检索结果-南通市图书馆

Fast target-aware learning for few-shot video object segmentation

维普期刊数据库评论

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Science China(Information Sciences) 2022年第8期65卷 71-86页

作者： yadang chen Chuanyan HAO Zhi-Xin YANG Enhua WU Engineering Research Center of Digital Forensics Ministry of EducationSchool of Computer and SoftwareNanjing University of Information Science and Technology School of Education Science and Technology Nanjing University of Posts and Telecommunications State Key Laboratory of Internet of Things for Smart City Department of Electromechanical EngineeringUniversity of Macau State Key Laboratory of Computer Science Institute of SoftwareUniversity of Chinese Academy of Sciences Faculty of Science and Technology University of Macau

Few-shot video object segmentation(FSVOS) aims to segment a specific object throughout a video sequence when only the first-frame annotation is given. In this study, we develop a fast target-aware learning approach for FSVOS, where the proposed approach adapts to new video sequences from its firstframe annotation through a lightweight procedure. The proposed network comprises two models. First, the meta knowledge model learns the general semantic features for the input video image and up-samples the coarse predicted mask to the original image size. Second, the target model adapts quickly from the limited support set. Concretely, during the online inference for testing the video, we first employ fast optimization techniques to train a powerful target model by minimizing the segmentation error in the first frame and then use it to predict the subsequent frames. During the offline training, we use a bilevel-optimization strategy to mimic the full testing procedure to train the meta knowledge model across multiple video *** proposed method is trained only on an individual public video object segmentation(VOS) benchmark without additional training sets and compared favorably with state-of-the-art methods on DAVIS-2017, with a J &F overall score of 71.6%, and on YouT ubeVOS-2018, with a J &F overall score of 75.4%. Meanwhile,a high inference speed of approximately 0.13 s per frame is maintained.

关键词： video object segmentation few-shot target-aware meta knowledge bilevel-optimization

Image completion with perspective constraint based on a single image

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Science China(Information Sciences) 2015年第9期58卷 110-121页

作者： HAO ChuanYan chen yadang WU Wen WU EnHua Faculty of Science and Technology University of Macau State Key Laboratory of Computer Science Institute of Software Chinese Academy of Sciences

Although image completion has been used in many image filling and editing problems for a long time, it is rarely applied in the scenario regarding the perspective effect which is demonstrated largely in the design of urban architectures. Thus this paper proposes a method to capture the perspective information automatically based on a single image and employs it to image completion tasks effectively. Actually it is difficult to automatize the extraction from only a single image, but it has been observed that architectures usually contain plenty of repetitions and symmetries which could be described through periodicity. Therefore,our approach first determines the initial vanishing points through the cascaded hough transformation. Then, a refinement scheme based on SIFT feature matching is proposed to obtain the accurate vanishing points. At last,a good perspective mesh is generated from the geometry calculation. The experiments show that this method could be used to automatically extract perspective information from a single image and recover the image in a visually plausible way.

关键词： perspective constraint image completion image inpainting texture synthesis mesh vanishing point hough transform SIFT features

Robust dense reconstruction by range merging based on confidence estimation

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Science China(Information Sciences) 2016年第9期59卷 22-32页

作者： yadang chen Chuanyan HAO Wen WU Enhua WU School of Computer and Software Nanjing University of Information Science and Technology Department of Computer and Information Science Faculty of Science and Technology University of Macau School of Education Science and Technology Nanjing University of Posts and Telecommunications State Key Laboratory of Computer Science Institute of Software Chinese Academy of Sciences

Although the stereo matching problem has been extensively studied during the past decades, automatically computing a dense 3D reconstruction from several multiple views is still a difficult task owing to the problems of textureless regions, outliers, detail loss, and various other factors. In this paper, these difficult problems are handled effectively by a robust model that outputs an accurate and dense reconstruction as the final result from an input of multiple images captured by a normal camera. First, the positions of the camera and sparse 3D points are estimated by a structure-from-motion algorithm and we compute the range map with a confidence estimation for each image in our approach. Then all the range maps are integrated into a fine point cloud data set. In the final step we use a Poisson reconstruction algorithm to finish the reconstruction. The major contributions of the work lie in the following points: effective range-computation and confidence-estimation methods are proposed to handle the problems of textureless regions, outliers and detail loss. Then, the range maps are merged into the point cloud data in terms of a confidence-estimation. Finally, Poisson reconstruction algorithm completes the dense mesh. In addition, texture mapping is also implemented as a post-processing work for obtaining good visual effects. Experimental results are presented to demonstrate the effectiveness of the proposed approach.

