检索结果-内蒙古大学图书馆

LPFSformer: Location Prior Guided Frequency and Spatial Interactive learning for Nighttime Flare Removal

IEEE TRANSACTIONS ON CIRCUITS AND SYSTEMS FOR VIDEO TECHNOLOGY 2025年第4期35卷 3706-3718页

作者： Chen, Guang-Yong Dong, Wei Fan, Guodong Su, Jian-Nan Gan, Min Chen, C. L. Philip Fuzhou Univ Coll Comp & Data Sci Fuzhou 350116 Peoples R China Minist Educ Fujian Key Lab Network Comp & Intelligent Informat Key Lab Intelligent Metro Univ Fujian Fuzhou 350108 Peoples R China Minist Educ Engn Res Ctr Big Data Intelligence Fuzhou 350108 Peoples R China Shandong Technol & Business Univ Sch Comp Sci & Technol Yantai 264003 Peoples R China Putian Univ New Engn Ind Coll Putian 351100 Fujian Peoples R China Putian Univ Putian Elect Informat Ind Technol Res Inst Putian 351100 Fujian Peoples R China Qingdao Univ Coll Comp Sci & Technol Qingdao 266071 Peoples R China South China Univ Technol Sch Comp Sci & Engn Guangzhou 510641 Peoples R China

When capturing images under strong light sources at night, intense lens flare artifacts often appear, significantly degrading visual quality and impacting downstream computer vision tasks. Although transformer-based methods have achieved remarkable results in nighttime flare removal, they fail to adequately distinguish between flare and non-flare regions. This unified processing overlooks the unique characteristics of these regions, leading to suboptimal performance and unsatisfactory results in real-world scenarios. To address this critical issue, we propose a novel approach incorporating Location Prior Guidance (LPG) and a specialized flare removal model, LPFSformer. LPG is designed to accurately learn the location of flares within an image and effectively capture the associated glow effects. By employing Location Prior Injection (LPI), our method directs the model's focus towards flare regions through the interaction of frequency and spatial domains. Additionally, to enhance the recovery of high-frequency textures and capture finer local details, we designed a Global Hybrid Feature Compensator (GHFC). GHFC aggregates different expert structures, leveraging the diverse receptive fields and CNN operations of each expert to effectively utilize a broader range of features during the flare removal process. Extensive experiments demonstrate that our LPFSformer achieves state-of-the-art flare removal performance compared to existing methods. Our code and a pre-trained LPFSformer have been uploaded to GitHub for validation.

关键词： Lenses Light sources Visualization image restoration Transformers Optical scattering Optical reflection Optical imaging image reconstruction Hardware Nighttime flare removal low-level computer vision deep learning

来源：评论

学校读者我要写书评

暂无评论

Autism Detection for Toddlers using Facial Features with deep learning 3

Autism Detection for Toddlers using Facial Features with Dee...

引用

3rd International Conference on Applied Artificial Intelligence and Computing, ICAAIC 2024

作者： Karthik, M.D. Jeba Priya, S. Mathu, T. Karunya Institute of Technology and Sciences Department of Cse Coimbatore India Karunya Institute of Technology and Sciences Department of Cse India

ISBN: (纸本)9798350375190

Autism Spectrum Disorder is a neuro-developmental disorder linked to challenges in social interaction, communication, and behavior. Children with autism often struggle with speech, exhibit challenges in maintaining eye contact, and face difficulties expressing emotions. Early detection and intervention in ASD are necessary to reduce the adverse effects of this disorder. Traditionally, ASD diagnosis involves costly and time-consuming brain imaging procedures. But nowadays, deep learning algorithms have advanced, providing a new way to detect ASD early by analyzing facial images. Unique facial features are found in Autistic toddlers that could be utilized with deep learning, but there has been limited attention given to this area. This research aims to improve the effectiveness of deep learning hybrid models, such as Vision Transformer (ViT) combined with Principal Component Analysis (PCA) and Support Vector Machine (SVM), ViT with CatBoost and SHAP, ViT with XGBoost and SHAP, and VGG16 with XGBoost. These modifications are specifically designed for image recognition tasks, such as detecting ASD in children's facial images. The proposed hybrid models bring about significant improvements over traditional architectures, decreasing the risk of overfitting and improving the models' capacity to identify subtle facial features linked to ASD. The proposed approach involves important preprocessing steps, which makes it appropriate for real-life scenarios. The accuracy of the hybrid models' performance is measured, and the findings show that the suggested hybrid models perform better than other conventional models at identifying ASD from images of children's faces. © 2024 IEEE.

