检索结果-内蒙古大学图书馆

Lightweight Low-Power u-net architecture for Semantic Segmentation

CIRCuITS SYSTEMS AND SIGNAL PROCESSING 2025年第4期44卷 2527-2561页

作者： Modiboyina, Chaitanya Chakrabarti, Indrajit Ghosh, Soumya Kanti Indian Inst Technol Dept Elect & Elect Commun Engn Kharagpur 721302 West Bengal India Indian Inst Technol Dept Comp Sci & Engn Kharagpur 721302 West Bengal India

The u-net is a popular deep-learning model for semantic segmentation tasks. This paper describes an implementation of the u-net architecture on FPGA (Field Programmable Gate Array) for real-time image segmentation. The proposed design uses a parallel-pipelined architecture to achieve high throughput and also focuses on addressing the resource and power constraints in edge devices by compressing CNN (Convolutional Neural networks) models and improving hardware efficiency. To this end, we propose a pruning technique based on parallel quantization that reduces weight storage requirements by quantizing u-net layers into a few segments, which in turn leads to the light weight of the u-net model. The system requires approximate to 1.5Mb\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\approx 1.5Mb$$\end{document} of memory for storing weights. The Electron Microscopy Dataset and BraTs Dataset has demonstrated the proposed u-net architecture, achieving an Intersection over union (Iou) of 90.31% and 94.1% when utilizing 4-bit quantized weights. Additionally, we designed a shift-based u-net accelerator that replaces multiplications with simple shift operations, further improving efficiency. The proposed u-net architecture achieves a 3.5 x\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\times$$\end{document} reduction in power consumption and a 35% reduction in area compared to previous architectures. To further reduce power consumption, we omit the computation for zero weights. Overall, the present work puts forward an effective method for optimizing CNN models in edge devices while meeting their computational and powe

关键词： CNN Quantization Pruning FPGA implementation u-net architecture Semantic segmentation

来源：评论

学校读者我要写书评

暂无评论

VMG: Rethinking u-net architecture for Video Super-Resolution

引用

IEEE TRANSACTIONS ON BROADCASTING 2025年第1期71卷 334-349页

作者： Tang, Jun Niu, Lele Liu, Linlin Dai, Hang Ding, Yong Zhejiang Univ Coll Integrated Circuits Hangzhou 310000 Peoples R China Univ Glasgow Sch Comp Sci Glasgow G12 8QQ Scotland

The u-net architecture has exhibited significant efficacy across various vision tasks, yet its adaptation for Video Super-Resolution (VSR) remains underexplored. While the Video Restoration Transformer (VRT) introduced u-net into the VSR domain, it poses challenges due to intricate design and substantial computational overhead. In this paper, we present VMG, a streamlined framework tailored for VSR. Through empirical analysis, we identify the crucial stages of the u-net architecture contributing to performance enhancement in VSR tasks. Our optimized architecture substantially reduces model parameters and complexity while improving performance. Additionally, we introduce two key modules, namely the Gated MLP-like Mixer (GMM) and the Flow-Guided cross-attention Mixer (FGM), designed to enhance spatial and temporal feature aggregation. GMM dynamically encodes spatial correlations with linear complexity in space and time, and FGM leverages optical flow to capture motion variation and implement sparse attention to efficiently aggregate temporally related information. Extensive experiments demonstrate that VMG achieves nearly 70% reduction in GPu memory usage, 30% fewer parameters, and 10% lower computational complexity (FLOPs) compared to VRT, while yielding highly competitive or superior results across four benchmark datasets. Qualitative assessments reveal VMG's ability to preserve remarkable details and sharp structures in the reconstructed videos.

关键词： Computer architecture Data mining Superresolution Mixers Feature extraction Computational modeling Transformers Logic gates Correlation Decoding Video super-resolution u-net architecture spatial-temporal complexity MLP

来源：评论

学校读者我要写书评

暂无评论

Scalable model for segmenting Cells' Nuclei using the u-net architecture

引用

MuLTIMEDIA TOOLS AND APPLICATIONS 2024年第23期83卷 63655页

作者： Ghnemat, Rawan Almodawar, Abedlrahman Al Saraireh, Jaafer Princess Sumaya Univ Technol Fac Comp Sci Comp Sci Dept Amman Jordan

