检索结果-内蒙古大学图书馆

International conference on Perspective Technologies and Methods in MEMS Design (MEMSTECH)

作者： Yaroslav Sokolovskyy Denys Manokhin Olha Mokrytska Department of Computer Design Systems Lviv Polytechnic National University Lviv Ukraine Department of Information Systems Ivan Franko National University of Lviv Lviv Ukraine

ISBN: (数字)9798350378627

ISBN: (纸本)9798350378634

This paper explores the task of automatic intracranial hemorrhage (ICH) segmentation based on computer tomography (CT) data, with potential applications in biomedical engineering. The main focus is, on improving segmentation accuracy by incorporating texture enhancement techniques based on fractional order derivatives. The study looks at the segmentation of ICH using $\mathrm{U}-\mathrm{Net}$, a deep learning model that is widely used in the field of the segmentation of medical images. The training process employs a parallel algorithm using CUDA technology. Afterwards, an investigation is conducted into a texture enhancement technique that relies on Riesz fractional order derivatives. The goal is to capture intricate details and subtle textures with the potential to enhance segmentation accuracy. To assess the impact of this preprocessing method on the automatic ICH segmentation problem, the U-Net model undergoes retraining and validation. The texture-enhanced images are analyzed to interpret the obtained results. The findings indicate a subtle yet discernible enhancement in accuracy, as gauged by the Jaccard and Dice coefficients. This emphasizes the auspicious potential of the explored texture enhancement method in improving intracranial hemorrhage segmentation within the realm of biomedical engineering.

关键词： Deep learning Training image segmentation Accuracy Graphics processing units artificial neural networks Hemorrhaging

来源：评论

学校读者我要写书评

暂无评论

Inducing Inductive Bias in Vision Transformer for EEG Classification

Inducing Inductive Bias in Vision Transformer for EEG Classi...

引用

International conference on Acoustics, Speech, and Signal processing (ICASSP)

作者： Rabindra Khadka Pedro G. Lind Gustavo Mello Michael A. Riegler Anis Yazidi Department of Computer Science Oslo Metropolitan University Oslo Norway Nordic Centre for Sustainable and Trustworthy Artificial Intelligence Research Oslo Norway Simula Metropolitan Center for Digital Engineering Oslo Norway

Human brain signals are highly complex and dynamic in nature. Electroencephalogram (EEG) devices capture some of this complexity, both in space and in time, with a certain resolution. Recently, transformer-based models have been explored in various applications with different modalities of data. In this work, we introduce a transformer-based model for the classification of EEG signals, inspired by the recent success of the Vision Transformer (ViT) in image classification. Driven by the distinctive characteristics of the EEG data, we design a module that enables us to (1) extract spatio-temporal tokens inherent in the EEG signals and (2) integrate additional non-linearities to capture intricate and non-linear patterns in EEG signals. To that end, we introduce a new lightweight architectural component that combines our proposed attention model with convolution. This convolutional tokenization module forms the basis of our vision backbone referred to as Brain Signal Vision Transformer (BSVT). This architecture takes into account the spatial and temporal features in EEG datasets, leading to token embeddings that effectively capture the fusion of spatial-temporal information. Moreover, while transformer-based models typically perform well when provided with large datasets, here we show that our combination of the inherent inductive bias of Convolutional neural networks (CNN) with the transformer enables efficient training from scratch using relatively small datasets, with as few as 0.75M parameters. On the publicly available EEG dataset from Temple University Hospital (TUH Abnormal), our model achieves results comparable or superior to its counterpart ViT model with patchify stem. The implementation is available at https://***/IamRabin/***

关键词：

来源：评论

学校读者我要写书评

暂无评论

GNet: 3D Object Detection from Point Cloud with Geometry-Aware Network

GNet: 3D Object Detection from Point Cloud with Geometry-Awa...