关键词： Support vector machines

来源：评论

学校读者我要写书评

暂无评论

Potcapsnet: an explainable pyramid dilated capsule network for visualization of blight diseases

引用

Neural Computing and Applications 2024年第36期36卷 23251-23274页

作者： Gupta, Sachin Tripathi, Ashish Kumar Pandey, Avinash Chandra Malaviya National Institute of Technology Jaipur Jaipur India PDPM Indian Institute of Information Technology Design and Manufacturing Jabalpur India

Potato tuber is crucial as a primary vegetable food crop across the globe. However, its quality and quantity are constantly threatened by fungal blight diseases, particularly late and early blight, posing a significant risk to food security. To identify blight diseases, agrarians visually inspect potato leaf color variations, which is time-consuming and computationally expensive. To address this challenge, various advanced technologies, such as image processing, machine learning, and deep neural networks, have been widely applied in agricultural domains, especially for autonomous disease identification and classification. However, there is an urgent need to develop computational models that can rapidly and objectively detect these diseases, even in their early stages. Therefore, this paper introduces the "PotCapsNet" method that utilizes parallel atrous convolutional layers with varying dilation rates for multi-scale feature extraction, a shuffled convolutional block attention module for effective feature selection, and a capsule network for early disease classification. The effectiveness of the proposed model was validated using the publicly available PlantVillage and PLD datasets for multi-class potato leaf disease identification. The experimental findings show that the proposed method outperforms other methods in terms of accuracy, specificity, F1 score, and sensitivity. The proposed method attained average recognition accuracy of 97.81% and 97.95% on the PlantVillage and PLD datasets, respectively, the best among all the considered methods. The code and datasets of the proposed method are available at https://***/ersachingupta11/PotCapsNet. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2024.

关键词： deep neural networks

来源：评论

学校读者我要写书评

暂无评论

Fault transient waveform recognition in actual transmission lines using multi-scale convolution and lightweight channel attention DenseNet

引用

INTERNATIONAL JOURNAL OF COMPUTING SCIENCE AND MATHEMATICS 2024年第4期19卷 366-382页

作者： Zhang, Nailong Chen, Jie Gao, Chao Tan, Xiao State Grid Jiangsu Elect Power Co Ltd Res Inst Nanjing 210000 Peoples R China State Grid Jiangsu Elect Power Co Ltd Nanjing 210000 Peoples R China

Ensuring the proper functioning of power transmission lines is crucial for society and everyday life. Accurate fault identification methods are essential for line maintenance and inspection. Existing methods based on simulated data overlook the complexity of real-world fault data, which limits their reliability. This study presents a method for identifying the data types of real transmission line fault transient waveforms using a multi-scale convolutional channel attention DenseNet. The method consists of two main components: data preprocessing and fault recognition. The imbalance of fault samples is addressed using the synthetic minority over-sampling technique (SMOTE), and the Gramian angular field (GAF) method converts complex time series data into image data. Multi-scale convolution further extracts fault feature information, while a lightweight channel attention block ReLU-alpha-sigma activation map (RASimAM) enhances discriminative capacity. Experimental results demonstrate that the method effectively addresses sample imbalance issues and performs well on small datasets. It accurately identifies fault types and exhibits excellent generalisation performance.

关键词： transmission lines fault identification deep learning DenseNet multi scale channel attention

来源：评论

学校读者我要写书评

暂无评论

An Innovative Way of Empirical Analysis of Various Structures used in DL Algorithms for the Better Practices 4

An Innovative Way of Empirical Analysis of Various Structure...

引用

4th IEEE International Conference on Advance Computing and Innovative Technologies in Engineering, ICACITE 2024

作者： Kumar, Himanshu Jasim, Laith Gehlot, Anita Devi, Guntoju Kalpana Yadav, Kanchan Kumar, Neelam Sanjeev Chitkara University Chitkara Centre for Research and Development Himachal Pradesh 174103 India The Islamic University Najaf Iraq Uttaranchal Institute of Technology Uttaranchal University Department of Electronics & Communication Engineering Dehradun248007 India Cmr Technical Campus Department of Cse Telangana Hyderabad India Gla University Department of Electrical Engineering Mathura India Srm Institute of Science & Technology Department of Computer Science & Engineering Chennai Vadapalani26 India

ISBN: (纸本)9798350360165

The importance of deep learning is growing with artificial intelligence. It operates quickly. Entangled nerve fibres (CNNs), a well-known deep learning method, have shown impressive results in painting, classification, and semantic division. The actuator function-a crucial part of CNNs -was looked at in the context of low velocities in this book study, Engineering Vehicle Rear System Programme. The text begins with a comprehensive analysis of all four conventional activation functions, including the well recognised sigmoid function, and a summary of their respective benefits. It then discusses the potential benefits. Swish, FTS, and Relu- are the three new operating systems that are offered with SoftPlus. The research then includes a combination of new activators and extra activation functions. It operates at the TensorFlow framework level. On the CIFAR-10 dataset, more image recognition comparisons are carried out. According to the results, using a combination of novel activation functions produces better recognition accuracy than using just one new activation function. Moreover, the hybrid technique shows higher convergence speeds and shorter average processing times, indicating its effectiveness in real-world applications. © 2024 IEEE.

