检索结果-内蒙古大学图书馆

20th International Conference on Intelligent Computing (ICIC)

作者： Liu, Haotian Zhong, Yixin Chen, Yuehui Cao, Yi Zhao, Yaou Univ Jinan Sch Informat Sci & Engn Artificial Intelligence Inst Jinan Peoples R China Univ Jinan Shandong Prov Key Lab Network Based Intelligent C Sch Informat Sci & Engn Jinan Peoples R China

ISBN: (纸本)9789819756773;9789819756780

Chinese character stroke segmentation aims to extract all strokes from a Chinese character image. Since strokes can intersect each other, the algorithm tends to divide one single stroke into several segments by the intersection point. Traditional methods extract the strokes based on pre-defined matching rules, which is not effective for complex characters. Deep learning-based methods can automatically learn the stroke segmentation rules through labelled data set. However, many of these methods rely solely on convolutions feature extraction. As convolution focuses more on the local patterns and is weak in capturing distant correlations, the strokes generated by the convolutional neural networks are often incomplete. In addition, the number of strokes among Chinese characters differ greatly, which implicitly unbalances the training data. Model performance varies considerably according to the number of strokes in the character. To address the above issues, this paper proposes a novel stroke segmentation network, namely, SGAN, which is based on generative adversarial networks to directly produce images containing extracted strokes. The strokes are grouped into 7 categories, each of which is generated by an output channel of SGAN. The axial attention and external attention are applied in SGAN in order to capture global features and improve the segmentation accuracy. To enhance the fidelity of generated stroke images, a series of post-processing algorithms are proposed to remove redundancies and missing parts. The effectiveness of the proposed method is verified over mainstream datasets, where SGAN has achieved outstanding performance.

关键词： Chinese character stroke segmentation GAN Attention

来源：评论

学校读者我要写书评

暂无评论

Interval Type-2 Fuzzy Restricted Boltzmann Machine

引用

IEEE TRANSACTIONS ON FUZZY SYSTEMS 2021年第5期29卷 1133-1142页

作者： Hosseini-Pozveh, Marzieh Sadat Safayani, Mehran Mirzaei, Abdolreza Isfahan Univ Technol Dept Elect & Comp Engn Esfahan *** Iran

Restricted Boltzmann machine (RBM) is an energy-based artificial neural network (ANN), applied in several applications like image processing, topic modeling, classification, regression, and pattern recognition. The fuzzy version of RBM is a new approach in this field, with parameters considered as fuzzy numbers. In this article, a fuzzy RBM is extended through interval type-2 membership functions, named the interval type-2 fuzzy RBM (IT2FRBM). The additional uncertainties in the structures of the membership functions are embedded in this model. This is formulated as a maximum likelihood problem which allows the parameters of the type-2 fuzzy numbers to be learned. The capabilities of this proposed approach as a discriminative or generative model are assessed. The robustness of this method against noise is analyzed. The results indicate that this IT2FRBM outperforms RBM and its different fuzzy versions.

关键词： Uncertainty Fuzzy systems Fuzzy sets artificial neural networks image processing Computational modeling Fuzzy numbers fuzzy RBM neural network restricted Boltzmann machine type-2 fuzzy

来源：评论

学校读者我要写书评

暂无评论

Energy-Efficient Architecture and Dataflow Optimization for Spiking neural Network Accelerators

Energy-Efficient Architecture and Dataflow Optimization for ...

引用

作者： Lee, Jeong-Jun University of California Santa Barbara

学位级别：Ph.D., Doctor of Philosophy

Spiking neural networks (SNNs) offer a promising biologically-plausible computing model and lend themselves to ultra-low-power event-driven processing on neuromorphic processors. Compared with the conventional artificial neural networks, SNNs are well-suited for processing complex spatiotemporal *** this dissertation, we aim to address key difficulties in accelerating SNNs: developing bio-plausible and hardware friendly algorithm, efficient processing of the added temporal dimension, and handling unstructured sparsity emergent in both space and time. First, training SNNs to reach the same performancesof conventional deep artificial neural networks (ANNs), particularly with error backpropagation (BP) algorithms, poses a significant challenge due to inherent complex dynamics and non-differentiable spike activities of spiking neurons. In this dissertation, we present the first study on realizing competitive spike-train level backpropagation (BP) like algorithms to enable on-chip training of SNNs. We propose a novel spike-train level direct feedback alignment (ST-DFA) algorithm, which is much more bio-plausible and hardware friendly than BP. Algorithm and hardware co-optimization and efficient online neural signal computation are explored for on-chip implementation of ST-DFA. On the Xilinx ZC706 FPGA board, the proposed hardware-efficient ST-DFA shows excellent performance vs. overhead tradeoffs for real-world speech and image classification applications. Despite its significance, dataflow optimization of spiking neural accelerator architectures has not been extensively studied. Recognizing the need for efficient processing of complex spatiotemporal data while considering the all-or-none nature of spiking activities, we propose holistic reconfigurable dataflow optimization for systolic array acceleration of spiking convolutional networks (S-CNNs). A novel scheme is introduced for parallel acceleration of computation across multiple time points, which further allows