引用

IEEE International conference on Cyborg and Bionic Systems (CBS)

作者： Hao Ruan Bingrong Xu Junbin Gao Lianguang Liu Junting Lv Yin Sheng Zhigang Zeng School of Artificial Intelligence and Automation Huazhong University of Science and Technology Wuhan China Key Laboratory of Image Processing and Intelligent Control of Education Ministry of China Wuhan China School of Automation Wuhan University of Technology Wuhan China

3D object detection based on point clouds in real-world applications has been a challenging task because the raw point cloud may be sparse or occluded. It may be caused by bad weather or limited sensor resolution, leading to decreasing detection performance. To this end, a novel highperformance Geometry-aware Network (GNet) is proposed and used to accurately detect 3D objects from sparse point clouds. Different from previous work that only regresses the offset of 3D bounding boxes numerically between ground truth (GT) and prediction, GNet considers the geometric information of 3D bounding boxes and combines the two effectively to achieve a better regression performance. In particular, Point Nets blocks learned voxel-wise features and the proposals are generated by the 3D voxel convolution neural network (voxel CNN). By utilizing the prior knowledge and the geometric properties of the 3D bounding boxes, the proposed novel OS-loss can improve detection performance and accelerate speed of inference. Furthermore, GNet takes advantage of prior geometry properties rather than only regressing the offset numerically. These speed up the convergence of the regression network. Finally, our proposed GNet was evaluated on the KITTI dataset for 3D objects and Bird’s Eye View (BEV) detection aspects. The results demonstrate that GNet can improve object detection performance for sparse point clouds.

关键词： Point cloud compression Location awareness Geometry Three-dimensional displays neural networks Object detection Proposals

来源：评论

学校读者我要写书评

暂无评论

PC-BNA: Parallel Convolution Binary neural Network Accelerator

PC-BNA: Parallel Convolution Binary Neural Network Accelerat...

引用

IEEE International conference on High Performance Computing and Communications (HPCC)

作者： Haihui Liu Yuang Zhang Xiangwei Zheng Mengxin Yu Baoxu An Xiaojie Liu Shandong Provincial Key Laboratory for Distributed Computer Software Novel Technology School of Information Science and Engineering Shandong Normal University Jinan China

Binary neural network (BNN) is widely used in speech recognition, image processing and other fields to save memory and speed up computing. However, the accuracy of the existing binarization scheme in the realistic dataset is obviously low, and the input layer uses 32-bit floating point to avoid excessive precision loss, which requires additional computing units and increases the computational burden. Therefore, it is very important to improve the input layer, save computing resources and reduce precision loss. In this paper, we propose a parallel convolution binary neural network accelerator architecture (PC-BNA). Based on our proposed BNN model, we design an efficient field programmable gate array (FPGA) based accelerator. The input of the first layer is binarized, the traditional binary convolution layer is replaced by a parallel binary convolution, and the network building blocks are improved. The experimental results show that the proposed PC-BNA has higher accuracy and better performance on CIFAR-10. The image recognition accuracy is as high as 91.4%, which is superior to the state-of-art of the BNN accelerator. Compared to the state-of-the-art BNN model, when using the same model size, the use of look-up-table (LUT) saves 9.08%, and the digital signal processor (DSP) saves 27.7%. The results suggest that it is promising for PC-BNA in future high performance mobile artificial intelligence applications.

关键词： Performance evaluation Convolution neural networks Digital signal processors Speech recognition Logic gates Mobile handsets

来源：评论

学校读者我要写书评

暂无评论

Empirical Evaluation of artificial Intelligence (AI) Powered Brain Stroke Detection Methodology by Using Learning Based image processing Techniques

Empirical Evaluation of Artificial Intelligence (AI) Powered...

引用

Innovative Computing, Intelligent Communication and Smart Electrical Systems (ICSES), International conference on

作者： S. Agnes Shifani M. Deepa Suresh Franklin ThamilSelvi M. S M. Paramaiyappan Vivek Korde H Attar Department of ECE Saveetha School of Engineering Saveetha Institute of Medical and Technical Sciences (SIMATS) Saveetha University Chennai Division of Mathematics School of Sciences SRM Institute of Science and Technology Trichy Department of Mathematics Sathyabama Institute of Science and Technology Chennai Department of ECE J.P. College of Engineering Tenkasi Division of Mechanical Engineering Yeshwantrao Chavan College of Engineering Nagpur Department of Energy Engineer Zarqa University zarqa Jordan College of Engineering University of Business and Technology Jeddah Saudi Arabia