关键词： deep learning

来源：评论

学校读者我要写书评

暂无评论

Enhancing Privacy and Security of Autonomous UAV Navigation 2

Enhancing Privacy and Security of Autonomous UAV Navigation

引用

2nd IEEE Conference on Artificial Intelligence (CAI)

作者： Aggarwal, Vatsal Kaushik, Arjun Ramesh Jutla, Charanjit Ratha, Nalini Univ Buffalo Getzville NY 14068 USA IBM Res Yorktown Hts NY USA

ISBN: (纸本)9798350354102;9798350354096

Autonomous Unmanned Aerial Vehicles (UAVs) have become essential tools in defense, law enforcement, disaster response, and product delivery. These autonomous navigation systems require a wireless communication network, and of late are deep learning based. In critical scenarios such as border protection or disaster response, ensuring the secure navigation of autonomous UAVs is paramount. But, these autonomous UAVs are susceptible to adversarial attacks through the communication network or the deep learning models - eavesdropping / man-in-the-middle / membership inference / reconstruction. To address this susceptibility, we propose an innovative approach that combines Reinforcement learning (RL) and Fully Homomorphic Encryption (FHE) for secure autonomous UAV navigation. This end-to-end secure framework is designed for real-time video feeds captured by UAV cameras and utilizes FHE to perform inference on encrypted input images. While FHE allows computations on encrypted data, certain computational operators are yet to be implemented. Convolutional neural networks, fully connected neural networks, activation functions and OpenAI Gym Library are meticulously adapted to the FHE domain to enable encrypted data processing. We demonstrate the efficacy of our proposed approach through extensive experimentation. Our proposed approach ensures security and privacy in autonomous UAV navigation with negligible loss in performance.

关键词： Autonomous Unmanned Aerial Vehicles Fully Homomorphic Encryption Privacy Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

AIfES: A Next-Generation Edge AI Framework

引用

IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE 2024年第6期46卷 4519-4533页

作者： Wulfert, Lars Kuehnel, Johannes Krupp, Lukas Viga, Justus Wiede, Christian Gembaczka, Pierre Grabmaier, Anton Fraunhofer Inst Microelect Circuits & Syst D-47057 Dusiburg Germany Rhein Westfal TH Aachen D-52076 Aachen Germany

Edge Artificial Intelligence (AI) relies on the integration of Machine learning (ML) into even the smallest embedded devices, thus enabling local intelligence in real-world applications, e.g. for image or speech processing. Traditional Edge AI frameworks lack important aspects required to keep up with recent and upcoming ML innovations. These aspects include low flexibility concerning the target hardware and limited support for custom hardware accelerator integration. Artificial Intelligence for Embedded Systems Framework (AIfES) has the goal to overcome these challenges faced by traditional edge AI frameworks. In this paper, we give a detailed overview of the architecture of AIfES and the applied design principles. Finally, we compare AIfES with TensorFlow Lite for Microcontrollers (TFLM) on an ARM Cortex-M4-based System-on-Chip (SoC) using fully connected neural networks (FCNNs) and convolutional neural networks (CNNs). AIfES outperforms TFLM in both execution time and memory consumption for the FCNNs. Additionally, using AIfES reduces memory consumption by up to 54% when using CNNs. Furthermore, we show the performance of AIfES during the training of FCNN as well as CNN and demonstrate the feasibility of training a CNN on a resource-constrained device with a memory usage of slightly more than 100 kB of RAM.

关键词： Edge AI Framework embedded systems machine learning framework on-device training resource-constrained devices TinyML

来源：评论

学校读者我要写书评

暂无评论

Bragg Spot Finder (BSF): a new machine-learningaided approach to deal with spot finding for rapidly filtering diffraction pattern images

引用

JOURNAL OF APPLIED CRYSTALLOGRAPHY 2024年第3期57卷 670-680页

作者： Dong, Jianxiang Yin, Zhaozheng Kreitler, Dale Bernstein, Herbert J. Jakoncic, Jean SUNY Stony Brook Dept Comp Sci Coll Engn & Appl Sci Upton NY USA Brookhaven Natl Lab Natl Synchrotron Light Source 2 Bldg 745 Upton NY 11973 USA Brookhaven Natl Lab Ronin Inst Independent Scholarship NSLS 2 Bldg 745 Upton NY USA

