检索结果-内蒙古大学图书馆

3rd International Conference on Advances in Information and Communication Technology, ICTA 2024

作者： Phuong, Huy Nguyen Mai, Thuong Duong Thi Thai Nguyen University of Technology Thai Nguyen Viet Nam University of Information and Communication Technology Thai Nguyen Viet Nam

ISBN: (纸本)9783031809422

Nowadays, deep learning architectures like CNN have proven their superiority in image recognition tasks. To effectively deploy CNN networks in practice, especially for AIoT applications, it is essential to find a network model that offers good recognition performance with a small size and a limited number of parameters. Additionally, when deploying to hardware, to ensure rapid task execution, an FPGA-based approach is a suitable choice due to its parallel processing capabilities, low power consumption, low latency, and reconfigurability. In this paper, the authors propose a solution for searching and implementing an optimal CNN model on an FPGA chip for handwritten digit recognition using the MNIST dataset. The comparative results of recognizing 10,000 MNIST image samples on FPGA, CPU, and Raspberry Pi 4 microcontroller demonstrate that the proposed solution achieves high accuracy (98.78%) while maintaining low execution time (36 s). This indicates the potential application of the proposed solution in deploying deep learning tasks in practical applications. © 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

deep learning Applications for Identifying and Detecting Heritage Sites 7

Deep Learning Applications for Identifying and Detecting Her...

引用

7th International Conference on Inventive Computation Technologies, ICICT 2024

作者： Anand Dravid, T. Jayanth, J.T. Rajalakshmi Engineering College Department of Computer Science Chennai Tamil Nadu Thandalam India

ISBN: (数字)9798350359299

ISBN: (纸本)9798350359299

This research work introduces a novel application of Artificial Intelligence (AI) in monument identification, utilizing the state-of-the-art object detection model YOLOv8. With the aim of revolutionizing the study, preservation, and understanding of historical sites, the proposed method employs image and video data for detection. A comprehensive dataset of historical temples was collected and annotated for training the system. Convolutional Neural Networks (CNNs) and the YOLOv8 algorithm, advanced deep learning techniques, were employed to locate and identify individual monuments within intricate historical contexts. Furthermore, sophisticated metric evaluation methods were integrated into the system to provide real-time insights into model behavior, detection reliability, potential biases, and ongoing monitoring, eliminating the need for third-party platforms. This feature enables rapid adjustments and enhancements, resulting in a dependable and effective monument detection system. Additionally, the model incorporates a Text-To-Speech (TTS) system to vocalize the names, historical contexts, and other relevant information of the identified heritage monuments. © 2024 IEEE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Cognitive Twins for Predictive Maintenance and Security in IoT Software Systems 4

Cognitive Twins for Predictive Maintenance and Security in I...

引用

4th IEEE International Conference on Mobile Networks and Wireless Communications, ICMNWC 2024

作者： Swamy, B. Venkata Barmola, Pareshwar Prasad Thangavel, Senthil Kaliappan, S. Patel, Hardik Abhyankar, Girish B V Raju Institute of Technology Department of Chemistry Medak India Graphic Era Deemed to be University Department of Computer Science & Engineering Dehradun India Paypal Inc CA United States Lovely Professional University Division of Research and Development Phagwara India Parul institute of Engineering and Technology Parul University Faculty of Engineering and Technology Department of Electronics & Communication Engineering Post Limda India Faculty of Law Pune India

ISBN: (纸本)9798350352931

The increasing complexity and interconnectedness of Internet of Things (IoT) software systems necessitate the development of intelligent solutions for predictive maintenance and security. Conventional techniques often fail to provide real-time insights and proactive responses due to the diverse and dynamic nature of IoT environments. To address these challenges, cognitive technologies offer promising avenues for enhancing the operational efficiency and security of IoT networks. This paper introduces Cognitive Twins, an AI-driven framework designed to optimize predictive maintenance and strengthen security in IoT software systems. Cognitive Twins leverage advanced machine learning models and real-time data streams to create dynamic digital replicas of IoT devices and software components. The framework employs a combination of deep learning-based anomaly detection, reinforcement learning for proactive maintenance scheduling, and natural language processing (NLP) for automated security log analysis. By continuously learning from device interactions and evolving threat patterns, Cognitive Twins predict potential failures and detect security threats before they occur, enabling real-time decision-making and automated responses. Cognitive Twins were evaluated on a large-scale IoT network, consisting of 500 nodes across multiple application domains. The framework achieved a predictive maintenance accuracy of 96.8%, reducing downtime by 35% compared to traditional models. In security applications, Cognitive Twins identified cyber threats with a detection rate of 98.3%, lowering the false positive rate to 1.5%. © 2024 IEEE.

