检索结果-内蒙古大学图书馆

Image detection method for multi-category lesions in wireless capsule endoscopy based on deep learning models

World Journal of Gastroenterology 2024年第48期30卷 5111-5129页

作者： Zhi-Guo Xiao Xian-Qing Chen Dong Zhang Xin-Yuan Li Wen-Xin Dai Wen-Hui Liang School of Computer Science Technology Changchun UniversityChangchun 130022Jilin ProvinceChina School of Computer Science Technology Beijing Institute of TechnologyBeijing 100811China

BACKGROUND Wireless capsule endoscopy(WCE)has become an important noninvasive and portable tool for diagnosing digestive tract diseases and has been propelled by advancements in medical imaging ***,the complexity of the digestive tract structure,and the diversity of lesion types,results in different sites and types of lesions distinctly appearing in the images,posing a challenge for the accurate identification of digestive tract *** To propose a deep learning-based lesion detection model to automatically identify and accurately label digestive tract lesions,thereby improving the diagnostic efficiency of doctors,and creating significant clinical application *** In this paper,we propose a neural network model,WCE_Detection,for the accurate detection and classification of 23 classes of digestive tract lesion ***,since multicategory lesion images exhibit various shapes and scales,a multidetection head strategy is adopted in the object detection network to increase the model's robustness for multiscale lesion ***,a bidirectional feature pyramid network(BiFPN)is introduced,which effectively fuses shallow semantic features by adding skip connections,significantly reducing the detection error *** the basis of the above,we utilize the Swin Transformer with its unique self-attention mechanism and hierarchical structure in conjunction with the BiFPN feature fusion technique to enhance the feature representation of multicategory lesion *** The model constructed in this study achieved an mAP50 of 91.5%for detecting 23 *** than eleven single-category lesions achieved an mAP50 of over 99.4%,and more than twenty lesions had an mAP50 value of over 80%.These results indicate that the model outperforms other state-of-the-art models in the end-to-end integrated detection of human digestive tract lesion *** The deep learning-based object detection network detects multiple digestive tract lesi

关键词： Human digestive tract Artificial intelligence Deep learning Wireless capsule endoscopy Object detection

来源：评论

学校读者我要写书评

暂无评论

Self-Supervised Monocular Depth Estimation via Discrete Strategy and Uncertainty

引用

IEEE/CAA Journal of Automatica Sinica 2022年第7期9卷 1307-1310页

作者： Zhenyu Li Junjun Jiang Xianming Liu School of Computer Science and Technology Harbin Institute of TechnologyHarbin 150001 Peng Cheng Laboratory Shenzhen 518052China

Dear Editor,This letter is concerned with self-supervised monocular depth *** estimate uncertainty simultaneously,we propose a simple yet effective strategy to learn the uncertainty for self-supervised monocular depth estimation with the discrete strategy that explicitly associates the prediction and the uncertainty to train the ***,we propose the uncertainty-guided feature fusion module to fully utilize the uncertainty *** will be available at https://***/zhyever/***-supervised monocular depth estimation methods turn into promising alternative trade-offs in both the training cost and the inference ***,compound losses that couple the depth and the pose lead to a dilemma of uncertainty calculation that is crucial for critical safety *** solve this issue,we propose a simple yet effective strategy to learn the uncertainty for self-supervised monocular depth estimation using the discrete bins that explicitly associate the prediction and the uncertainty to train the *** strategy is more pluggable without any additional changes to self-supervised training losses and improves model ***,to further exert the uncertainty information,we propose the uncertainty-guided feature fusion module to refine the depth estimation.

关键词： estimation utilize uncertainty

来源：评论

学校读者我要写书评

暂无评论

Robust 3D reconstruction and cell-type deconvolution approach for spatial transcriptomics by hybrid graph learning and iterative computation

Robust 3D reconstruction and cell-type deconvolution approac...