关键词： neural network accelerators Dataflow optimization Energy-efficient architecture

来源：评论

学校读者我要写书评

暂无评论

A deep learning framework for early diagnosis of Alzheimer's disease on MRI images

引用

MULTIMEDIA TOOLS AND applications 2023年第2期83卷 3767-3799页

作者： Arafa, Doaa Ahmed Moustafa, Hossam El-Din Ali, Hesham A. Ali-Eldin, Amr M. T. Saraya, Sabry F. Mansoura Univ Fac Engn Comp Engn & Control Syst Dept Mansoura Egypt Mansoura Univ Fac Engn Elect & Commun Engn Dept Mansoura Egypt Delta Univ Sci & Technol Fac Artificial Intelligence Mansoura Egypt

Numerous medical studies have shown that Alzheimer's disease (AD) was present decades before the clinical diagnosis of dementia. As a result of the development of these studies with the discovery of many ideal biomarkers of symptoms of Alzheimer's disease, it became clear that early diagnosis requires a high-performance computational tool to handle such large amounts of data, as early diagnosis of Alzheimer's disease provides us with a healthy opportunity to benefit from treatment. The main objective of this paper is to establish a complete framework that is based on deep learning approaches and convolutional neural networks (CNN). Four stages of AD, such as (I) preprocessing and data preparation, (II) data augmentation, (III) cross-validation, and (IV) classification and feature extraction based on deep learning for medical image classification, are implemented. In these stages, two methods are implemented. The first method uses a simple CNN architecture. In the second method, the VGG16 model is the pre-trained model that is trained on the imageNet dataset but applies the same model to the different datasets. We apply transfer learning, meaning, and fine-tuning to take advantage of the pre-trained models. Seven performance metrics are used to evaluate and compare the two methods. Compared to the most recent effort, the proposed method is proficient of analyzing AD, moreover, entails less labeled training samples and minimal domain prior knowledge. A significant performance gain on classification of all diagnosis groups was achieved in our experiments. The experimental findings demonstrate that the suggested designs are appropriate for basic structures with minimal computational complexity, overfitting, memory consumption, and temporal regulation. Besides, they achieve a promising accuracy, 99.95% and 99.99% for the proposed CNN model in the classification of the AD stage. The VGG16 pre-trained model is fine-tuned and achieved an accuracy of 97.44% for AD stage clas

关键词： Alzheimer's Disease (AD) Convolution neural Network (CNN) Deep Learning (DL) Transfer Learning (TL) Imaging Pre-processing

来源：评论

学校读者我要写书评

暂无评论

Gesture Analysis Using image processing: For Detection of Suspicious Human Actions 3rd

Gesture Analysis Using Image Processing: For Detection of Su...

引用

3rd Congress on Intelligent Systems (CIS)

作者： Bhagat, Prachi Bhalchandra, Anjali S. Govt Coll Engn Dept Elect Engn Yavatmal India Govt Coll Engn Dept Elect Engn Aurangabad Maharashtra India

ISBN: (纸本)9789811992247;9789811992254

In recent years, interest in real-time video sequences for human action recognition has risen rapidly, and discrimination is required to identify which type of human action appears in the videos. The use of video for human action recognition is becoming widespread in many applications like monitoring of traffic, pedestrian detection, anomalous behavior identification of the human being, health monitoring, human-computer interaction, robotics, etc. Various combinations of feature representation are used to implement different feature detectors for the identification of features and to classify human actions. For testing the given data, Machine learning-based techniques such as unsupervised and supervised learning are utilized to estimate the task of identifying human behaviors. For classification, unsupervised learning techniques like data clustering, hierarchical learning, etc., are implemented, whereas to handle action classification, supervised learning techniques like SVM, random forest, KNN, Naive-Bayes, and artificial neural networks are used. In this paper, real-time applications for suspicious action detection, tracking, and recognition are surveyed and classified. We propose a new technology like MediaPipe Holistic which provides pose, face, and hand mark detectionmodels where the suspicious movement of the human being is being detected by the system from a video that may be stored in memory or from real-time monitoring.

关键词： Feature detectors Machine learning Deep learning SVM CNN Kalman filtering MediaPipe

来源：评论

学校读者我要写书评

暂无评论

Continuous Action Learning Automata Game Optimizer Applied to Convolutional neural networks

Continuous Action Learning Automata Game Optimizer Applied t...

