检索结果-内蒙古大学图书馆

2nd IEEE International Conference on Networks, Multimedia and Information Technology, NMITCON 2024

作者： Deepa, N. Kumar, Bura Vijay Sowmya, M. Al-Jawahry, Hassan M. Bhushan, Yannam Bharath Saveetha Institute of Medical and Technical Sciences Saveetha School of Engineering Department of Computer Science and Engineering Chennai India Sr University School of Computer Science and Artificial Intelligence Warangal India Nitte Meenakshi Institute of Technology Department of Ai & Ds Bengaluru India The Islamic University College of Technical Engineering Department of Computers Techniques Engineering Najaf Iraq Gokaraju Rangaraju Institute of Engineering and Technology Department of Ai&ml Bachupally Hyderabad India

ISBN: (纸本)9798350372892

In daily life, the moving vehicle plays a significant part in effective system of transportation and traffic visual surveillance systems. Most critical task is to detect the vehicle image and most of the existing method, concentrated on maximizing the prediction percentage only. Also it lacks in classifying the imbalance data. To overcome this kind of problem, the Gaussian Multivariate Mixture Model (G-MMM) and VGG19 model is implemented to detect and classify the vehicle types. The proposed method contained four process namely, preprocessing performed by Flexible Histogram Equalization method (FHE), vehicle detection using G-MMM, feature extraction using Steerable Pyramid Transform (SPT) and Weber Local Descriptor (WLD) and classification using VGG19 ensemble deep learning model. G-MMM significantly improves detection sensitivity and it transformed the correlation between attributes through the training process. VGG19 is pretrained model, so it allows the fast learning recognition on new images of vehicle. The proposed method tested on Beijing Institute of Technology (BIT) dataset and MIO vision traffic camera dataset. When compared to the others methods like modified AlexNet, Deep Convolution Neural Network (DCNN), this proposed G-MMM-VGG19 gives 99.47 % of output rate in classification phase. © 2024 IEEE.

关键词： Equalizers

来源：评论

学校读者我要写书评

暂无评论

Advancements in Brain Disease Diagnosis for Comparing Traditional Methods and Deep Learning Approaches in Medical Imaging 1

Advancements in Brain Disease Diagnosis for Comparing Tradit...

引用

1st International Conference on Emerging Technologies and Innovation for Sustainability, EmergIN 2024

作者： Kumari, Sunita Singh, Laxman Pasha, Moghal Irfan Kumaraswamy, M. Choudhary, Amar Agarwal, Monika Department Computer Science and Engineering ABES Institute of Technology Ghaziabad India School of Computer Science Engineering and Technology Bennett University Greater Noida India Department of CSE-AI&ML CVR College of Engineering Hyderabad India Department of CST Sasi Institute of Technology and Engineering Tadepalligudem India Alliance University Bengaluru India Department of Computer Science and Engineering Raj Kumar Goel Institute of Technology Ghaziabad India

ISBN: (纸本)9798350391268

This work explores improvements in the diagnosis of brain disease through traditional vs deep learning base techniques with respect to Medical Imaging. The dataset containing 3,000 images of brain tumors was preprocessed to resize (224x224), normalize and augmentation techniques like rotation, flipping were further implemented. The dataset are splitted into70% training and 30%. testing across four deep learning models, Convolutional Neural Networks (CNN), VGG16. The models were trained to classify brain tumor images into more or less complex categories. After training VGG19 gave top accuracy of 98.6 per cent, followed byVVG16 at second place with an accuracy rate of 96.75%, while the ResNet50 and CNN were rated third and fourth in terms of their accuracies;as neither performed well with just92.3%and89.3%. VGG19 showed the best precision (98.8%), recall(98.5%) and F1-score(98.65) among metrics of all models utilised in this research(which ever they might be). VGG19 had low false positives and negatives which are important for accurate diagnosis as can be seen in the confusion matrices. The type of research exemplifies how deep learning can improve the accuracy and certainty within medical imaging diagnostics, particularly in difficult tasks such as brain tumor classification. This study highlights the strong possibilities deep learning could lead to transformative improvements in clinical neuroscience by employing advanced neural network architectures and meticulous preprocessing strategies. © 2024 IEEE.