引用

2024 IEEE International Conference on Bioinformatics and Biomedicine, BIBM 2024

作者： Zhang, Zhengqian Chen, Binghong Wang, Jialiang Wang, Haiyu Yu, Tianshuo J., Ren M., Zhang C., Liu W., Jiang Y., Liu Harbin Institute of Technology School of Computer Science and Technology Harbin China Harbin Institute of Technology School of Mathematics Harbin China Harbin Institute of Technology Department of Fundamental Education Harbin China University of Glasgow James Watt School of Engineering Glasgow United Kingdom Future Technology Institute Harbin Institute of Technology Harbin China

ISBN: (纸本)9798350386226

Spatial transcriptomics data provides new methods of exploring gene expressions. However, the results usually do not contain cell-level proportions, leaving a gap between the new technologies and the corresponding analysis. To address the issues, we propose a novel framework, OptiGraph3D, in this paper. Our approach starts with a pivoted sample-level sequential registration, eliminating potential cumulative errors in multiple registrations. The registration results provide sufficient spatial prior for our cell-type deconvolving process, which leverages the power of ADMM method by modelling the cell-type deconvolution task as a matrix inverse problem and proposes efficient solvers with guaranteed convergence. Furthermore, the cell-type deconvolution results can in turn perform as supervision for cluster recognition in 3D reconstruction. Our model introduces a semi-supervision training manner by introducing a specific class token in the multi-headed graph attention network, bridging the gap between gene-level expressions of spatial transcriptomics and presenting new angle for sample-wise analysis. Extensive experiments are performed on the human dorsolateral prefrontal cortex (DLPFC) and the developing human embryonic heart dataset, suggesting our approach's superiority over other methods. © 2024 IEEE.

关键词： Gene expression

来源：评论

学校读者我要写书评

暂无评论

Event-based sliding mode control under denial-of-service attacks

引用

science China(Information sciences) 2022年第6期65卷 125-141页

作者： Yingxin TIAN Xiang LI Bo DONG Yabin GAO Ligang WU Department of Control Science and Engineering Harbin Institute of Technology School of Software Harbin Institute of Technology

This paper deals with the problem of estimator-based sliding mode control against denial-of-service(DoS) attacks and discrete events via a time-delay approach. A networked system is considered an uncertain dynamical system with matched and mismatched perturbations and exogenous disturbance in the network environment. A network-resource-aware event-triggering mechanism is designed with aperiodically releasing system measurements. Furthermore, to describe the DoS attack duration and inter-event time, a time-delay modeling approach considers the DOS attack duration and inter-event time as a “time delay” of the measurements between the sensor and controller over the network is proposed. Consequently, an intervaltime-delay system with uncertainties is formulated. A state-observer-based sliding mode controller, by which the ideal sliding mode can be achieved, is proposed against the DoS attacks. The resulting sliding motion is proved to be robust and stable with an Lgain performance. Finally, the effectiveness and applicability of the present sliding mode control are validated in a simulated pendulum system.

关键词： sliding mode control time delay uncertain system state observer event trigger

来源：评论

学校读者我要写书评

暂无评论

A comprehensive survey of image compression methods: from prediction models to advanced techniques

引用

Multimedia Tools and Applications 2025年 1-39页

作者： Devadason, Joshua Rajah D, Aju School of Computer Science and Engineering Vellore Institute of Technology Vellore India

Images are used widely nowadays. Images are used in many fields such as medicine to terrain mapping. There is a need to compress the images and represent them in shorter form for effective transmission. Several techniques are reviewed in this paper. The first group of techniques is the data compression technique. The data compression methods are Huffman encoding, Lempel–Ziv Welch (LZW), arithmetic encoding, run length encoding, and Shannon-Fano encoding. The second group of methods is prediction, here the neighboring pixels are utilized in the prediction of the present pixel and the error is encoded using an entropy encoder. In wavelet-based techniques, the source image is transformed from the spatial domain to the wavelet domain, and the coefficients are encoded. In learning-based techniques, a neural network is used to model the image data for compression. Fractal encoding is a compression method that searches for similar blocks in an image and compresses the difference between the blocks. With an average bitrate of 4.87 bpp (bits per pixel) on the first 10 photos of the Digital Imaging and Communications in Medicine (DICOM) image collection, the results indicate that LZW is the top performing data compression technique among the data compression techniques. In prediction-based techniques, Context Adaptive Lossless Image Codec (CALIC) is the best performing with 3.5 bpp. The modern neural method, compressive autoencoder outperforms established methods such as Joint Photographic Experts Group (JPEG), by achieving a PSNR of 35 on the Kodak dataset compared with 32 PSNR on the same dataset at 1.0 bpp. © The Author(s), under exclusive licence to Springer science+Business Media, LLC, part of Springer Nature 2025.

