检索结果-内蒙古大学图书馆

Global-Local Multigranularity Transformer for Hyperspectral image Classification

IEEE JOURNAL OF SELECTED TOPICS IN APPLIED EARTH OBSERVATIONS AND REMOTE SENSING 2025年 18卷 112-131页

作者： Meng, Zhe Yan, Qian Zhao, Feng Chen, Gaige Hua, Wenqiang Liang, Miaomiao Xian Univ Posts & Telecommun Sch Commun & Informat Engn Xian 710121 Peoples R China Xian Univ Posts & Telecommun Sch Artificial Intelligence Xian 710121 Peoples R China Xian Univ Posts & Telecommun Sch Comp Sci & Technol Xian 710121 Peoples R China Jiangxi Univ Sci & Technol Sch Informat Engn Jiangxi Prov Key Lab Multidimens Intelligent Perce Ganzhou 341000 Peoples R China

Hyperspectral image (HSI) classification is a challenging task in remote sensing applications, aiming to determine the category of each pixel by utilizing rich spectral and spatial information in HSI. Convolutional neural networks (CNNs) have been effective in processing HSI data by extracting local features, but they are deficient in capturing global contextual information. Recently, transformer has become proficient in attending to global information due to their self-attention mechanisms, yet they may fall short in capturing multiscale features of HSI. To address these limitations, a global-local multigranularity transformer (GLMGT) network is proposed for HSI classification. The GLMGT combines CNN with the transformer to comprehensively capture multigranularity spectral and spatial features across global and local scales. Specifically, we introduce a multigranularity spatial feature extraction block to extensively extract spatial information at different granularities, including multiscale local spatial features and global spatial features. In addition, we introduce a multigranularity spectral feature extraction block to fully leverage spectral information across different granularities. The validity of the proposed method is demonstrated through experimental validation using seven publicly available datasets, which include two Chinese satellite hyperspectral datasets (ZY1-02D Huanghekou and GF-5 Yancheng) and one UAV-based hyperspectral dataset.

关键词： Feature extraction Transformers Convolution Convolutional neural networks Data mining Three-dimensional displays Hyperspectral imaging Kernel Spectral analysis Computer vision CNN multigranularity transformer hyperspectral image (HSI) classification

来源：评论

学校读者我要写书评

暂无评论

Satellite image classification with neural quantum kernels

引用

MACHINE LEARNING-SCIENCE AND TECHNOLOGY 2025年第1期6卷

作者： Rodriguez-Grasa, Pablo Farzan-Rodriguez, Robert Novelli, Gabriele Ban, Yue Sanz, Mikel Univ Basque Country UPV EHU Dept Phys Chem Apartado 644 Bilbao 48080 Spain Univ Basque Country UPV EHU Quantum Ctr Apartado 644 Bilbao 48080 Spain TECNALIA Basque Res & Technol Alliance BRTA Derio 48160 Spain GMV Dept Artificial Intelligence & Big Data Isaac Newton 11 Madrid 28760 Spain Univ Carlos III Madrid Dept Fis Avda Univ 30 Leganes 28911 Spain CSIC Inst Ciencia Mat Madrid E-28049 Madrid Spain IKERBASQUE Basque Fdn Sci Plaza Euskadi 5 Bilbao 48009 Spain Basque Ctr Appl Math BCAM Alameda Mazarredo 14 Bilbao 48009 Spain

Achieving practical applications of quantum machine learning (QML) for real-world scenarios remains challenging despite significant theoretical progress. This paper proposes a novel approach for classifying satellite images, a task of particular relevance to the earth observation industry, using QML techniques. Specifically, we focus on classifying images that contain solar panels, addressing a complex real-world classification problem. Our approach begins with classical pre-processing to reduce the dimensionality of the satellite image dataset. We then apply neural quantum kernels-quantum kernels derived from trained quantum neural networks-for classification. We evaluate several strategies within this framework, demonstrating results that are competitive with the best classical methods. Key findings include the robustness of or results and their scalability, with successful performance achieved up to 8 qubits.

