033 USA

Crop growth prediction can help agricultural workers to make accurate and reasonable decisions on farming activities. Existing crop growth prediction models focus on one crop and train a single model for each crop. In this article, we develop a ubiquitous growth prediction model for multiple crops, aiming at training a single model for multiple crops. A ubiquitous vision and sensor transformer (ViST) model for crop growth prediction with image and sensor data is developed to achieve the goals. In the proposed model, a cross-attention mechanism is proposed to facilitate the fusion of multimodal feature maps to reduce computational costs and balance the interactive effects among features. To train the model, we combine the data from multiple crops to create a single (ViST) model. A sensor network system is established for data collection on the farm where rice, soybean, and maize are cultivated. Experimental results show that the proposed ViST model has an excellent ubiquitous ability for crop growth prediction with multiple crops.

关键词： Crop growth prediction ubiquitous model multimodal learning transformer module cross-attention mechanism

来源：评论

学校读者我要写书评

暂无评论

Self-attention transformer unit-based deep learning framework for skin lesions classification in smart healthcare

引用

DISCOVER APPLIED SCIENCES 2024年第1期6卷 3页

作者： Rezaee, Khosro Zadeh, Hossein Ghayoumi Meybod Univ Dept Biomed Engn Meybod Iran Vali e Asr Univ Rafsanjan Dept Elect Engn Rafsanjan Iran

Rising mortality rates in recent years have elevated melanoma to the ranks of the world's most lethal cancers. Dermoscopy images (DIs) have been used in smart healthcare applications to determine medical features using deep transfer learning (DTL). DI-related lesions are widespread, have local features, and are associated with uncertainty. There are three components to our bi-branch parallel model: (1) the transformer module (TM), (2) the self-attention unit (SAU), and (3) a convolutional neural network (CNN). With CNN and TM able to extract local and global features, respectively, a novel model has been developed to fuse global and local features using cross-fusion to generate fine-grained features. Parallel systems between the branches are merged using a feature-fusion architecture, resulting in a pattern that identifies the characteristics of a variety of lesions. Moreover, this paper proposes an optimized and lightweight CNN architecture version (optResNet-18) that discriminates skin cancer lesions with high accuracy. To verify the proposed method, the procedure evaluated the accuracy for the ISIC-2019 and the PH2 datasets as 97.48 and 96.87%, respectively, a significant difference over traditional CNN networks (e.g., ResNet-50 and ResNet-101) and the TM. The proposed model outperforms state-of-the-art performance metrics such as AUC, F1-score, specificity, precision, and recall. The proposed method can also be used as a generalizable model to diagnose different lesions in DIs with smart healthcare applications by combining DTL and medical imaging. With the proposed e-Health platform, skin diseases can be detected in real-time, which is crucial to speedy and reliable diagnostics. A novel approach to classifying multiple lesions in DIs is presented using a transformer module (TM) with the self-attention unit (SAU). A lightweight and optimized version of deep transfer learning is designed to alleviate overfitting. Test multi-tissue classifying accuracy was 97.48%

关键词： Melanoma cancer Smart healthcare Bi-directional feature fusion transformer module Convolutional neural network

来源：评论

学校读者我要写书评

暂无评论

TF-SOD: a novel transformer framework for salient object detection

引用

NEURAL COMPUTING & APPLICATIONS 2022年第14期34卷 11789-11806页

作者： Wang, Zhenyu Zhang, Yunzhou Liu, Yan Wang, Zhuo Coleman, Sonya Kerr, Dermot Northeastern Univ Fac Robot Sci & Engn Shenyang Peoples R China Northeastern Univ Coll Informat Sci & Engn Shenyang Peoples R China Natl Univ Def Technol Coll Intelligence Sci & Technol Changsha Peoples R China Univ Ulster Intelligent Syst Res Ctr Derry BT52 1SA North Ireland

Most of existing salient object detection models are based on fully convolutional network (FCN), which learn multi-scale/level semantic information through convolutional layers to obtain high-quality predicted saliency maps. However, convolution is locally interactive, it is difficult to capture remote dependencies, and FCN-based methods suffer from coarse object boundaries. In this paper, to solve these problems, we propose a novel transformer framework for salient object detection (named TF-SOD), which mainly consists of the encoder part of the FCN, fusion module (FM), transformer module (TM) and feature decoder module (FDM). Specifically, FM is a bridge connecting the encoder and TM and provides some foresight for the non-local interaction of TM. Besides, FDM can efficiently decode the non-local features output by TM and achieve deep fusion with local features. This architecture enables the network to achieve a close integration of local and non-local interactions, making information complementary to each other, deeply mining the associated information between features. Furthermore, we also propose a novel edge reinforcement learning strategy, which can effectively suppress edge blurring from local and global aspects by means of powerful network architecture. Extensive experiments using five datasets demonstrate that the proposed method outperforms 19 state-of-the-art methods

关键词： Salient object detection Fusion module transformer module Feature decoder Edge reinforcement learning strategy

来源：评论

学校读者我要写书评

暂无评论

MedLeafNet: A Feature Fusion Network Integrating transformer and CNN modules for Indian Medical Leaf Classification 2

MedLeafNet: A Feature Fusion Network Integrating Transformer...

引用

2nd International Conference on Intelligent Data Communication Technologies and Internet of Things, IDCIoT 2024

作者： Tiwari, Raj Gaurang Maheshwari, Himani Gautam, Vinay Agarwal, Ambuj Kumar Trivedi, Naresh Kumar Jain, Vishal Chitkara University Institute of Engineering and Technology Chitkara University Punjab India School of Computing Graphic Era Hill University Uttrakhand Dehradun India School of Engineering and Technology Department of Computer Science and Engineering Sharda University Greater Noida India

ISBN: (纸本)9798350327533

A new feature fusion network called 'MedLeafNet' is presented in this study specifically for the categorization of Indian medicinal leaves. Using a hybrid architecture, MedLeafNet integrates the best features of Convolutional Neural Network (CNN) modules for spatial feature extraction and transformer modules for sequence modeling. Because of the complexity of classifying Indian medicinal leaves-which are famous for their varied and powerful medicinal properties-the suggested methodology is well suited to the task. For this research, 80 different types of Indian leaves were analyzed, each with its well-known therapeutic properties. To improve its capacity to detect complex patterns in leaves, Me dLeafNet incorporates transformer and CNN modules to efficiently gather both global and local information. The suggested model was optimized using several different optimizers, such as Adam, Adadelta, Gradient Descent, AdaGrad, RMSProp, Adamax, Momentum, and Adaptive Moment Estimation. When trained with the Adam optimizer, MedLeafNet attained its maximum accuracy of 98.97%, as shown by extensive testing. MedLeafNet is also compared against some of the most well-known deep learning architectures, such as ResNet, Inception, VGG, Xception, and MobileNet. In terms of classification accuracy, the findings show that it is superior to these architectures, proving that it is effective for the difficult job of classifying Indian medicinal leaves. © 2024 IEEE.

关键词： Automated Plant Identification Deep Architectures Deep Learning Feature Fusion Network Medicinal Properties Optimizer Diversity transformer module

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：