关键词： Neuroimaging

来源：评论

学校读者我要写书评

暂无评论

Optimized Image Compression Using Multiple Compressed Sensing Techniques

引用

SN computer science 2025年第4期6卷 1-18页

作者： Puttegowda, Kiran Mohan, B.A. Veeraprathap, V. Vijay, C.P. Sudheesh, K.V. Kumar, D. S. Sunil Department of ECE Vidyavardhaka College of Engineering Karnataka Mysuru India Department of ISE B.M.S. Institute of Technology and Management Yelahanka Karnataka Bengaluru India Department of ECE ATME college of Engineering Karnataka Mysuru India Department of Computer Science Engineering (AI & ML) Vidyavardhaka College of Engineering Mysuru Karnataka India Department of Computer Applications East Point College of Higher Education Banglore India

In the digital era, the demand for efficient image compression techniques is ever-increasing due to the exponential growth of visual data. Compressed sensing (CS), a groundbreaking theory in signal processing, offers a promising solution by exploiting the inherent sparsity of images. The proposed work introduces a novel approach to image compression, leveraging compressed sensing techniques to achieve optimized results. Our method works by expressing image data in a higher-order non-raw and non-dense format hence dramatically decreasing the size of the entire data. This significantly reduces storage requirements while preserving image quality. By combining advanced signal processing with innovative image analysis techniques, our approach maintains low computational complexity, enabling real-time applications without sacrificing the fidelity of reconstructed images, even after compression. This makes our approach especially valuable for applications where storage space is limited but image integrity remains crucial. We present experimental results demonstrating the effectiveness of our technique across various image types and compression ratios. Our findings underscore the potential of compressed sensing as a transformative paradigm in the realm of image compression, paving the way for more efficient storage and transmission of visual content in diverse applications. © The Author(s), under exclusive licence to Springer Nature Singapore Pte Ltd. 2025.

关键词： Compressed Sensing Compression Compression ratio Images Reconstruction

来源：评论

学校读者我要写书评

暂无评论

A Deep Learning and Raspberry PI Investigation for Epilepsy Seizure Detection 3

A Deep Learning and Raspberry PI Investigation for Epilepsy ...

引用

3rd International Conference on Artificial Intelligence, Computational Electronics and Communication System, aiCECS 2024

作者： Raibag, Mohammad Asif Rashmi, B.C. Gorikhan, Irshad Ahmad Sarkar, Rashel Deka, Nilakshi Yenepoya Institute of Technology Moodbidri Computer Science & Engineering Mangalore India RNS Institute of Technology CSE( AI & ML) Bangalore India CSE(AI & ML) AGMRCET Varur Hubli Dharwad India Assam Royal Global University Computer Science & Engineering Guwahati India

ISBN: (纸本)9798350391244

A common neurological condition that affects over 65 million people globally is epilepsy. Most of the time, patients with this ailment do not respond to medicine or surgery. However, therapeutic intervention that anticipates a seizure before it happens can help avoid it. Research indicates that the preictal state - a state of abnormal brain activity - begins just prior to the actual seizure. It is still challenging to accurately and very sensitively forecast the preictal condition of a seizure, despite multiple experiments. The proposed work uses deep learning technology to develop an automated seizure prediction system. The EEG data is preprocessed and some pertinent characteristics are extracted before being classified using a deep learning system. Furthermore, a precise seizure detection system constructed around a Raspberry Pi is developed to anticipate the start of seizures. © 2024 IEEE.

关键词： ANN Deep Learning EEG Epilepsy Raspberry Pi

来源：评论

学校读者我要写书评

暂无评论

PEAC: Unsupervised Pre-training for Cross-Embodiment Reinforcement Learning 38

PEAC: Unsupervised Pre-training for Cross-Embodiment Reinfor...

引用

38th Conference on Neural Information Processing Systems, NeurIPS 2024

作者： Ying, Chengyang Hao, Zhongkai Zhou, Xinning Xu, Xuezhou Su, Hang Zhang, Xingxing Zhu, Jun Department of Computer Science & Technology Institute for AI BNRist Center Tsinghua-Bosch Joint ML Center THBI Lab Tsinghua University China Guangzhou China

Designing generalizable agents capable of adapting to diverse embodiments has achieved significant attention in Reinforcement Learning (RL), which is critical for deploying RL agents in various real-world applications. Previous Cross-Embodiment RL approaches have focused on transferring knowledge across embodiments within specific tasks. These methods often result in knowledge tightly coupled with those tasks and fail to adequately capture the distinct characteristics of different embodiments. To address this limitation, we introduce the notion of Cross-Embodiment Unsupervised RL (CEURL), which leverages unsupervised learning to enable agents to acquire embodiment-aware and task-agnostic knowledge through online interactions within reward-free environments. We formulate CEURL as a novel Controlled Embodiment Markov Decision Process (CE-MDP) and systematically analyze CEURL's pre-training objectives under CE-MDP. Based on these analyses, we develop a novel algorithm Pre-trained Embodiment-Aware Control (PEAC) for handling CEURL, incorporating an intrinsic reward function specifically designed for cross-embodiment pre-training. PEAC not only provides an intuitive optimization strategy for cross-embodiment pre-training but also can integrate flexibly with existing unsupervised RL methods, facilitating cross-embodiment exploration and skill discovery. Extensive experiments in both simulated (e.g., DMC and Robosuite) and real-world environments (e.g., legged locomotion) demonstrate that PEAC significantly improves adaptation performance and cross-embodiment generalization, demonstrating its effectiveness in overcoming the unique challenges of CEURL. The project page and code are in https://***/ceurl. © 2024 Neural information processing systems foundation. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Smart System for Missing Child Identification