关键词： quantum neural networks quantum kernels quantum machine learning satelite image classification real dataset

来源：评论

学校读者我要写书评

暂无评论

Efficient Deep Learning Infrastructures for Embedded Computing Systems: A Comprehensive Survey and Future Envision

引用

ACM TRANSACTIONS ON EMBEDDED COMPUTING SYSTEMS 2025年第1期24卷 1-100页

作者： Luo, Xiangzhong Liu, Di Kong, Hao Huai, Shuo Chen, Hui Xiong, Guochu Liu, Weichen Nanyang Technol Univ Nanyang Singapore Norwegian Univ Sci & Technol Dept Comp Sci Trondheim Norway Nanyang Technol Univ Sch Comp Sci & Engn Singapore Singapore

Deep neural networks (DNNs) have recently achieved impressive success across a wide range of real-world vision and language processing tasks, spanning from image classification to many other downstream vision tasks, such as object detection, tracking, and segmentation. However, previous well-established DNNs, despite being able to maintain superior accuracy, have also been evolving to be deeper and wider and thus inevitably necessitate prohibitive computational resources for both training and inference. This trend further enlarges the computational gap between computation-intensive DNNs and resource-constrained embedded computing systems, making it challenging to deploy powerful DNNs in real-world embedded computing systems towards ubiquitous embedded intelligence. To alleviate this computational gap and enable ubiquitous embedded intelligence, we focus in this survey on discussing recent efficient deep learning infrastructures for embedded computing systems, spanning from training to inference, from manual to automated, from convolutional neural networks to transformers, from transformers to vision transformers, from vision models to large language models, from software to hardware, and from algorithms to applications. Specifically, we discuss recent efficient deep learning infrastructures for embedded computing systems from the lens of (1) efficient manual network design for embedded computing systems, (2) efficient automated network design for embedded computing systems, (3) efficient network compression for embedded computing systems, (4) efficient on-device learning for embedded computing systems, (5) efficient large language models for embedded computing systems, (6) efficient deep learning software and hardware for embedded computing systems, and (7) efficient intelligent applications for embedded computing systems. We also envision promising future directions and trends, which have the potential to deliver more ubiquitous embedded intelligence. We believe th

关键词： Embedded computing systems embedded intelligence artificial intelli- gence efficient deep learning algorithms efficient network design efficient neural architecture search effi- cient model compression efficient on-device learning efficient large language models efficient deep learning software and hardware intelligent embedded applications

来源：评论

学校读者我要写书评

暂无评论

IoT-Based Multisensors Fusion for Activity Recognition via Key Features and Hybrid Transfer Learning

引用

IEEE ACCESS 2025年 13卷 14727-14742页

作者： Jalal, Ahmad Khan, Danyal Sadiq, Touseef Alotaibi, Moneerah Alotaibi, Sultan Refa Aljuaid, Hanan Rahman, Hameedur Air Univ Fac Comp & AI Islamabad 44000 Pakistan Korea Univ Coll Informat Dept Comp Sci & Engn Seoul 02841 South Korea Univ Agder Ctr Artificial Intelligence Res Dept Informat & Commun Technol N-4879 Grimstad Norway Shaqra Univ Dept Comp Sci Coll Sci & Humanities Dawadmi Shaqra Saudi Arabia Princess Nourah Bint Abdulrahman Univ Coll Comp & Informat Sci Dept Comp Sci POB 84428 Riyadh 11671 Saudi Arabia

Human activity recognition (HAR) has attracted significant attention in various fields, including healthcare, smart homes, and human-computer interaction. Accurate HAR can enhance user experience, provide critical health insights, and enable sophisticated context-aware applications. This paper presents a comprehensive system for HAR utilizing both RGB videos and inertial measurement unit (IMU) sensor data. The system employs a multi-stage processing pipeline involving preprocessing, segmentation, feature extraction, and classification to achieve high accuracy in activity recognition. In the preprocessing stage, frames are extracted from RGB videos, and IMU sensor data undergoes denoising. The segmentation phase applies Naive Bayes segmentation for video frames and Hamming windows for sensor data to prepare them for feature extraction. Key features are extracted using techniques such as ORB (Oriented FAST and Rotated BRIEF), MSER (Maximally Stable Extremal Regions), DFT (Discrete Fourier Transform), and KAZE for image data, and LPCC (Linear Predictive Cepstral Coefficients), PSD (Power Spectral Density), AR Coefficient, and entropy for sensor data. Feature fusion is performed using Linear Discriminant Analysis (LDA) to create a unified feature set, which is then classified using ResNet50 (Residual neural Network) to recognize activities such as using a smartphone, cooking, and reading a newspaper. The system was evaluated using the LARa and HWU-USP datasets, achieving classification accuracies of 92% and 93%, respectively. These results demonstrate the robustness and effectiveness of the proposed HAR system in diverse scenarios.