ISBN: (数字)9798331543624

ISBN: (纸本)9798331543631

Stroke is a leading cause of death and disability worldwide, making early detection and accurate diagnosis critical for patient outcomes. This study presents a Hybrid Learning Assisted Tumor Detection Scheme (HLTDS) for stroke detection from brain CT images. The proposed model integrates Convolutional neural networks (CNNs) with advanced feature extraction techniques, including AutoEncoders and edge detection, to enhance accuracy and computational efficiency. HLTDS was evaluated against nine existing models: CNN, ResNet50, VGG16, MobileNet, DenseNet, Inception, Xception, and EfficientNet, based on key performance metrics such as accuracy, precision, recall, and F1-score. The results demonstrate that HLTDS outperforms all compared models with an accuracy of 96.45%, achieving the highest precision (95.82%) and recall (94.67 % ), making it the most reliable model for identifying stroke-related anomalies. The model's efficiency in terms of training time (58 minutes) and inference time (145 milliseconds) positions it as an optimal choice for real-time clinical applications. Furthermore, HLTDS achieves a balance between model size and memory usage, making it suitable for deployment in resource-constrained environments. These results highlight the importance of hybrid learning approaches in improving stroke detection while ensuring scalability. Future work will focus on expanding the model's capability to handle multiclass classification and further reducing computational costs.

关键词： Training Accuracy Computational modeling Computed tomography Autoencoders Stroke (medical condition) Brain modeling Computational efficiency Convolutional neural networks Hybrid learning

来源：评论

学校读者我要写书评

暂无评论

A contactless automated dynamic monitoring method for organoid morphology on the time axis

A contactless automated dynamic monitoring method for organo...

引用

IEEE International conference on Data Mining Workshops (ICDM Workshops)

作者： Xun Deng Lun Hu Zekun Jiang Lu Liu Peng-Wei Hu The Xinjiang Technical Institute of Physics and Chemistry Chinese Academy of Sciences Urumqi China University of Chinese Academy of Sciences Beijing China Xinjiang Laboratory of Minority Speech and Language Information Processing Urumqi China Shanghai Artificial Intelligence Research Institute Shanghai China Shanghai Roche Pharmaceuticals Ltd Shanghai China

Organoids, as in vitro 3D cultures with structures and functional characteristics similar to corresponding tissues, offer crucial biotechnological support for physiology, pathology, and drug testing. In the field of organoid imaging, the primary approach involves observing the dynamic changes in organoid morphology and movement during cultivation. Currently, there are limited and somewhat restrictive methods for organoid image analysis. Common techniques, such as those based on fluorescence staining, can impact organoid viability. Methods relying on bounding-box detection and tracking fail to capture organoid morphological characteristics, while threshold-based or deep neural network methods often suffer from segmentation inaccuracies. this paper presents the MacrOrga model, which is based on a combination of a multi-axis attention mechanism and convolutional residual networks. MacrOrga represents an innovative approach to organoid segmentation. MacrOrga effectively captures long-distance pixel relationships within organoid images, facilitating precise organoid segmentation. It demonstrates robust generalization capabilities across various datasets. Based on the results of the exact segmentation, the tracking of individual organoids is achieved using the Kuhn– Munkres algorithm to minimize the cost matrix of organoid assignment. The combination of precise segmentation and tracking functions offers a contactless, automated monitoring method for organoids in drug experiments, physiological assessments, and other research applications.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Employing Conversational Agents in Palliative Care: A Feasibility Study and Preliminary Assessment

Employing Conversational Agents in Palliative Care: A Feasib...

