检索结果-内蒙古大学图书馆

5th IEEE International Conference for Emerging Technology, INCET 2024

作者： Devi, M. Shyamala Harivishnu, K.C. Suriya, Jaya V. Jayasuriya, K. Jason, Infant G.A. Farih, M.D. Department of Computer Science and Engineering Panimalar Engineering College Tamilnadu Chennai India Department of Artificial Intelligence and Data Science Panimalar Engineering College Tamilnadu Chennai India

ISBN: (纸本)9798350361155

Among the edible varieties of fungi, mushrooms possess the strongest nutrients found in the plant. Nevertheless, because of their great demand for food and significant benefits to medical research, it is imperative to identify the different varieties of mushrooms within the current species. The deep learning technology has recently rendering the CNN models towards classification of various object in interdisciplinary fields. This paper proposes Regularized Dense Layered LeNet (RDL-LeNet) that categorizes the mushroom types with high accuracy. The Mushrooms Classification dataset, which includes 3150 mushroom pictures was used for this execution. The Mushrooms Classification dataset was fitted with the existing CNN models to select the best CNN model. To determine the most efficient CNN framework, the Mushrooms Classification dataset was fitted using the available CNN models. When compared to the other CNN model, the LeNet performs better, with an accuracy rate of 90.67%. LeNet has now been chosen to adjust the quantity of dense layers. LeNet typically consist of three convolutions, two pooling layers, and two dense layers. In order to suggest the recommended framework RDL-LeNet, the LeNet is now changed by adding an additional dense layer that is made up of a single block of batch normalization, relu optimization, and drop out at the end that further compresses the output into vector size of 84. The RDL-LeNet is made up of three convolutional layers, two pooling layers, and three dense layers. The third dense layer consists of a single block that is made up of batch normalization, relu optimization, and dropout, ultimately leading to the final output layer. The suggested RDL-LeNet and the current CNN have been implemented with the Mushroom Classification dataset. The experiment demonstrates that the suggested RDL-LeNet model works better in the categorization of mushroom types, with a high accuracy of 98.68%. © 2024 IEEE.

关键词： Fungi

来源：评论

学校读者我要写书评

暂无评论

Classification of Tuberculosis Bacilli from Microscopic Sputum Smear Images Using Swin Transformer

Classification of Tuberculosis Bacilli from Microscopic Sput...

引用

2025 International Conference on Multi-Agent Systems for Collaborative Intelligence, ICMSCI 2025

作者： Vallabhaneni, Praneeth Satapathy, Ashutosh Indugula, Manisha Velagapudi Ramakrishna Siddhartha Engineering College Department of Artificial Intelligence and Data Science Vijayawada India Velagapudi Ramakrishna Siddhartha Engineering College Department of Computer Science and Engineering Vijayawada India

ISBN: (纸本)9798331509828

Tuberculosis (TB) has been a great challenge in the health world, and proper treatment requires proper diagnosis at the right time. This paper has classified the bacilli in sputum samples into single/simple and clump formations using deep-learning transformer techniques. The method starts with semantic segmentation, where TransUNet model has been used on sputum images to detect regions containing the bacilli. The superior visual task performer, the Swin Transformer, further classifies the regions identified. Accuracy, precision, recall, F1-score, and AUC-ROC are the metrics used to evaluate the model's performance. This classification contributes to TB diagnosis by quantifying bacterial load and categorizing infection severity. This type of differentiation may guide clinicians in developing more precise treatment strategies as the variability of infection severity may be observed in patients suffering from TB. Integrating the recent transformer models, such as Swin Transformer and TransUNet, into the microbiological diagnosis process will improve the speed of TB diagnosis, and automated bacterial classification will occur through an organized process. This study will help diagnose TB by improving the capacity of diagnosis, which subsequently helps in better decision-making regarding treatment and management, independent of human interpretation. The successful classification of TB bacilli resulted in an average accuracy of 93%, average precision of 93%, average recall of 92.5%, and average F1-score of 92.5%. © 2025 IEEE.

关键词： Semantic Segmentation

来源：评论

学校读者我要写书评

暂无评论

ReLU Activated Attention UNet Encipher Decipher Framework based Species Classification of Butterfly

ReLU Activated Attention UNet Encipher Decipher Framework ba...