关键词： Feature extraction Videos image segmentation Human activity recognition Accuracy Noise Convolutional neural networks Data mining Residual neural networks Noise reduction Human activity recognition (HAR) RGB videos IMU sensor data Naive Bayes segmentation recurrent neural network (RNN) Shapley additive explanations (SHAP)

来源：评论

学校读者我要写书评

暂无评论

Frequency-Domain Super-Resolution With Reconstruction Using Compressed Representation (FDSR-RCR) Algorithm for Remote Sensing Satellite images

引用

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING 2025年 63卷

作者： Miao, Jiaqing Zhou, Xiaobing Li, Guibing Li, Gaoping Zeng, Li Liu, Xiaoguang Tan, Ying Southwest Minzu Univ Sch Math Chengdu 610054 Peoples R China Montana Technol Univ Geol Engn Dept Butte MT 59701 USA Southwest Minzu Univ Coll Comp Sci & Artificial Intelligence Key Lab Comp Syst State Ethn Affairs Commiss Chengdu 610041 Peoples R China Southwest Jiaotong Univ Sch Elect Engn Chengdu 611756 Peoples R China

In remote sensing image processing for Earth and environmental applications, super-resolution (SR) is a crucial technique for enhancing the resolution of low-resolution (LR) images. In this study, we proposed a novel algorithm of frequency-domain super-resolution with reconstruction from compressed representation. The algorithm follows a multistep procedure: first, an LR image in the space domain is transformed to the frequency domain using a Fourier transform. The frequency-domain representation is then expanded to the desired size (number of pixels) of a high-resolution (HR) image. This expanded frequency-domain image is subsequently inverse Fourier transformed back to the spatial domain, yielding an initial HR image. A final HR image is then reconstructed from the initial HR image using a low-rank regularization model that incorporates a nonlocal smoothed rank function (SRF). We evaluated the performance of the new algorithm by comparing the reconstructed HR images with those generated by several commonly used SR algorithms, including: 1) bicubic interpolation;2) sparse representation;3) adaptive sparse domain selection and adaptive regularization;4) fuzzy-rule-based (FRB) algorithm;5) SR convolutional neural networks (SRCNNs);6) fast SR convolutional neural networks (FSRCNNs);7) practical degradation model for deep blind image SR;8) the frequency separation for real-world SR (FSSR);and 9) the enhanced SR generative adversarial networks (ESRGANs). The algorithms were tested on Landsat-8 and Moderate Resolution Imaging Spectroradiometer (MODIS) multiresolution images over various locations, as well as on images with artificially added noise to assess the robustness of each algorithm. Results show that: 1) the proposed new algorithm outperforms the others in terms of the peak signal-to-noise ratio, structure similarity, and root-mean-square error and 2) it effectively suppresses noise during HR reconstruction from noisy low-resolution (LR) images, overcoming a key

关键词： Compressed representation frequency domain low-rank regularization model low-rank regularization model remote sensing images remote sensing images smoothed rank function (SRF) smoothed rank function (SRF) super-resolution (SR) super-resolution (SR) super-resolution (SR)

来源：评论

学校读者我要写书评

暂无评论

Automated Extraction and Classification of Fossil Whorl Patterns Using Deep Learning: A Novel Approach to Paleontological image Analysis

引用

TRAITEMENT DU SIGNAL 2025年第1期42卷 67-77页

作者： Bi, Ruifang Shi, Hualin Shanxi Inst Technol Dept Art & Design Sci Yangquan 045000 Peoples R China Huayuan Int Land Port Grp Co Ltd Enterprise Technol Ctr Taiyuan 030000 Peoples R China Key Lab Intelligent Logist Shanxi Prov Taiyuan 030000 Peoples R China