Smart System for Missing Child Identification

引用

International Conference on ent Computing and Control Systems (ICICCS)

作者： Vineela Madireddy Jonnalagadda Varsha K. Ram Prasad T. Sai Bhargava Department of Computer Science and Engineering (AI & ML) Vignan Institute of Technology and Science Nalgonda Telangana India

ISBN: (数字)9798331512088

ISBN: (纸本)9798331512095

Identifying and recovering missing children is a significant challenge faced by law enforcement agencies and social organizations worldwide. Every year, many individuals go missing, and a considerable number of them are never found. The complexity of these cases increases when missing individuals are located far from their original location, requiring substantial time and effort to confirm their identities. Traditional identification methods, which depend heavily on manual efforts and police resources, are often slow and inefficient. Delays in these processes can lead to serious consequences, reducing the chances of finding a missing child promptly. To overcome these challenges, a System is introduced that uses Convolutional Neural Networks (CNN) for extracting facial features and the K-Nearest Neighbors (KNN) algorithm for matching images. This approach aims to enhance the identification process by comparing photos of found children with a centralized database of missing children. The system allows the public to upload images of children who might be missing, which are then automatically analyzed and matched against the stored images. Additionally, users can provide relevant information such as location details, landmarks, and contact information. Designed to adapt to variations in age and image quality, the CNN model ensures reliable matching. Upon receiving an image, the system performs an automated comparison—either identifying a match or storing the image for future reference. If a match is found, the system promptly provides relevant recovery details to assist in reuniting the child with their family. By leveraging automation and machine learning, this approach significantly reduces identification time, increases efficiency, and enhances the success rate of locating missing children. Optimizations like FaiSS indexing, and dimensionality reduction enhance efficiency, ensuring real-time identification with reduced computational overhead and improved scalability.

关键词： Law enforcement Scalability Organizations Manuals Nearest neighbor methods Smart systems Reliability engineering Real-time systems Convolutional neural networks Optimization

来源：评论

学校读者我要写书评

暂无评论

Fourier Controller Networks for Real-Time Decision-Making in Embodied Learning 41

Fourier Controller Networks for Real-Time Decision-Making in...

引用

41st International Conference on Machine Learning, ICml 2024

作者： Tan, Hengkai Liu, Songming Ma, Kai Ying, Chengyang Zhang, Xingxing Su, Hang Zhu, Jun Department of Computer Science and Technology Institute for AI Tsinghua-Bosch Joint ML Center THBI Lab BNRist Center Tsinghua University Beijing100084 China

Transformer has shown promise in reinforcement learning to model time-varying features for obtaining generalized low-level robot policies on diverse robotics datasets in embodied learning. However, it still suffers from the issues of low data efficiency and high inference latency. In this paper, we propose to investigate the task from a new perspective of the frequency domain. We first observe that the energy density in the frequency domain of a robot’s trajectory is mainly concentrated in the low-frequency part. Then, we present the Fourier Controller Network (FCNet), a new network that uses Short-Time Fourier Transform (STFT) to extract and encode time-varying features through frequency domain interpolation. In order to do real-time decision-making, we further adopt FFT and Sliding DFT methods in the model architecture to achieve parallel training and efficient recurrent inference. Extensive results in both simulated (e.g., D4RL) and real-world environments (e.g., robot locomotion) demonstrate FCNet’s substantial efficiency and effectiveness over existing methods such as Transformer, e.g., FCNet outperforms Transformer on multienvironmental robotics datasets of all types of sizes (from 1.9M to 120M). The project page and code can be found https://***/fcnet. Copyright 2024 by the author(s)

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Search if you don't know! Knowledge-Augmented Korean Grammatical Error Correction with Large Language Models

Search if you don't know! Knowledge-Augmented Korean Grammat...