引用

2024 International Conference on Advancements in Power, Communication and Intelligent Systems, APCI 2024

作者： Devi, M. Shyamala Jeeva, R. Jagadheeswaran, R. Kumar V, Hemand Erwin Nicholas, M. Keerthi Vasan, P. Panimalar Engineering College Department of Computer Science and Engineering Tamilnadu Chennai India Panimalar Engineering College Department of Artificial Intelligence and Data Science Tamilnadu Chennai India

ISBN: (纸本)9798350363289

The recognition and categorization of butterflies is crucial for the preservation of butterfly species in the fields of entomology, computer vision and deep learning. Environmentalists have long utilized butterflies as model organisms in the research on the effects of habitat degradation, dispersion, and warming temperatures. This paper recommends Relu Activated Attention UNet (RAA-UNet) that categorizes the nine butterfly species with high precision. The Butterfly dataset contains 1664 butterfly images that were used in this investigation. The dataset consists of 832 photos of butterflies and 832 related images that have been segmented. The proposed RAA-UNet starts by creating masked butterfly pictures by masking the original image with a segmented butterfly image. To classify the butterfly species, the masked butterfly pictures are assigned to Attention UNet with Decoder and Encoder combined with the Attention gate and activated using Relu Activation function. The proposed RAA-UNet and conventional CNN models are fitted to the masked butterfly pictures. As demonstrated by the experiments, the RAA-UNet model works better with a high accuracy of 97.75% in the classification of butterfly species. © 2024 IEEE.

关键词： Invertebrates

来源：评论

学校读者我要写书评

暂无评论

Traffic flow prediction based on graph convolutional networks with a parallel attention network and stacked gate recurrent units

引用

Multimedia Tools and Applications 2025年第15期84卷 14329-14358页

作者： Xia, Dawen Ao, Yuce Wei, Xiaoduo Li, Yunsong Chen, Yan Hu, Yang Li, Yantao Li, Huaqing College of Data Science and Information Engineering Guizhou Minzu University Guiyang550025 China Department of Automotive Engineering Guizhou Traffic Technician and Transportation College Guiyang550008 China College of Computer Science Chongqing University Chongqing400044 China College of Electronic and Information Engineering Southwest University Chongqing400715 China

Accurate traffic flow prediction is essential to address traffic issues and assist traffic managers make informed decisions in intelligent transportation systems. Extracting potential features from traffic data is challenging due to the complex topology of urban road networks and the time-varying traffic flow. To capture the global spatiotemporal characteristics of traffic flow, we propose a novel model based on graph convolutional networks with a parallel attention network and stacked gated recurrent units (PAGCN-SGRU). First, the parallel attention (PA) network enhances the feature representation of global traffic road nodes and road segments. Then, the graph convolutional networks (GCN) are designed to extract spatial characteristics. Next, the stacked gate recurrent units (SGRU) are employed to capture temporal features. Finally, PAGCN-SGRU discovers global spatiotemporal features for traffic flow prediction. The experimental results demonstrate that the accuracy of PAGCN-SGRU under the SZ-dataset is improved by 9.76%, 72.54%, 5.76%, 16.07%, 2.07%, 1.82%, 3.35%, and 6.59%, respectively, compared to that of HA, ARIMA, SVR, GCN, T-GCN, A3T-GCN, ST-GCN, and DCRNN. In the Los-dataset, the accuracy values increase by 7.11%, 8.94%, 6.66%, 7.77%, 3.43%, 2.45%, 2.28%, and 4.09%, respectively. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2024.

关键词： Forecasting

来源：评论

学校读者我要写书评

暂无评论

Auxiliary Dense Layered LeNet Deep Learning based Mango Variety Classification 2

Auxiliary Dense Layered LeNet Deep Learning based Mango Vari...

引用

2nd International Conference on Advancement in Computation and computer Technologies, InCACCT 2024

作者： Shyamala Devi, M. Janani, G. Fernandes, Jonita Maria Harshini, T. Nivetha, R. Joshika Rao, O.R. Panimalar Engineering College Department of Computer Science and Engineering Tamilnadu Chennai India Panimalar Engineering College Department of Artificial Intelligence and Data Science Tamilnadu Chennai India

ISBN: (纸本)9798350371314

In the agriculture sector, physical classification of fruits is a costly process that can produce inconsistent outcomes due to human negligence. Fruit categorization from snapshots is an extremely difficult venture, especially for mango fruits because various species of mango fruits share significant resemblance. This paper proposes Auxiliary Dense Layered LeNet (ADL-LeNet) that classifies the mango type with high accuracy. The Mango Varieties Classification and Grading dataset, which includes 1200 mango pictures was used for this work. The Mango image dataset is applied with the existing CNN model to select the best CNN model. The LeNet outperforms with the accuracy of 92.77% when compared to the other CNN model. Now LeNet is selected for finetuning the number of dense layers. Generally, the LeNet is composed of three convolutions, two subsampling followed by two dense layers. Now, the LeNet is modified by adding one extra dense layer at the end that compress the output further into vector size of 42 to suggest the proposed model ADL-LeNet. The ADL-LeNet composed of three convolutional, two subsampling and three dense layers ending with the final output layer. The Mango Image dataset has been applied with the existing CNN and proposed ADL-LeNet. The implementation shows that the proposed ADL-LeNet model outperforms with the high accuracy of 97.6% in mango type classification. © 2024 IEEE.