The rapid advancement of artificial Intelligence (AI) has led to the displacement of traditional fossil pattern recognition methods in paleontological studies, particularly through the application of image processing technologies. This study focuses on the fossilized whorls of ancient organisms from the Yangquan region, employing state-of-the-art AI-driven techniques to identify and extract distinctive features from these fossils for automated pattern recognition. Existing paleontological databases of whorl fossils were reviewed, and a deep learning model was developed using convolutional neural networks (CNNs) to facilitate the extraction and classification of fossil whorl patterns. The model incorporates multi-level feature abstraction through various image preprocessing techniques to enhance both the accuracy and robustness of the recognition process. A transfer learning strategy based on CNNs was introduced, allowing for rapid adaptation to new fossil patterns despite limited sample sizes. Furthermore, an improved feature extraction algorithm leveraging Scale-Invariant Feature Transform (SIFT) for feature point matching was implemented, significantly accelerating the speed and accuracy of the feature extraction process. In the experimental phase, over 300 images of fossilized whorls were utilized for model training and validation, achieving a recognition accuracy exceeding 95%, Awhich represents an improvement of nearly 30% over traditional manual methods. The generalization ability of the model was also evaluated, confirming its stability and reliability across diverse fossil data sets. This research underscores the transformative potential of AI-based image processing technologies in the extraction and analysis of paleontological patterns, offering new tools for the study of Yangquan fossils while also contributing to broader applications in cultural heritage preservation and scientific education. This work provides a solid foundation for the further integrati

关键词： paleontological whorl fossils AI image processing CNN feature point extraction pattern recognition

来源：评论

学校读者我要写书评

暂无评论

An end-to-end deep convolutional neural network-based data-driven fusion framework for identification of human induced pluripotent stem cell-derived endothelial cells in photomicrographs

引用

ENGINEERING applications OF artificial INTELLIGENCE 2025年 139卷

作者： Iqbal, Imran Ullah, Imran Peng, Tingying Wang, Weiwei Ma, Nan Inst Act Polymers Helmholtz Zentrum Hereon Dept PLR D-14513 Teltow Germany German Res Ctr Environm Hlth Helmholtz Munich D-85764 Munich Germany

Deep learning is a very powerful analytic tool to recognize the patterns in data to make appropriate predictions. It has tremendous potential in data analyses, particularly for cell biology domain, caused by the growing scale and inherent complexity of biological data. The core purpose of this research work is to design, implement, and calibrate an efficient deep convolutional neural network (DCNN) architecture in the context of binary-class classification problem. This diversified network is developed to precisely identify human induced pluripotent stem cell-derived endothelial cells (hiPSC-derived EC) based on photomicrograph. The proposed architecture is cerebrally developed with numerous convolutional modules, multiple kernel sizes, various pooling layers, activation functions and strides, nevertheless fewer trainable parameters to strengthen the network and enhance its performance. The proposed feature fusion framework is compared with the classifier fusion approach in terms of Matthews's correlation coefficient (MCC), training time, inference time, number of layers, number of parameters, graphics processing unit (GPU) memory utilization, and floating-point operations (FLOPS). Specifically, it achieves 94.6% sensitivity, 94.5% specificity, and 94.7% precision. Induced pluripotent stem cell (iPS) dataset is also introduced in this research work that has 16278 images which are labelled by three independent and experienced human experts of cell biology domain to facilitate future research. Experimental results show that the proposed framework offers an innovative and attainable algorithm for accelerating and systematizing the classification task along with saving time and effort.

关键词： Computer vision Deep convolutional neural networks Endothelial cells Human induced pluripotent stem cells image processing Information fusion Machine learning Photomicrograph

来源：评论

学校读者我要写书评

暂无评论

Efficient neural Architecture Search: Computational Cost Reduction Mechanisms in DeepGA 17th

Efficient Neural Architecture Search: Computational Cost Red...

引用

17th Workshop of Hybrid Intelligent Systems, HIS 2024, 17th Workshop on Intelligent Learning Environments, WILE 2024 and 6th Workshop on New Trends in Computational Intelligence and applications, CIAPP 2024, held in conjunction with Mexican International Conference on artificial Intelligence, MICAI 2024

作者： Barradas-Palmeros, Jesús-Arnulfo López-Herrera, Carlos-Alberto Acosta-Mesa, Héctor-Gabriel Mezura-Montes, Efrén Artificial Intelligence Research Institute University of Veracruz Xalapa Mexico