引用

International Conference on Communications, Signal processing, and their applications (ICCSPA)

作者： James Lindsay Sidney Givigi Department of Computer and Electrical Engineering Royal Military College of Canada Kingston Canada School of Computing Queen's University Kingston Canada

ISBN: (数字)9798350384819

ISBN: (纸本)9798350384826

This paper explores the application of the Continuous Action Learning Automata (CALA) game optimizer to Convolutional neural networks (CNNs) for image classification tasks. The CALA game optimizer, initially developed for training artificial neural networks (ANNs), offers a non-gradient descent-based optimization approach that can adapt to different network architectures and activation functions. Leveraging the versatility of the CALA game optimizer, we investigate its performance on CNNs, specifically targeting image recognition within the MNIST dataset. The paper discusses the rationale behind using the CALA game optimizer for CNNs, including its ability to accommodate various activation functions and deeper network architectures. Experimental results demonstrate the efficacy of CALA in training CNNs, showcasing its flexibility and effectiveness in optimizing network parameters for image classification tasks.

关键词： Training image recognition Learning automata Games Network architecture Signal processing Convolutional neural networks Optimization image classification

来源：评论

学校读者我要写书评

暂无评论

Novel image processing strategies for probing photoelectron dichroism effects in chiral molecules

Novel image processing strategies for probing photoelectron ...

引用

作者： Sparling, Chris Heriot-Watt University

学位级别：博士

The phenomenon of photoelectron circular dichroism (PECD) has revolutionised the study of chiral molecules in the gas-phase. PECD presents itself as an asymmetry in the photoelectron angular distribution (PAD) of chiral molecules upon photoionization with circularly polarized light, with measurable dichroism signals typically of order ±10%. To study and extract the subtle features (including those due to PECD effects) present in PADs, advanced experimental techniques, such as velocity map imaging (VMI), and complementary sophisticated data-processing techniques are required. The use of elliptically polarized laser pulses, or mixing together linear and circular polarizations for creating the PADs further complicates the analysis of experimental data, and often requires the use of comparatively time-consuming tomographic imaging techniques. This thesis investigates and develops a number of novel image processing approaches for tackling some of the challenges common to VMI experiments. This includes exploring applications of machine learning and artificial neural networks for solving ill-posed problems that frequently arise during the analysis of experimental data. New tomographic reconstruction approaches are also investigated for the first time within the context of VMI experiments. It is found that the Hankel Transform Reconstruction (HTR) approach, in particular, has significant advantages over other reconstruction options, and allows for PADs to be determined tomographically with the minimum possible volume of imaging data. These novel image processing strategies are tested by accurately reconstructing the full three-dimensional PADs in a study of the photoelectron elliptical dichroism (PEELD) of (R)-camphor following three photon ionization with elliptically polarized 400 nm light. This measurement revealed clear features in the PADs of chiral molecules that have previously been thought to be negligible in other related experiments. Finally, the first experimenta

关键词：

来源：评论

学校读者我要写书评

暂无评论

Analysis of Video Quality Datasets via Design of Minimalistic Video Quality Models

引用

IEEE TRANSACTIONS ON PATTERN ANALYSIS AND MACHINE INTELLIGENCE 2024年第11期46卷 7056-7071页

作者： Sun, Wei Wen, Wen Min, Xiongkuo Lan, Long Zhai, Guangtao Ma, Kede Shanghai Jiao Tong Univ Inst Image Commun & Informat Proc Shanghai 200240 Peoples R China City Univ Hong Kong Dept Comp Sci Kowloon Hong Kong Peoples R China Natl Univ Def Technol Inst Quantum Informat Coll Comp Sci & Technol Changsha 410073 Peoples R China Natl Univ Def Technol Coll Comp Sci & Technol State Key Lab High Performance Comp Changsha 410073 Peoples R China Shanghai Jiao Tong Univ AI Inst MoE Key Lab Artificial Intelligence Shanghai 200240 Peoples R China City Univ Hong Kong Shenzhen Res Inst Dept Comp Sci Kowloon Hong Kong Peoples R China

Blind video quality assessment (BVQA) plays an indispensable role in monitoring and improving the end-users' viewing experience in various real-world video-enabled media applications. As an experimental field, the improvements of BVQA models have been measured primarily on a few human-rated VQA datasets. Thus, it is crucial to gain a better understanding of existing VQA datasets in order to properly evaluate the current progress in BVQA. Towards this goal, we conduct a first-of-its-kind computational analysis of VQA datasets via designing minimalistic BVQA models. By minimalistic, we restrict our family of BVQA models to build only upon basic blocks: a video preprocessor (for aggressive spatiotemporal downsampling), a spatial quality analyzer, an optional temporal quality analyzer, and a quality regressor, all with the simplest possible instantiations. By comparing the quality prediction performance of different model variants on eight VQA datasets with realistic distortions, we find that nearly all datasets suffer from the easy dataset problem of varying severity, some of which even admit blind image quality assessment (BIQA) solutions. We additionally justify our claims by comparing our model generalization capabilities on these VQA datasets, and by ablating a dizzying set of BVQA design choices related to the basic building blocks. Our results cast doubt on the current progress in BVQA, and meanwhile shed light on good practices of constructing next-generation VQA datasets and models.