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

FieldSeg: A scalable agricultural field extraction framework based on the Segment Anything Model and 10-m Sentinel-2 imagery

引用

COMPUTERS AND ELECTRONICS IN AGRICULTURE 2025年 232卷

作者： Ferreira, Lucas B. Martins, Vitor S. Aires, Uilson R. V. Wijewardane, Nuwan Zhang, Xin Samiappan, Sathish Mississippi State Univ MSU Dept Agr & Biol Engn Mississippi State MS 39762 USA Univ Tennessee Dept Biosyst Engn & Soil Sci Knoxville TN 37996 USA

Accurate delineation of agricultural fields from satellite imagery is crucial for digital agriculture and conservation. The Segment Anything Model (SAM), a state-of-the-art image segmentation model, brings new possibilities for this task. However, its feasibility under different agricultural contexts remains unclear, and there are open questions regarding model parametrization, image preprocessing, and integration into an operational framework. This study proposes a new SAM-assisted crop field extraction framework (FieldSeg) using 2022 Sentinel-2 temporal composites and presents the lessons learned using this foundational model in eight agricultural regions across the world. Through rigorous experiments, this study optimized FieldSeg in three stages: input data preparation, model parametrization and patch management, and final fine parametrization. This study explored different bands and temporal metrics combinations and defined a set of optimal configurations for the framework based on performance and processing time. Non-agricultural objects segmented using SAM were removed using an annual crop mask derived from Google Dynamic World. While performance was low to moderate in regions with small fields (<5ha in China, South Africa, and Spain), FieldSeg achieved a promising performance in the study areas with medium-large fields (>= 5 ha in Argentina, Australia, Brazil, USA-California, and USA-Iowa), with the rates of correctly extracted fields ranging from 0.541 to 0.814. The extracted fields showed a good segmentation quality, with mean dice coefficients ranging from 0.735 to 0.847. The large-scale applicability of FieldSeg was also demonstrated in four countries (1 million square kilometers), showing promising results and the ability to generalize across different regions.

关键词： SAM Remote sensing deep learning Crop monitoring Transformers

来源：评论

学校读者我要写书评

暂无评论

Optimizing deep learning on Sustainable Embedded Systems: A Study of Handwritten Digit Recognition with CNN and OpenCV

Optimizing Deep Learning on Sustainable Embedded Systems: A ...

引用

IEEE Region 10 International Conference TENCON

作者： Chiang Liang Kok Chee Kit Ho R Vicknesh Charles Lee Yit Yan Koh College of Engineering The University of Newcastle Australia Engineering Cluster Singapore Institute of Technology Singapore

ISBN: (数字)9798350350821

ISBN: (纸本)9798350350838

This senior thesis develops a real-time handwritten digit identification system using a Raspberry Pi 3B+ with a camera module, leveraging a lightweight CNN optimized with MNIST. The project highlights the effective implementation of deep learning on edge computing devices through seamless integration of CNN, TensorFlow Lite, and OpenCV's real-time image processing. The system is both cost-effective and precise, enabling real-time digit recognition tasks. This proposed work illustrates the potential of AI applications in education, industry, and commerce, setting the stage for future advancements in embedded AI systems.

关键词： deep learning Industries Handwriting recognition Embedded systems image processing real-time systems Convolutional neural networks Artificial intelligence IEEE Regions Edge computing

来源：评论

学校读者我要写书评

暂无评论

Emotion-Aware Multimedia Synthesis: A Generative AI Framework for Personalized Content Generation based on User Sentiment Analysis 2

Emotion-Aware Multimedia Synthesis: A Generative AI Framewor...