ISBN: (纸本)9783031838811

neural Architecture Search (NAS) aims to automate the design process of Deep neural networks (DNN) without requiring profound domain knowledge. The Deep Genetic Algorithm (DeepGA) was proposed to find the architectures of Convolutional neural networks (CNNs) for image processing, and its applications have covered a variety of data domains. Nonetheless, one of the main impediments of NAS is its computational cost, which is produced by evaluating candidate architectures. This work proposes using two cost-reduction mechanisms applied to DeepGA: using memory to avoid repeated evaluations and reducing the number of epochs for training in a low-fidelity estimation scheme for evaluations. The results indicate that the execution time of the algorithm was extensively reduced without diminishing the accuracy performance of the resulting architecture. The previous derives in a more efficient NAS procedure. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Complex-Valued Federated Learning with Differential Privacy and MRI applications 9th

Complex-Valued Federated Learning with Differential Privacy ...

引用

27th International Conference on Medical image Computing and Computer Assisted Intervention (MICCAI)

作者： Riess, Anneliese Ziller, Alexander Kolek, Stefan Rueckert, Daniel Schnabel, Julia Kaissis, Georgios Helmholtz Munich Inst Machine Learning Biomed Imaging Neuherberg Germany Tech Univ Munich Sch Computat Informat & Technol Munich Germany Helmholtz Munich Helmholtz AI Oberschleissheim Germany Tech Univ Munich Artificial Intelligence Healthcare & Med Munich Germany Ludwig Maximilians Univ Munchen Math Foundat AI Munich Germany Imperial Coll London Dept Comp London England Kings Coll London Sch Biomed Engn & Imaging Sci London England Tech Univ Munich Inst Diagnost & Intervent Radiol Munich Germany

ISBN: (纸本)9783031776090;9783031776106

Federated learning enhanced with Differential Privacy (DP) is a powerful privacy-preserving strategy to protect individuals sharing their sensitive data for processing in fields such as medicine and healthcare. Many medical applications, for example magnetic resonance imaging (MRI), rely on complex-valued signal processing techniques for data acquisition and analysis. However, the appropriate application of DP to complex-valued data is still underexplored. To address this issue, from the theoretical side, we introduce the complex-valued Gaussian mechanism, whose behaviour we characterise in terms of f-DP, (epsilon, delta)-DP and Renyi-DP. Moreover, we generalise the fundamental algorithm DP stochastic gradient descent to complex-valued neural networks and present novel complex-valued neural network primitives compatible with DP. Experimentally, we showcase a proof-of-concept by training federated complex-valued neural networks with DP on a real-world task (MRI pulse sequence classification in k-space), yielding excellent utility and privacy. Our results highlight the relevance of combining federated learning with robust privacy-preserving techniques in the MRI context.

关键词： Federated learning Differential Privacy Complex Numbers

来源：评论

学校读者我要写书评

暂无评论

Learning Truly Monotone Operators with applications to Nonlinear Inverse Problems

引用

SIAM JOURNAL ON IMAGING SCIENCES 2025年第1期18卷 735-764页

作者： Belkouchi, Younes Pesquet, Jean-Christophe Repetti, Audrey Talbot, Hugues Univ Paris Saclay CentraleSupelec Inria F-91190 Gif Sur Yvette France Heriot Watt Univ Sch Math & Comp Sci Edinburgh EH14 4AS Scotland Heriot Watt Univ Sch Engn & Phys Sci Edinburgh EH14 4AS Scotland Bayes Ctr Maxwell Inst Math Sci Edinburgh EH9 3FD Scotland

This article introduces a novel approach to learning monotone neural networks (NNs) through a newly defined penalization loss. The proposed method is particularly effective in solving classes of variational problems, specifically monotone inclusion problems, commonly encountered in image processing tasks. The forward-backward-forward (FBF) algorithm is employed to address these problems, offering a solution even when the Lipschitz constant of the NN is unknown. Notably, the FBF algorithm provides convergence guarantees under the condition that the learned operator is monotone. Building on plug-and-play methodologies, our objective is to apply these newly learned operators to solving nonlinear inverse problems. To achieve this, we initially formulate the problem as a variational inclusion problem. Subsequently, we train a monotone NN to approximate an operator that may not inherently be monotone. Leveraging the FBF algorithm, we then show simulation examples where the nonlinear inverse problem is successfully solved.

关键词： monotone operator optimization inverse problem deep learning forward-backward-forward plug and play (PnP)

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：