关键词： Pixels

来源：评论

学校读者我要写书评

暂无评论

CCM-FL:Covert communication mechanisms for federated learning in crowd sensing IoT

引用

Digital Communications and Networks 2024年第3期10卷 597-608页

作者： Hongruo Zhang Yifei Zou Haofei Yin Dongxiao Yu Xiuzhen Cheng Institute of Intelligent Computing School of Computer Science and TechnologyShandong UniversityQingdao266237China

The past decades have witnessed a wide application of federated learning in crowd sensing,to handle the numerous data collected by the sensors and provide the users with precise and customized ***,how to protect the private information of users in federated learning has become an important research *** with the differential privacy(DP)technique and secure multiparty computation(SMC)strategy,the covert communication mechanism in federated learning is more efficient and energy-saving in training the ma-chine learning *** this paper,we study the covert communication problem for federated learning in crowd sensing Internet-of-Things *** from the previous works about covert communication in federated learning,most of which are considered in a centralized framework and experimental-based,we firstly proposes a centralized covert communication mechanism for federated learning among n learning agents,the time complexity of which is O(log n),approximating to the optimal ***,for the federated learning without parameter server,which is a harder case,we show that solving such a problem is NP-hard and prove the existence of a distributed covert communication mechanism with O(log logΔlog n)times,approximating to the optimal solution.Δis the maximum distance between any pair of learning *** analysis and nu-merical simulations are presented to show the performance of our covert communication *** hope that our covert communication work can shed some light on how to protect the privacy of federated learning in crowd sensing from the view of communications.

关键词： Covert communications Federated learning Crowd sensing SINR model

来源：评论

学校读者我要写书评

暂无评论

Skin cancer detection using optimized mask R-CNN and two-fold-deep-learning-classifier framework

引用

Multimedia Tools and Applications 2025年 1-28页

作者： Reddy, Akepati Sankar M.P, Gopinath School of Computer Science and Engineering Vellore Institute of Technology Vellore India

Skin cancer is a serious and potentially life-threatening condition caused by DNA damage in the skin cells, leading to genetic mutations and abnormal cell growth. These mutations can cause the cells to divide and grow uncontrollably, forming a tumor on the skin. To prevent skin cancer from spreading and potentially leading to serious complications, it's critical to identify and treat it as early as possible. An innovative two-fold deep learning based skin cancer detection model is presented in this research work. Five main stages make up the proposed model: Preprocessing, segmentation, feature extraction, feature selection, and skin cancer detection. Initially, the Min–max contrast stretching and median filtering used to pre-process the collected raw image. From the pre-processed image, the Region of Intertest (ROI) is identified via optimized mask Region-based Convolutional Neural Network (R-CNN). Then, from the identified ROI areas, the texture features like Illumination-invariant Binary Gabor Pattern (II-BGP), Local Binary Pattern (LBP), Gray-Level Co-occurrence Matrix (GLCM), Color feature such as Color Correlogram and Histogram Intersection, and Shape feature including Moments, Area, Perimeter, Eccentricity, Average bending energy are extracted. To choose the optimal features from the extracted ones, the Golden Eagle Mutated Leader Optimization (GEMLO) is used. The proposed Golden Eagle Mutated Leader Optimization (GEMLO) is the conceptual amalgamation of the standard Mutated Leader Algorithm (MLA) and Golden Eagle Optimizer are used to select best features (GEO). The skin cancer detection is accomplished via two-fold-deep-learning-classifiers, that includes the Fully Convolutional Neural Networks (FCNs) and Multi-Layer Perception (MLP). The final outcome is the combination of the outcomes acquired from Fully Convolutional Neural Networks (FCNs) and Multi-Layer Perception (MLP). The PYTHON platform is being used to implement the suggested model. Using the curre