关键词： Fruits

来源：评论

学校读者我要写书评

暂无评论

An implementation of personalized therapy in Clinical Decision Support System using adaptive transformer and hybrid deep learning network

引用

Australian Journal of Electrical and Electronics engineering 2025年第1期22卷 105-120页

作者： Praveena Rachel Kamala, S. T, Saraswathi. Kaliraj, V. Nithya, A. Hema, R. Ramyadevi, R. Department of Information Technology Easwari Engineering college Tamil Nadu Chennai India Department of CSE S.A Engineering College Tamil Nadu Chennai India Department of Computer Science and Engineering St.Joseph’s College of Engineering Tamil Nadu Chennai India Department of Electronics and Communication Engineering Easwari Engineering College Tamil Nadu Chennai India Department of Computer Science and Applications SRM Institute of Science and Technology Tamil Nadu Chennai India

This paper suggests a new mechanism from deep learning concept for personalised therapy in Clinical Decision Support Systems (CDSS). Basically, the texts used for the observation are acquired from the standard data sources and then forwarded to the text preprocessing task. In the pre-processing phase, the punctuation and special character removal, stop word removal, and stemming process are applied to remove noise and help to eliminate the redundant information in order to improve the data quality. Further, the pre-processed text is applied to the Adaptive Transformer Net (ATN) for the feature extraction purpose, where the attributes in this task are optimally determined with the aid of the Adaptive Walrus Optimization Algorithm (AWOA). Finally, the resultant text is subjected to the Hybrid Deep Learning Network (HDLNet). The HDLNet model is implemented by integrating the ‘Residual Long Short-Term Memory (Residual LSTM) with Dilated Recurrent Neural Network (Dilated RNN)’. From the results, the sensitivity analysis performed in the implemented technique secured 3.7% more efficient than LSTM, 7.76% improved than MobileNet, 6.7% superior to residual LSTM, and 0.90% effective than dilated RNN in dataset 1. Throughout the validation, the conventional techniques are evaluated with the suggested personalised therapy in CDSS to prove its efficacy. ©, Engineers Australia.

关键词： Long short-term memory

来源：评论

学校读者我要写书评

暂无评论

Diluie: constructing diverse demonstrations of in-context learning with large language model for unified information extraction

引用

Neural Computing and Applications 2024年第22期36卷 13491-13512页

作者： Guo, Qian Guo, Yi Zhao, Jin Department of Computer Science and Engineering East China University of Science and Technology Shanghai200237 China Shanghai Key Laboratory of Data Science School of Computer Science Fudan University Shanghai200433 China

Large language models (LLMs) have demonstrated promising in-context learning capabilities, especially with instructive prompts. However, recent studies have shown that existing large models still face challenges in specific information extraction (IE) tasks. Moreover, it could have more effectively utilized various prompts such as instruction tuning, diverse demonstrations of in-context learning, and long-range token sequences for assisting language modeling in understanding context. In this study, we propose DILUIE, a unified information extraction framework based on in-context learning with diverse demonstration examples. DILUIE is encoded with an EVA attention mechanism and incremental encoding technology. Based on the constructed diverse demonstrations, we expand the size of instances efficiently in both instruction tuning and in-context learning to gain insights into the potential benefits of utilizing diverse information extraction datasets. To deepen the understanding of context, we further design three auxiliary tasks to assist in aligning contextual semantics. Experimental results demonstrate that DILUIE achieves 2.23 and 2.53% improvements in terms of Micor-/Macor-F1 on average relative to the current state-of-the-art baseline, which also significantly outperforms the GPT-3.5-turbo in zero-shot settings, and the average token length of achieving the best performance over tasks is around 15k. Furthermore, we observe that in-context learning shows enhanced performance when provided with more demonstrations during multiple-shot instruction tuning (8 k). Additionally, increasing the length of instructions (10 k) can result in a more substantial improvement in the upper limits of scaling for in-context learning. Code is available on https://***/Phevos75/DILUIE. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2024.