Medical image segmentation significantly influences medicine for diagnostic purposes where higher precision and accuracy are demanded. One of the most cutting-edge architectures in Deep Learning (DL) designed to perform on a few training biomedical images is the u-Shape network (unet). unfortunately, thousands of training biological images are out of reach in most situations, limiting the number of attainable training sets and, as a result, the ability to achieve better accuracy. Therefore, this research proposes a scalable model where the u-net architecture has been tuned to improve its ability to perform on fewer training samples, mitigating the limited trainable data problem. Several model layers, nodes, hyper-parameters, and normalization methods have been experimented with to tune the original u-net architecture well, considering possible underfitting and/or overfitting scenarios while maintaining reasonable accuracy. The 2018 Data Science Bowl dataset has experimented with several smaller subsets, 10%-100% of the original dataset, which has been trained separately. The model's accuracy ranged between 96.98% and 97.93% for the experimented subsets. More importantly, the reported variance in the accuracy given different training data sizes is minimal (around 1%).

关键词： Image segmentation Deep learning u-net architecture Hyper-parameters Normalization

来源：评论

学校读者我要写书评

暂无评论

Analysis of fractal dimension of segmented blood vessels in fundus images using u-net architecture

引用

INTERNATIONAL JOuRNAL OF BIOMEDICAL ENGINEERING AND TECHNOLOGY 2025年第3期47卷 214-240页

作者： Mariyappan, Saranya Sunitha, K. A. Arjunan, Sridhar P. SRM Inst Sci & Technol Coll Engn & Technol Dept Elect & Instrumentat Engn Kattankulathur 603203 Tamil Nadu India SRM Univ AP Dept Elect & Commun Engn Amaravati 522502 Andhra Prades India

Precise segmentation of retinal blood vessels (RBVs) is pivotal in ophthalmology research, aiding in detecting diverse retinal abnormalities. This study proposes a contrast-limited adaptive histogram equalisation (CLAHE) technique to improve retinal image quality and visibility of microvascular structures. We aimed to determine the complexity of blood vessels using fractal dimensions (FD) and compare different metrics for their effectiveness. We employed the unet architecture to separate blood vessels, and our results on the DRIVE retinal fundus image standard dataset showed an impressive accuracy rate of 97.24%, surpassing traditional filtering methods. Box counting, information, capacity, correlation, and probability dimensions are used in the FD analysis to help us understand the complex and irregular structures of retinal blood vessels. These metrics are valuable for detecting and monitoring retinal diseases in clinical settings. Our comparison with other techniques reveals promising results, particularly in the capacity and information dimensions, with statistical significance (P < 0.05). The potential of fractal dimensions as a screening tool for diabetic retinopathy underscores their importance in epidemiological studies.

关键词： blood vessels fundus image DRIVE dataset filter techniques u-net architecture fractal dimension diabetic retinopathy statistical analysis deep learning retinal vessel augmentation structural complexity

来源：评论

学校读者我要写书评

暂无评论

Brain Tumor Segmentation of Magnetic Resonance Imaging (MRI) Images using Deep Neural network Driven unmodified and Modified u-net architecture

引用

INTERNATIONAL JOuRNAL OF ADVANCED COMPuTER SCIENCE AND APPLICATIONS 2024年第10期15卷 778-786页

作者： Arianti, Nunik Destria Muda, Azah Kamilah Univ Teknikal Malaysia Melaka Fak Teknol Maklumat Dan Komunikasi Melaka Malaysia Nusa Putra Univ Dept Informat Syst Sukabumi Indonesia

Accurately separating healthy tissue from tumorous regions is crucial for effective diagnosis and treatment planning based on magnetic resonance imaging (MRI) data. Current manual detection methods rely heavily on human expertise, so MRI-based segmentation is essential to improving diagnostic accuracy and treatment outcomes. The purpose of this paper is to compare the performance of detecting brain tumors from MRI images through segmentation using an unmodified and modified u-net architecture from deep neural network (DNN) that has been modified by adding batch normalization and dropout on the encoder layer with and without the freeze layer. The study utilizes a public 2D brain tumor dataset containing 3064 T1-weighted contrast-enhanced images of meningioma, glioma, and pituitary tumors. Model performance was evaluated using intersection over union (Iou) and standard metrics such as precision, recall, f1-score, and accuracy across training, validation, and testing stages. Statistical analysis, including ANOVA and Duncan's multiple range test, was conducted to determine the significance of performance differences across the architectures. Results indicate that while the modified architectures show improved stability and convergence, the freeze layer model demonstrated superior Iou and efficiency, making it a promising approach for more accurate and efficient brain tumor segmentation. The comparison of the three methods revealed that the modified unet architecture with a freeze layer significantly reduced training time by 81.72% compared to the unmodified u-net while maintaining similar performance across validation and testing stages. All three methods showed comparable accuracy and consistency, with no significant differences in performance during validation and testing.