Macromolecular crystallography contributes significantly to understanding diseases and, more importantly, how to treat them by providing atomic resolution 3D structures of proteins. This is achieved by collecting X-ray diffraction images of protein crystals from important biological pathways. Spotfinders are used to detect the presence of crystals with usable data, and the spots from such crystals are the primary data used to solve the relevant structures. Having fast and accurate spot finding is essential, but recent advances in synchrotron beamlines used to generate X-ray diffraction images have brought us to the limits of what the best existing spotfinders can do. This bottleneck must be removed so spotfinder software can keep pace with the X-ray beamline hardware improvements and be able to see the weak or diffuse spots required to solve the most challenging problems encountered when working with diffraction images. In this paper, we first present Bragg Spot Detection (BSD), a large benchmark Bragg spot image dataset that contains 304 images with more than 66 000 spots. We then discuss the open source extensible U-Net-based spotfinder Bragg Spot Finder (BSF), with image pre-processing, a U-Net segmentation backbone, and post-processing that includes artifact removal and watershed segmentation. Finally, we perform experiments on the BSD benchmark and obtain results that are (in terms of accuracy) comparable to or better than those obtained with two popular spotfinder software packages (Dozor and DIALS), demonstrating that this is an appropriate framework to support future extensions and improvements.

关键词： Bragg Spot Finder BSF Bragg Spot Detection BSD Bragg reflections weak intensities machine learning filtering X-ray diffraction microbeams diffraction spot search real time processing

来源：评论

学校读者我要写书评

暂无评论

Bayesian Source Identification With Dual Hierarchical Neural Networks for Urban Air Pollution

引用

JOURNAL OF ADVANCES IN MODELING EARTH SYSTEMS 2025年第4期17卷

作者： Al Aawar, Elissar Resifi, Sofien Jebari, Hatem Hoteit, Ibrahim King Abdullah Univ Sci & Technol Phys Sci & Engn Div Thuwal Saudi Arabia

Identifying urban air pollution sources is essential for public health and environmental sustainability. In this study, we propose a novel hierarchical method for urban air pollution source identification, leveraging deep learning (DL) within an efficient Bayesian inference framework. We rely on observations in the form of two-dimensional (2D) pollutant concentration distributions, and adopt the Wasserstein W2 $\left({W}_real\right)$ distance to model the likelihood probability distribution. The hierarchical nature of the framework stems from the integration of two neural networks (NNs). The first one acts as an emulator that replicates the physical dispersion model to predict future pollution observations recursively over a defined timeframe. These predictions are then used as inputs for the second NN that approximates the W2 ${W}_real$ distance between predicted and observed pollutant concentration distributions to rapidly compute the likelihood probability. The approach adopts a multi-model strategy to mitigate the accumulation of errors, particularly those arising from the recursive prediction steps across multiple time intervals, ensuring the reliability of predictions over extended periods. The proposed framework is implemented on graphics processing units (GPUs), enabling scalable computations for real-world applications and rapid decision making. Through extensive numerical experiments, we demonstrate the suggested method's effectiveness in accurately estimating pollution source parameters, including location, emission rate, and duration, using synthetic observational data. Sensitivity analyses further explore the impact of observational horizons and sampling on solution convergence and accuracy. Numerical results demonstrate robust performances and computational efficiency compared to the conventional approach, particularly in scenarios with limited computational resources and observations availability.

关键词： Bayesian inference deep learning source identification surrogate models uncertainty quantification urban air pollution

来源：评论

学校读者我要写书评

暂无评论

Fault Classification in Aircraft Electrical Systems Using Transients and FPGA-Based deep learning

Fault Classification in Aircraft Electrical Systems Using Tr...

引用

2024 IEEE International Conference on Green Energy and Smart Systems, GESS 2024

作者： Guzman, Ian Babiceanu, Radu F. Embry-Riddle Aeronautical University Daytona BeachFL United States

ISBN: (纸本)9798350379617

The shift of aviation industry towards More Electric Aircraft (MEA) has led to increasingly complex aircraft power systems characterized by highly inter-connected electrical and electronic components. Electri-cal monitoring systems are indispensable equipment to ensure flight safety and reliability in aircraft environments. Field-Programmable Gate Arrays (FPGAs) offer compelling solutions for the implementation of these systems in aircraft due to their signal processing-based capabilities, reduced power consumption and compact size suitable for integration into avionics. Additionally, FPGAs can process large amounts of data in real-time at high speeds. This paper explores the FPGA implementation of a deep learning model to detect and classify faults in aircraft power systems using transients that are introduced by electrical faults such as short and open circuits. The proposed approach employs an LSTM neural network with a classification accuracy of 94.09% deployed on a Xilinx Zynq UltraScale+ MPSoC ZCU102 development board. Experimental results demonstrate the effectiveness of the FPGA implementation in classifying electrical faults based on high-frequency transients. © 2024 IEEE.

关键词： Aircraft power systems

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：