关键词： Demonstrations

来源：评论

学校读者我要写书评

暂无评论

Interpretable Detection of Malicious Behavior in Windows Portable Executables Using Multi-Head 2D Transformers

引用

Big data Mining and Analytics 2024年第2期7卷 485-499页

作者： Sohail Khan Mohammad Nauman Computer Science Department Effat College of EngineeringEffat UniversityJeddah 23341Kingdom of Saudi Arabia

Windows malware is becoming an increasingly pressing problem as the amount of malware continues to grow and more sensitive information is stored on *** of the major challenges in tackling this problem is the complexity of malware analysis,which requires expertise from human *** developments in machine learning have led to the creation of deep models for malware ***,these models often lack transparency,making it difficult to understand the reasoning behind the model’s decisions,otherwise known as the black-box *** address these limitations,this paper presents a novel model for malware detection,utilizing vision transformers to analyze the Operation Code(OpCode)sequences of more than 350000 Windows portable executable malware samples from real-world *** model achieves a high accuracy of 0.9864,not only surpassing the previous results but also providing valuable insights into the reasoning behind the *** model is able to pinpoint specific instructions that lead to malicious behavior in malware samples,aiding human experts in their analysis and driving further advancements in the *** report our findings and show how causality can be established between malicious code and actual classification by a deep learning model,thus opening up this black-box problem for deeper analysis.

关键词： machine learning malware vision transformers Windows Protable Executable(PE)

来源：评论

学校读者我要写书评

暂无评论

Revolutionising facility layout: a case study of dynamic facility layout in cable production

引用

International Journal of Web engineering and Technology 2024年第3期19卷 267-309页

作者： Vineetha, G.R. Rasheedali, Shiyas Chekkot Division of Computer Science and Engineering Department of Mechanical Engineering Cochin University College of Engineering CUSAT Kuttanadu Kerala Alappuzha India

In the competitive landscape of globalised markets, businesses must prioritise cost reduction for sustained competitiveness. This study delves into the dynamic facility layout problem (DFLP) within a cable production company in Kerala, emphasising adaptability to changing production demands. Addressing material handling costs and rearrangement expenses, the research evaluates the efficacy of the current static layout and explores the benefits of transitioning to a dynamic layout. The case study reveals potential cost savings through the strategic restructuring of machine arrangements. The innovative machine learning-based genetic algorithm (ML-GA) integrates machine learning algorithms, genetic algorithms, and a local search method, offering a cutting-edge solution to dynamic facility layout challenges. By considering demand variability and relocation costs, the study provides insights for informed decision-making, emphasising the significance of material flow patterns. This research contributes to enhancing efficiency and profitability, providing practical implications for businesses navigating the complexities of modern manufacturing. Copyright © 2024 Inderscience Enterprises Ltd.

关键词： Plant layout

来源：评论

学校读者我要写书评

暂无评论

Classification of autism severity levels using facial features and eye gaze patterns

引用

Multimedia Tools and Applications 2024年第16期84卷 1-14页

作者： Jiji, G. Wiselin Department of Computer Science and Engineering Dr. Sivanthi Aditanar College of Engineering Tamil Nadu Thiruchendur India

Autism Spectrum Disorder (ASD) is a complex neurodevelopmental condition with varying degrees of severity. Early diagnosis and classification of autism severity are crucial for personalized intervention and support. This study proposes a novel approach using image processing techniques to analyze facial features and eye gaze patterns to differentiate between mild, moderate, and severe autism cases. A comprehensive dataset comprising facial videos and images of individuals aged 2 to 12 years with ASD was collected, along with normally developing children. Features such as facial asymmetry is calculated using SIFT feature, eye size, inter-eye distance, and eye openness were extracted for both groups using canny and adaptive thresholding techniques. Support Vector Machine (SVM) classifiers were employed to classify autism cases into three severity levels. Results revealed that eye gaze patterns were significantly lower for autism cases and higher for severe cases. Facial asymmetry was higher for autism cases, showing greater deviations from mild to severe cases. Severe autism cases exhibited extreme stiffness in facial muscle control, leading to the absence of facial expressions. Additionally, inter-eye distance increased, and eye openness decreased for severe autism cases. The proposed method demonstrates promising discrimination performance, as evidenced by high accuracy of 98.70 and sensitivity and specificity of 100 and 97.30. This research contributes valuable insights into the potential use of image processing techniques for early autism diagnosis and effective severity classification. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2024.

关键词： Classification (of information)

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：