关键词： Accuracy brain tumor DNN u-net architecture comparison performance

来源：评论

学校读者我要写书评

暂无评论

u-net architecture for Detecting Glaucoma with Retinal Fundus Images 1

U-Net Architecture for Detecting Glaucoma with Retinal Fundu...

引用

1st International Conference on Advances in Electrical, Electronics and Computational Intelligence, ICAEECI 2023

作者： Nandhini, S. Jagadeesan, J. Anitha, E. Vinayaka Mission's Research Foundation Aarupadai Veedu Institute of Technology Department of Computer Science and Engineering Tamil Nadu India Sathyabama Institute of Science and Technology Department of Computer Science and Engineering Tamil Nadu India Loyola - ICAM College of Engineering and Technology Department of Information Technology Chennai India

ISBN: (纸本)9798350342796

Ophthalmologists use retinal images to determine eye disease, for example glaucoma disease. This diagnosis is made by measuring changes in the cup-to-disc ratio. Eye Cup Segmentation Helps Fossil Ophthalmologists Calculate CDRs of Retinal Background Images. In this article, we propose a machine learning approach to perform segmentation tasks using the u-net architecture. This proposed technique was evaluated using 650 colour retinal fundus images. Next, u-net has 160 opportunities. Additionally, we compared the segmentation results with ground reality images. According to the test results, the eyecup's segmentation performance reaches 98.42% of the membrane, with a loss of 0.15. Fundus images help detect glaucoma. In the past, several experiments have been conducted to detect this disease. This paper proposes an ideal framework aimed at building intelligent solutions that help identify glaucoma using MATLAB software tools. The diagnosis of glaucoma is made by measuring changes in the bowl-to-disc ratio. Eye Cup Segmentation Helps Fossil Ophthalmologists Calculate CDRs of Retinal Background Images. As we propose a deep learning method using u-net architecture to perform segmentation tasks, we collected many fundus images as data. Subsequent pre-processing of the occlusion figure is accompanied by classification and segmentation. The pre-processing module uses several filters such as Gaussian filter and guided image filter to remove noise. The division into healthy and sick eyes helps with further administration. © 2023 IEEE.

关键词： Deep Learning Glaucoma detection KNN Algorithm Machine Learning u-net architecture

来源：评论

学校读者我要写书评

暂无评论

Analysis of depth variation of u-net architecture for brain tumor segmentation

引用

MuLTIMEDIA TOOLS AND APPLICATIONS 2023年第7期82卷 10723-10743页

作者： Jena, Biswajit Jain, Sarthak Nayak, Gopal Krishna Saxena, Sanjay Int Inst Informat Technol Bhubaneswar India

u-net is a fully convolutional network (FCN) architecture designed to research the segmentation of biomedical images. The depth of the u-net is one of the major constraints of this model while computing the performances. The larger depth of the u-net means that its computational complexity is high as well. In certain cases, this large depth, as in the original model, is not justified for biomedical imaging modalities. In this paper, we have done an efficient analysis of u-net architecture's depth variation, i.e., after removing different layers. For the analysis, the datasets BraTS-2017 and BraTS-2019, which consist of High-Grade Glioma (HGG) and Low-Grade Glioma (LGG) MR Scans, have been used for tumor segmentation. We have achieved a dice coefficient of at least 0.8866 and as high as 0.8887 on the discovery cohort, and at least 0.8895 and as high as 0.8911 cross-validation replication cohort. The results show that there are the least significant changes occurring in the performance parameters while moving from the higher to the lower depth of the model. Hence, in this paper, we presented that the large depth of u-net, which costs more in terms of computational complexity, is not always required. Moreover, the u-net models with depth reduction, which decreases the computational complexity, can achieve nearly the same results as in the case of the original u-net.

关键词： u-net architecture Fully convolutional network (FCN) Brain tumor segmentation Biomedical image segmentation Deep learning Convolutional neural network

来源：评论

学校读者我要写书评

暂无评论

Refining u-net architecture Through Genetic Algorithms for Improved Skin Lesion Image Segmentation 4th

Refining U-Net Architecture Through Genetic Algorithms for I...