关键词： stereo matching 3D reconstruction textureless regions outliers details loss range map

Secure P2P topology based on a multidimensional DHT space mapping

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Science China(Information Sciences) 2013年第5期56卷 224-237页

作者： SUN ZhiXin LUO BingQing chen yadang BU Kai Key Laboratory of Broadband Wireless Communication and Sensor Network Technology(Nanjing University of Posts and Telecommunications) Ministry of Education State Key Laboratory for Novel Software Technology Nanjing University

Distributed search and routing algorithms based on the DHT(distributed hash table) protocol have attracted considerable attention in structured P2P(peer-to-peer) research as a result of favorable properties such as distribution,self-organization,and high *** with a traditional C/S(client/server) network,the probability of peers initiating malicious behavior increases dramatically because of their self-governing and dynamic characteristics,which also make it harder to satisfy the peers’ security required by *** this paper,we propose a new secure DHT protocol based on a multidimensional mapping *** mechanism maps peers to a multidimensional space by dividing the identifiers into ***,a series of secure methods and routing algorithms are provided to achieve secure DHT in smaller *** with state-ofthe-art approaches,the theoretical analysis and experimental results show that the multidimensional mapping mechanism can effectively improve the average success rate of a resource search by inhibiting malicious behavior.

关键词： multidimensional mapping DHT(distributed hash table) structured P2P(peer-to-peer) network routing attack security

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

多尺度特征融合的知识蒸馏异常检测方法

计算机辅助设计与图形学学报 2022年第10期34卷 1542-1549页

作者：陈亚当陈柳任余文斌朱加乐南京信息工程大学计算机学院、软件学院、网络空间安全学院南京210044 南京信息工程大学数字取证教育部研究中心南京210044

为提高一分类异常检测方法的泛化性,提出了多尺度特征融合的知识蒸馏异常检测方法.在训练阶段,使用训练好的教师网络教授学生网络学习正常样本特征信息;在测试阶段,由于教师网络强大的泛化能力,对未见过的异常样本也有较好的表示能力,... 详细信息

为提高一分类异常检测方法的泛化性,提出了多尺度特征融合的知识蒸馏异常检测方法.在训练阶段,使用训练好的教师网络教授学生网络学习正常样本特征信息;在测试阶段,由于教师网络强大的泛化能力,对未见过的异常样本也有较好的表示能力,而学生网络未见过异常样本,因此,两者对异常样本的表征会产生较大差异,从而实现异常检测.为了进一步提高检测精度,采用中层特征金字塔实现多尺度特征融合,以增强对不同大小异常点的检测能力;同时,为了增大学生和教师网络对异常样本的表征差异,设计了特征重构模块对中层特征进行重构.在公开数据集MVTecAD上的实验结果表明,文中方法分别达到较高的97.8%AUC像素级得分、97.7%AUC图像级得分.

关键词：异常检测知识蒸馏一分类问题特征融合

维普期刊数据库

A Phase Estimation Algorithm for Quantum Speed-Up Multi-Party Computing

在线全文

学校读者我要写书评

暂无评论

Computers, Materials & Continua 2021年第4期67卷 241-252页

作者： Wenbin Yu Hao Feng Yinsong Xu Na Yin yadang chen Zhi-Xin Yang Jiangsu Collaborative Innovation Center of Atmospheric Environment and Equipment Technology(CICAEET) Jiangsu Engineering Center of Network MonitoringSchool of Computer and SoftwareNanjing University of Information Science&TechnologyNanjing210044China Department of Computer Science and Engineering Michigan State UniversityEast Lansing48824MIUSA State Key Laboratory of Internet of Things for Smart City and Department of Electromechanical Engineering University of Macao999078Macao