引用

2024 Findings of the Association for Computational Linguistics, EMNLP 2024

作者： Koo, Seonmin Kim, Jinsung Park, Chanjun Lim, Heuiseok Department of Computer science and Engineering Korea University Korea Republic of Upstage AI Korea Republic of

ISBN: (纸本)9798891761681

Grammatical error correction (GEC) system is a practical task used in the real world, showing high achievements alongside the development of large language models (LLMs). However, these achievements have been primarily obtained in English, and there is a relative lack of performance for non-English data, such as Korean. We hypothesize that this insufficiency occurs because relying solely on the parametric knowledge of LLMs makes it difficult to thoroughly understand the given context in the Korean GEC. Therefore, we propose a Knowledge-Augmented GEC (KAGEC) framework that incorporates evidential information from external sources into the prompt for the GEC task. KAGEC first extracts salient phrases from the given source and retrieves non-parametric knowledge based on these phrases, aiming to enhance the context-aware generation capabilities of LLMs. Furthermore, we conduct validations for fine-grained error types to identify those requiring a retrieval-augmented manner when LLMs perform Korean GEC. According to experimental results, most LLMs, including ChatGPT, demonstrate significant performance improvements when applying KAGEC. © 2024 Association for Computational Linguistics.

关键词： Modeling languages

来源：评论

学校读者我要写书评

暂无评论

Enhanced Medical Tumor Detection

Enhanced Medical Tumor Detection

引用

Pervasive Computing and Social Networking (ICPCSN), International Conference on

作者： Vineela Madireddy Shaik Rehmat Sultana Muddasani Ashvitha Yadasu Ajay Department of Computer Science and Engineering (AI&ML) Vignan Institute of Technology and Science Nalgonda Telanagana

ISBN: (数字)9798331535193

ISBN: (纸本)9798331535209

Medical Imaging plays a crucial role in diagnosing and treating various diseases, but relying on a single imaging modality often limits the depth of the information available to doctors. This paper introduces a pioneering deep learning-based approach to enhance the integration of multiple imaging techniques, like MRI, CT, and PET for more accurate diagnosis and treatment planning. By intelligently gathering data from these different imaging sources, our approach provides a more detailed and holistic view of medical conditions, helping doctors detect tumours more accurately and reduce diagnostic uncertainties. This advancement not only aids healthcare professionals in making more informed decisions but also paves the way for personalize treatment strategies. To increase accuracy and precision, enhanced medical tumour detection uses deep learning models, sophisticated imaging methods, and optimal data processing. While machine learning and deep learning models such as CNNs, Vision Transformers (ViTs), and hybrid models guarantee reliable tumour classification and segmentation, multi-modal imaging (MRI, CT, and PET) improves visualization. Techniques like ensemble learning, feature extraction, and data pretreatment improve detection performance even more. Precise boundary detection is ensured by accurate segmentation utilizing U-Net, DeepLabV3+, and Mask R-CNN. Model performance is validated by evaluation criteria like sensitivity, specificity, AUC-ROC, Jaccard Index, and dice coefficient. According to experimental findings, employing hybrid deep learning techniques ensures accurate medical tumour diagnosis with accuracy above 95% and precision above 92%.

关键词： Deep learning Image segmentation Accuracy Uncertainty Magnetic resonance imaging Computed tomography Convolutional neural networks Positron emission tomography Medical diagnostic imaging Tumors

来源：评论

学校读者我要写书评

暂无评论

Hybrid CNN and RNN Variant Model for Indian Music Genre Classification

Hybrid CNN and RNN Variant Model for Indian Music Genre Clas...

引用

Smart Systems for applications in Electrical sciences (ICSSES), International Conference on

作者： Swathi Surapally Siva Ramakrishna Jeevakala Sriya Pranathi Aluri Computer Science and Engineering (AI & ML) Institute of Aeronautical Engineering Hyderabad India

ISBN: (数字)9798331507671

ISBN: (纸本)9798331507688

Music genre classification contributes largely to information retrieval for organizing growing collections of music. The process of classification of music is very difficult and unreliable. Many methods utilized handcrafted features to identify unique patterns but were still unable to determine original music characteristics. In comparison, music classification by applying deep learning models exhibits a dynamic and effective potential. Although there are a large number of neural networks, the combination of CNN and variants of RNN has not received adequate consideration. Moreover, on the shortcomings of the specific neural network classification model, we would like to propose a hybrid architecture of CNN and variant of RNN, such as a GRU. In addition, we compare the performances based on Mel-spectrogram and MFCC features.

关键词： Analytical models Time-frequency analysis Tagging Smart systems Spatial databases Robustness Convolutional neural networks Tuning Spectrogram Recommender 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：