关键词： Median filters

来源：评论

学校读者我要写书评

暂无评论

An Energy-Efficient Routing Algorithm for UAV Formation Based on Time-Aggregated Graph

引用

China Communications 2024年第11期21卷 28-39页

作者： Wang Gaifang Li Bo Yang Hongjuan Jiang Xu School of Information Science and Engineering Harbin Institute of TechnologyWeihai 264200China

The limited energy and high mobility of unmanned aerial vehicles(UAVs)lead to drastic topology changes in UAV *** existing routing protocols necessitate a large number of messages for route discovery and maintenance,greatly increasing network delay and control overhead.A energyefficient routing method based on the discrete timeaggregated graph(TAG)theory is proposed since UAV formation is a defined time-varying *** network is characterized using the TAG,which utilizes the prior knowledge in UAV *** energyefficient routing algorithm is designed based on TAG,considering the link delay,relative mobility,and residual energy of *** routing path is determined with global network information before requesting *** results demonstrate that the routing method can improve the end-to-end delay,packet delivery ratio,routing control overhead,and residual ***,introducing timevarying graphs to design routing algorithms is more effective for UAV formation.

关键词： energy-efficient route time-aggregated graph UAV formation

来源：评论

学校读者我要写书评

暂无评论

Stacked collaborative transformer network with contrastive learning for video moment localization

引用

Intelligent Data Analysis 2025年第1期29卷 220-234页

作者： Megala, G. Swarnalatha, P. School of Computer Science and Engineering Vellore Institute of Technology Vellore India

Video grounding intends to perform temporal localization in multimedia information retrieval. The temporal bounds of the target video span are determined for the given input query. A novel interactive multi-head self-attention (IMSA) transformer is proposed to localize an unseen moment in the untrimmed video for the given image. A new semantic-trained self-supervised approach is considered in this paper to perform cross-domain learning to match the image query – video segment. It normalizes the convolution function enabling efficient correlation and collecting of semantically related video segments across time based on the image query. A double hostile Contrastive learning with Gaussian distribution parameters method is advanced to learn the representations of video. The proposed approach performs dynamically on various video components to achieve exact semantic synchronization and localization among queries and video. In the proposed approach, the IMSA model localizes frames greatly compared to other approaches. Experiments on benchmark datasets show that the proposed model can significantly increase temporal grounding accuracy. The moment occurrence is identified in the video with a start and end boundary ascertains an average recall of 86.45% and a mAP of 59.3%. © 2024 – IOS Press. All rights reserved.

关键词： Semantic Segmentation

来源：评论

学校读者我要写书评

暂无评论

Two-Stage Approach for Targeted Knowledge Transfer in Self-Knowledge Distillation

引用

IEEE/CAA Journal of Automatica Sinica 2024年第11期11卷 2270-2283页

作者： Zimo Yin Jian Pu Yijie Zhou Xiangyang Xue the School of Computer Science Fudan University the Institute of Science and Technology for Brain Inspired Intelligence Fudan University IEEE

Knowledge distillation(KD) enhances student network generalization by transferring dark knowledge from a complex teacher network. To optimize computational expenditure and memory utilization, self-knowledge distillation(SKD) extracts dark knowledge from the model itself rather than an external teacher network. However, previous SKD methods performed distillation indiscriminately on full datasets, overlooking the analysis of representative samples. In this work, we present a novel two-stage approach to providing targeted knowledge on specific samples, named two-stage approach self-knowledge distillation(TOAST). We first soften the hard targets using class medoids generated based on logit vectors per class. Then, we iteratively distill the under-trained data with past predictions of half the batch size. The two-stage knowledge is linearly combined, efficiently enhancing model performance. Extensive experiments conducted on five backbone architectures show our method is model-agnostic and achieves the best generalization ***, TOAST is strongly compatible with existing augmentation-based regularization methods. Our method also obtains a speedup of up to 2.95x compared with a recent state-of-the-art method.

关键词： Cluster-based regularization iterative prediction refinement model-agnostic framework self-knowledge distillation(SKD) two-stage knowledge transfer

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：