引用

IEEE Symposium on Bioinformatics and Bioengineering (BIBE)

作者： Maria Chatzimina Lefteris Koumakis Kostas Marias Manolis Tsiknakis Institute of Computer Science Foundation for Research and Technology–Hellas (FORTH) Heraklion Crete Greece Dept. of Electrical and Computer Engineering Hellenic Mediterranean University & Institute of Computer Science Foundation for Research and Technology–Hellas (FORTH) Heraklion Crete Greece

ISBN: (纸本)9781728146188

Recording of patient-reported outcomes (PROs) enables direct measurement of the experiences of patients with chronic conditions, including cancer; thus, PROs are a critical element of high quality, person-centered care for cancer patients. A growing body of literature reports on the feasibility of using electronic tools for the collection of Patient reported Outcomes (ePROs), although the usability of available solutions does affect their acceptance and use. In parallel, recent advancement in artificial intelligence, machine learning and speech recognition have led to the growing interest in conversational agents, i.e. software applications that mimic written or spoken human speech. In the present manuscript we provide a review of current developments regarding the implementation of conversational agents and their application in the domain of palliative care for oncology patients and also present (i) a methodology for the implementation of a conversational agent able to collect ePRO health data and (ii) initial evaluation results from a relevant feasibility study. Our approach differs from other available systems since the conversational agent reported in the present work is not based on rules, but rather uses machine learning algorithms and more specifically recurrent neural networks (RNN) for identifying appropriate answers. Evaluation results of user experience provided promising results and highlight that users gave positive responds when interacting with the system. Based on the User Experience Questionnaire, pragmatic quality and overall quality were categorized as excellent and hedonic quality was categorized as good. The result of this research can be used as reference for the future development and improvement of the conversational agents in the healthcare domain.

关键词： Cancer Natural language processing Tools Computer science artificial intelligence

来源：评论

学校读者我要写书评

暂无评论

Deep Learning Segmentation Models Evaluation for Deforestation Monitoring Embedded Systems

Deep Learning Segmentation Models Evaluation for Deforestati...

引用

Intelligent Cybernetics Technology & applications (ICICyTA), International conference on

作者： Álvaro S. Careli Evandro C. Vilas Boas Eduardo H. Teixeira Elaine C. C. Silva Guilherme P. Aquino Felipe A. P. Figueiredo Cyber Security and IoT Laboratory CS&I Lab. National Institute of Telecommunication Inatel Santa Rita do Sapucaí Brazil IoT Research Group Laboratory National Institute of Telecommunication Inatel Santa Rita do Sapucaí Brazil Wireless and Artificial Intelligence Laboratory WAI Lab. National Institute of Telecommunication Inatel Santa Rita do Sapucaí Brazil

ISBN: (数字)9798331506490

ISBN: (纸本)9798331506506

This work evaluates deep learning segmentation models to propose a deforestation monitoring embedded system. The approach stands for environmental monitoring using remote sensing imagery, edge computing, and a deep learning segmentation model. Thus, the performance of you only look once architecture version 8 (YOLOv8) and Mask Region-based convolutional neural networks (Mask R-CNN) embedded in Raspberry Pi Model 4 regarding Intersection over Union (IoU), mean Average Precision (mAP), and time per image processing metrics is compared. The models are combined with a pixel-based algorithm that analyzes the temporal segmented images to define their forest area percentage for deforestation monitoring and detection. The results demonstrate YOLOv8x model achieved an IoU of 0.762, with a time per image of 0.4777 seconds, while Mask R-CNN R101 FPN 3x obtained an IoU of 0.763, with a time per image of 0.2669 seconds. The average times for YOLOv8 ranged from 0.0434 to 0.4777 seconds, and for Mask R-CNN from 0.1969 to 0.2669 seconds. Finally, this work proposes evaluating the model's performance when working with generative AI models Dall-e, Craiyon, and Tess-AI to create a synthetic dataset to augment the initial one with synthetic samples and improve the model's training with a large dataset. The Dall-e has been shown to outperform the others regarding the IoU metric, which was suggested to augment datasets with synthetic samples.

关键词： Deforestation Measurement Deep learning Training image segmentation Embedded systems Generative AI Computational modeling Environmental monitoring Synthetic data

来源：评论

学校读者我要写书评

暂无评论

Design-Space Exploration of Systolic Array for Edge Inferencing applications 23

Design-Space Exploration of Systolic Array for Edge Inferenc...