引用

4th International Conference on Intelligent Systems and Pattern Recognition-ISPR

作者： El Abassi, Fouzia Darouichi, Aziz Ouaarab, Aziz Cadi Ayyad Univ Comp & Syst Engn Lab L2IS FSTG Marrakech 40000 Morocco

ISBN: (纸本)9783031821554;9783031821561

Recently, various approaches using manually designed convolutional neural networks (CNNs) and u-shaped (encoder-decoder) models have yielded encouraging results in the automatic segmentation of medical images. Nevertheless, the manual construction of these models is laborious, requires a high level of expertise, is time-consuming, error-prone and can result in the loss of crucial details. To improve segmentation performance, these models often incorporate numerous parameters, which can lead to problems such as overfitting, vanishing and exploding gradient and increased computational complexity. In this study, we implement a genetic algorithm (GA) specifically designed for a u-net architecture to segment images of skin lesions, using two datasets ISIC-SMALL1 and ISIC-SMALL2 constructed from ISIC-2017 dataset. The main objective is to optimize internal block topologies within the u-net, in order to identify a high-performance deep neural network with a minimum of parameters. Compared to the u-net baseline, our proposed methodology demonstrates favorable outcomes in terms of both the number of trainable parameters and Intersection over union (Iou). For the ISIC-SMALL1 dataset, the leading model surpasses u-net with a reduced parameter count (0.352001M) and an elevated Iou score (0.826052 vs. 0.736352). Similarly, the second-best model outperforms u-net with fewer parameters (0.258781M vs. 31.043586M) and a higher Iou score (0.821956 vs. 0.736352). In the ISIC-SMALL2 dataset, both top models outshine u-net in Iou (0.753122 and 0.722079 vs. 0.725363, respectively) while maintaining lower parameter counts.

关键词： Medical Image Segmentation Convolutional Neural networks (CNNs) u-net architecture Genetic Algorithms (GA) Skin Lesion

来源：评论

学校读者我要写书评

暂无评论

Segmentation of Brain Tumours using Optimised u-net architecture 1

引用

4th Congress on Intelligent Systems (CIS)

作者： Jyothilakshmi, M. Rebecca, P. Preethy Joe, J. Wisely St Peters Coll Engn & Technol Chennai 600054 India

ISBN: (数字)9789819990436

ISBN: (纸本)9789819990429;9789819990436

Recent research has shown that convolutional networks can be much deeper, more accurate, and easier to train if they have dumpier links between layers close to the input and the output. In this paper, we investigate the optimal depth, number of convolutional channels, and post-processing technique for the "brain tumour segmentation" task in the BraTS21 challenge by testing for profound supervision loss, Focal loss, decoder attention, drop block, and residual connections. We tested our suggested architecture with deep supervision, drop block, and Focal loss, together with numerous u-net network variations, including the initially proposed u-net, the unetR, Residual u-net, and Attention u-net, in order to find the optimal training sequence.

关键词： Brain tumours u-net architecture BraTS21 Disease

来源：评论

学校读者我要写书评

暂无评论

Accurate Semantic Segmentation of Aerial Imagery using Attention Res u-net architecture 6

Accurate Semantic Segmentation of Aerial Imagery Using Atten...

引用

6th IEEE International Conference on Emerging Smart Computing and Informatics, ESCI 2024

作者： Gonthina, Nagamani Adunuri, Saikumar Mateti, Rishindra Allampalli, Sai Sudeep Narasimha Prasad, L.V. Institute of Aeronautical Engineering Hyderabad India

ISBN: (纸本)9798350306613

Semantic segmentation of photos holds importance in various fields such, as security, emergency response, agriculture, urban planning and surveillance. Researchers have shown increased interest in automating the analysis of photographs without intervention due to its critical role across multiple disciplines. A CNN variant called u-net has been successful in segmenting photos beyond the domain. However u-nets limited number of layers hampers its ability to extract information from photos accurately and can lead to flawed boundaries for objects with complex features. In this study we introduce an architecture called Attention Res u-net that addresses challenges in semantic segmentation such as low accuracy, for high resolution images and incorrect object boundary recognition. Our approach focuses on refining object boundaries. This approach successfully balances the extraction of attributes with the preservation of space thereby significantly improving aerial image analysis. © 2024 IEEE.

关键词： Aerial imagery Attention u-net architecture mechanisms Semantic segmentation u-net architecture

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：