关键词： Streaming media Computational modeling Distortion Video recording Quality assessment Visualization Crowdsourcing Blind video quality assessment datasets deep neural networks video processing

来源：评论

学校读者我要写书评

暂无评论

A novel offline handwritten text recognition technique to convert ruled-line text into digital text through deep neural networks

引用

MULTIMEDIA TOOLS AND applications 2022年第13期81卷 18223-18249页

作者： Qureshi, Faiza Rajput, Asif Mujtaba, Ghulam Fatima, Noureen Sukkur IBA Univ Ctr Excellence Robot Artificial Intelligence & Bl Dept Comp Sci Sukkur Pakistan

Offline Handwritten Text Recognition (HTR) has been an active area of research due to its wide range of applications and challenges. Recently, many offline HTR techniques have been developed. However, most of the existing techniques were trained on the datasets that contain the handwritten text images on plain pages. Nevertheless, in real life, the handwritten text can be written on either plain pages or ruled-line pages. Therefore, the approaches proposed in recent literature are unable to convert the digital text accurately written on ruled-line pages. Hence, this study proposes a tailor-made end-to-end offline HTR technique that can accurately convert the offline handwritten text written on ruled-line pages into digital text with the help of computer vision and deep neural network-based techniques. To Evaluate the performance of our proposed technique, we developed a relatively complex dataset that contains the hand-written text images on the ruled-line pages. Our experimental results show that our proposed technique is capable of converting the hand-written text on ruled-line pages into digital text with an overall accuracy of 76.7%. Moreover, the experimental results show that our proposed technique obtained 20% more accurate results compared to baseline techniques. We believe that our proposed technique will contribute positively in the body of knowledge in the field of offline HTR. Moreover, the modular design of our proposed technique allows tailored modifications with respect to data while eliminating the need to retrain the neural network-based models.

关键词： Offline hand-written text recognition Ruled-line handwritten text recognition Ruled-lines Deep learning Machine learning Digital image processing

来源：评论

学校读者我要写书评

暂无评论

Automatic damage detection of bridge joints and road pavements by artificial neural networks (ANNs) 13

Automatic damage detection of bridge joints and road pavemen...

引用

Conference on Earth Resources and Environmental Remote Sensing/GIS applications XIII Part of SPIE Remote Sensing Conference

作者： Gagliardi, V Bella, F. Sansonetti, G. Previti, R. Menghini, L. Roma Tre Univ Dept Engn Via Vito Volterra 62 I-00146 Rome Italy

ISBN: (纸本)9781510655409;9781510655393

Monitoring the actual conditions of critical infrastructure assets is a priority for administrators to guarantee high-standard in terms of structural stability, and operational safety and to prevent damages and deterioration. Nowadays, most protocols for assessing road pavement conditions, including viaducts and bridges, are based on visual inspections conducted by specialized operators and, rarely, the local application of ground-based technologies such as IoT Wireless Sensors, laser scanners, Falling Weight Deflectometers, and accelerometers. However, the high costs of maintenance operations and on-site surveys, which involves the partial or total temporary closure of the infrastructure for the duration of the tests, still limit the implementation of these procedures at the network-scale level. Accordingly, the definition of innovative, automatic, and low-cost procedures for continuous monitoring operations, especially for road pavements and concrete bridges, is still an open challenge. In this context, recent developments in artificial neural networks (ANNs) have opened new possibilities in the automatic recognition and geo-localization of damages affecting critical transportation assets, such as bridges and viaducts. This research presents an experimental application for the automatic detection of deteriorated bridge-deck expansion joints and the classification of pavement distresses based on artificial neural networks models. More specifically, the methodology is founded on the latest generation of ANNs algorithms, among which "YOLO v5". For this purpose, an experimental evaluation is conducted by the acquisition of a large dataset of imageries of road damages and bridge-deck joints, collected by open-source datasets, and the processing by ANNs. Furthermore, a field experiment was conducted in order to collect several videos of bridges, selected as case-studies, by means of an instrumented vehicle, through a High-Definition resolution camera. The acquired image

关键词： road pavement evaluation non-destructive analysis machine learning health-condition assessment pavement inspection bridge joint monitoring transport infrastructures roads inspection deep neural networks

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：