引用

Proceedings of the Third International conference on AI-ML Systems

作者： Prashanth H C Yash Dharmesh Mogal Madhav Rao International Institute of Information Technology Bangalore IN International Institute of information Technology Bangalore IN

ISBN: (纸本)9798400716492

Systolic array (SA) architectures are suitable hardware accelerators to run artificial intelligence (AI) algorithms due to their scalable and pipelined structure. Inferencing on edge devices enables decision-making capability in a small, power-efficient form factor suitable for many IoT-based applications. Convolutional neural networks (CNNs), which form a major part of AI algorithms, are generally memory and compute-intensive algorithms. Optimizing resource-constrained SA-based edge-computing system architecture is crucial. The primary challenge is to find an optimal SA configuration with sufficient on-chip and off-chip memory, processing elements, and optimal memory-bandwidth interface for achieving minimum latency and stall cycles while maximizing compute utilization and mapping efficiency. This work characterizes the popular ResNet50 CNN model on different SA architecture configurations that are structured using a realistic designer tool SCALE-SIM. The design-space exploration of SA was carried out to run the Resnet50 model by configuring i) array sizes (processing elements matrix), ii) SRAM memory sizes available to the computing elements, iii) varying interface bandwidth of memory to the computing elements. The compute latency in clock cycles, SA utilization, Mapping efficiency, number of Stall cycles, and SRAM and DRAM bandwidth requirements were extracted using cycle accurate runs. All three types of data-flow models, including output stationary (OS), weight stationary (WS), and input stationary (IS), were investigated in this work. Unlike previous research that relied on benchmarking various mathematical models, this study focuses on characterizing SA configurations using the ResNet50 model. It presents comprehensive results for all the investigated configurations and establishes the most efficient SA architecture for running CNN architectures. Moreover, it provides valuable insights for designing resource-constrained edge-inferencing systems.

关键词： AI Accelerators CNN Design Space Exploration Systolic Array

来源：评论

学校读者我要写书评

暂无评论

U-Infuse: Democratization of Customizable Deep Learning for Object Detection

引用

SENSORS 2021年第8期21卷 2611页

作者： Shepley, Andrew Falzon, Greg Lawson, Christopher Meek, Paul Kwan, Paul Univ New England Sch Sci & Technol Armidale NSW 2350 Australia Flinders Univ S Australia Coll Sci & Engn Adelaide SA 5001 Australia NSW Dept Primary Ind Vertebrate Pest Res Unit POB 530 Coffs Harbour NSW 2450 Australia Univ New England Sch Environm & Rural Sci Armidale NSW 2350 Australia Melbourne Inst Technol Sch IT & Engn Melbourne Vic 3000 Australia

image data is one of the primary sources of ecological data used in biodiversity conservation and management worldwide. However, classifying and interpreting large numbers of images is time and resource expensive, particularly in the context of camera trapping. Deep learning models have been used to achieve this task but are often not suited to specific applications due to their inability to generalise to new environments and inconsistent performance. Models need to be developed for specific species cohorts and environments, but the technical skills required to achieve this are a key barrier to the accessibility of this technology to ecologists. Thus, there is a strong need to democratize access to deep learning technologies by providing an easy-to-use software application allowing non-technical users to train custom object detectors. U-Infuse addresses this issue by providing ecologists with the ability to train customised models using publicly available images and/or their own images without specific technical expertise. Auto-annotation and annotation editing functionalities minimize the constraints of manually annotating and pre-processing large numbers of images. U-Infuse is a free and open-source software solution that supports both multiclass and single class training and object detection, allowing ecologists to access deep learning technologies usually only available to computer scientists, on their own device, customised for their application, without sharing intellectual property or sensitive data. It provides ecological practitioners with the ability to (i) easily achieve object detection within a user-friendly GUI, generating a species distribution report, and other useful statistics, (ii) custom train deep learning models using publicly available and custom training data, (iii) achieve supervised auto-annotation of images for further training, with the benefit of editing annotations to ensure quality datasets. Broad adoption of U-Infuse by ecological pra

关键词： animal identification artificial intelligence camera-trap images camera trapping deep convolutional neural networks deep learning environmental software wildlife ecology wildlife monitoring ecological object detection

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：