引用

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

作者： Sivasathiya, G. Anil Kumar, D. Harish Rangasamy, A.R. Kanishkaa, R. Chennai India

ISBN: (纸本)9798350327533

This research work introduces an innovative approach to multimedia content creation by incorporating emotion and sentiment analysis into a Generative Adversarial Network (GAN) framework. The system dynamically detects and interprets the user's emotional and sentiment cues, allowing for real-time adaptation and generation of multimedia content, including images, videos, and music. Leveraging advanced deep learning techniques, such as sentiment-aware GANs and emotion recognition through neural networks, the proposed framework establishes a seamless connection between user expression and media synthesis. By conditioning the generative process on the user's emotional state, the model learns to generate contextually relevant and emotionally resonant content. This research work encompasses an in-depth analysis of existing emotion recognition methods, the design and architecture of the proposed system, hardware and software requirements, as well as rigorous testing and performance evaluations. The outcome aims to redefine interactive multimedia experiences, empowering users to effortlessly communicate and translate the emotions into personalized and expressive digital content © 2024 IEEE.

关键词： deep learning image generative AI NLP Speech Recognition

来源：评论

学校读者我要写书评

暂无评论

SPECTRUM: A Multi-Component Pipeline for High-Quality image Synthesis 6

SPECTRUM: A Multi-Component Pipeline for High-Quality Image ...

引用

6th IEEE International Conference on Artificial Intelligence in Engineering and Technology, IICAIET 2024

作者： Kumar, Vishal Kavitha Nair, R. Amelesh, M. Choudhary, deepak A. Shree, Sachin Acharya Institute of Technology Artificial Intelligence and Machine Learning Bengaluru India

ISBN: (纸本)9798350389692

This research paper introduces a novel pipeline for image generation and enhancement, integrating advanced techniques from stable diffusion and deep neural network (DNN) super-resolution. The pipeline consists of three key components: Stable Diffusion v1.5, Stable Diffusion XL Refiner 1.0, and ESPCN (Enhanced Sub-Pixel Convolutional Network). Stable Diffusion v1.5 serves as the foundation for text-to-image generation, utilizing the innovative approaches of Self Attention Guidance (SAG) and Classifier-Free Guidance (CFG). CFG directs the model's output fidelity towards the provided prompt, ensuring faithful image generation. Meanwhile, SAG leverages internal self-attention maps to iteratively refine image details, resulting in higher-quality outputs. The Stable Diffusion XL Refiner 1.0 component enhances previously generated images, affording control over the influence of prior images on refinement. By adjusting this influence, users can tailor the refinement process to either maintain consistency with the previous image or explore new visual possibilities. ESPCN, a deep Neural Network Upscaler embedded within OpenCV, completes the pipeline by upscaling the generated image by a factor of two. Its compact design enables real-time operations, making it suitable for various applications. Through comprehensive experimentation and evaluation, this pipeline demonstrates remarkable capabilities in generating high-quality images from textual prompts and refining them to achieve desired visual outcomes. The integration of stable diffusion techniques with DNN super-resolution presents a promising avenue for advancing image generation and enhancement methodologies. © 2024 IEEE.

关键词： deep neural networks

来源：评论

学校读者我要写书评

暂无评论

Pepper bell leaf disease detection and classification using optimized convolutional neural network

引用

MULtimeDIA TOOLS AND APPLICATIONS 2023年第8期82卷 12065-12080页

作者： Mustafa, Hassan Umer, Muhammad Hafeez, Umair Hameed, Ahmad Sohaib, Ahmed Ullah, Saleem Madni, Hamza Ahmad Khwaja Fareed Univ Engn & Informat Technol Dept Comp Engn Rahim Yar Khan Pakistan Islamia Univ Bahawalpur Dept Comp Sci Bahawalpur Pakistan Khwaja Fareed Univ Engn & Informat Technol Dept Comp Sci Rahim Yar Khan Pakistan

Agriculture production plays a significant role in the country's economy. Diseases are quite natural and common among plants. Identification of diseases in plants is necessary for averting losses in the yield of agricultural products. Manual monitoring of plants requires expertise, immense effort, and excessive time. Automatic detection will not only help in reducing time and effort but will also help in detecting disease at an early stage, as soon as it will start appearing on plant leaves. Recently, image processing in agriculture has attained a surge of interest by researchers. This study presents a five-layered CNN model for automatic detection of plant disease utilizing leaf images. In order to better train a CNN model, 20,000 augmented images are generated. Experimental results demonstrate that proposed optimized-CNN model can predict pepper bell plant leaf as healthy or bacterial with 99.99% accuracy. Robust results make the proposed optimized-CNN model a preliminary warning tool that can be applied as a disease identification system in a real cultivation environment.

关键词： Leaf disease image classification deep learning Optimized convolutional neural network

来源：评论

学校读者我要写书评

暂无评论

A Multidimensional Communication Scheduling Method for Hybrid Parallel DNN Training

引用

IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS 2024年第8期35卷 1415-1428页

作者： Li, Shengwei Lu, Kai Lai, Zhiquan Liu, Weijie Ge, Keshi Li, Dongsheng Natl Univ Def Technol Coll Comp Natl Key Lab Parallel & Distributed Comp Changsha 410073 Peoples R China

The transformer-based deep neural network (DNN) models have shown considerable success across diverse tasks, prompting widespread adoption of distributed training methods such as data parallelism and pipeline parallelism. With the increasing parameter number, hybrid parallel training becomes imperative to scale training. The primary bottleneck in scaling remains the communication overhead. The communication scheduling technique, emphasizing the overlap of communication with computation, has demonstrated its benefits in scaling. However, most existing works focus on data parallelism, overlooking the nuances of hybrid parallel training. In this paper, we propose TriRace, an efficient communication scheduling framework for accelerating communications in hybrid parallel training of asynchronous pipeline parallelism and data parallelism. To achieve effective computation-communication overlap, TriRace introduces 3D communication scheduling, which adeptly leverages data dependencies between communication and computations, efficiently scheduling AllReduce communication, sparse communication, and peer-to-peer communication in hybrid parallel training. To avoid possible communication contentions, TriRace also incorporates a topology-aware runtime which optimizes the execution of communication operations by considering ongoing communication operations and real-time network status. We have implemented a prototype of TriRace based on PyTorch and Pipedream-2BW, and conducted comprehensive evaluations with three representative baselines. Experimental results show that TriRace achieves up to 1.07-1.45x speedup compared to the state-of-the-art pipeline parallelism training baseline Pipedream-2BW, and 1.24-1.81x speedup compared to the Megatron.

关键词： Training Computational modeling Processor scheduling Pipelines Data models Transformers Pipeline processing Distributed training deep learning hybrid parallelism communication scheduling

来源：评论

学校读者我要写书评

暂无评论

Ad Extension Augmentation through LSTM-NLP Synergy: A deep learning Approach 1

Ad Extension Augmentation through LSTM-NLP Synergy: A Deep L...

引用

1st IEEE International Conference on Cognitive Robotics and Intelligent Systems, ICC - ROBINS 2024

作者： Reddy, Medarametla Varshitha Nikhitha, Nagalla R.M.K. Engineering College Department of Artificial Intelligence and Data Science 601206 India

ISBN: (纸本)9798350372748

In the dynamic realm of digital advertising, enhancing click-through and conversion rates within ad extensions remains a significant challenge for agencies. Ad extensions play a pivotal role in amplifying traditional text ads by providing added information and context, thereby elevating the ad's informativeness, visibility, and click-through rates. This research addresses the hurdles faced by ad-agencies constrained by manual ad extension creation, the evolution of customer web content, and resource-intensive monitoring. This study presents a data-driven solution to revolutionize ad extension effectiveness and campaign conversion rates. This approach initiates with the automation of ad extension creation, eradicating the inconsistencies of manual content generation. By harnessing real-time insights from website scraping, up-to-the-minute data about customer brand offerings can be acquired. The extracted data is processed using several Natural Language processing (NLP) and clustering techniques. To effectively manage the dynamic nature of web content, Long Short-Term Memory (LSTM) neural networks are utilized. This deep learning model predicts word selection probabilities within ad extensions, facilitating the recommendation of terms in alignment with evolving web content. This solution also addresses the resource-intensive monitoring challenge through real-time sentiment analysis. This continual assessment provides timely insights for proactive adjustments, thus optimizing conversion rates. Hyperparameter tuning ensures optimal model accuracy. This model not only contributes to the digital advertising domain but also furnishes agencies with a transformative framework to surmount time-consuming manual processes, thereby enhancing their competitive edge and achieving accuracy of around 92%. © 2024 IEEE.

关键词： Sentiment analysis

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：