Security and privacy issues have attracted the attention of researchers in the field of IoT as the information processing scale grows in sensor *** computing,theoretically known as an absolutely secure way to store and transmit information as well as a speed-up way to accelerate local or distributed classical algorithms that are hard to solve with polynomial complexity in computation or *** this paper,we focus on the phase estimation method that is crucial to the realization of a general multi-party computing model,which is able to be accelerated by quantum algorithms.A novel multi-party phase estimation algorithm and the related quantum circuit are proposed by using a distributed Oracle operator with *** proved theoretical communication complexity of this algorithm shows it can give the phase estimation before applying multi-party computing efficiently without increasing any additional ***,a practical problem of multi-party dating investigated shows it can make a successful estimation of the number of solution in advance with zero communication complexity by utilizing its special statistic *** simulations present the correctness,validity and efficiency of the proposed estimation method.

关键词： Edge computing security multi-party computing quantum algorithm phase estimation communication complexity

维普期刊数据库评论

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

融合视觉词与自注意力机制的视频目标分割

中国图象图形学报 2022年第8期27卷 2444-2457页

作者：季传俊陈亚当车洵南京信息工程大学计算机学院、软件学院、网络空间安全学院南京210044 数字取证教育部工程研究中心南京210044 南京众智维信息科技有限公司南京210006

目的视频目标分割(video object segmentation, VOS)是在给定初始帧的目标掩码条件下,实现对整个视频序列中感兴趣对象的分割,但是视频中往往会出现目标形状不规则、背景中存在干扰信息和运动速度过快等情况,影响视频目标分割质量。对... 详细信息

目的视频目标分割(video object segmentation, VOS)是在给定初始帧的目标掩码条件下,实现对整个视频序列中感兴趣对象的分割,但是视频中往往会出现目标形状不规则、背景中存在干扰信息和运动速度过快等情况,影响视频目标分割质量。对此,本文提出一种融合视觉词和自注意力机制的视频目标分割算法。方法对于参考帧,首先将其图像输入编码器中,提取分辨率为原图像1/8的像素特征。然后将该特征输入由若干卷积核构成的嵌入空间中,并将其结果上采样至原始尺寸。最后结合参考帧的目标掩码信息,通过聚类算法对嵌入空间中的像素进行聚类分簇,形成用于表示目标对象的视觉词。对于目标帧,首先将其图像通过编码器并输入嵌入空间中,通过单词匹配操作用参考帧生成的视觉词来表示嵌入空间中的像素,并获得多个相似图。然后,对相似图应用自注意力机制捕获全局依赖关系,最后取通道方向上的最大值作为预测结果。为了解决目标对象的外观变化和视觉词失配的问题,提出在线更新机制和全局校正机制以进一步提高准确率。结果实验结果表明,本文方法在视频目标分割数据集DAVIS(densely annotated video segmentation)2016和DAVIS 2017上取得了有竞争力的结果,区域相似度与轮廓精度之间的平均值J&F-mean(Jaccard and F-score mean)分别为83.2%和72.3%。结论本文提出的算法可以有效地处理由遮挡、变形和视点变化等带来的干扰问题,实现高质量的视频目标分割。

关键词：视频目标分割(VOS) 聚类算法视觉词自注意力机制在线更新机制全局校正机制

维普期刊数据库

在线全文

学校读者我要写书评

暂无评论

动态双边网格实现的视频前景分割算法

计算机辅助设计与图形学学报 2018年第11期30卷 2101-2107页

作者：陈亚当郝川艳南京信息工程大学计算机与软件学院南京210044 南京邮电大学科学与教育学院南京210023

针对视频前景分割技术中前/背景颜色相似、物体遮挡、数据冗余等难点,提出用双边网格的数据结构形式来实现视频前景分割.首先为视频数据构造一个高维的映射空间;然后根据标签数据为空间中每一个节点计算相应的分割值,并采用图割算法进... 详细信息

针对视频前景分割技术中前/背景颜色相似、物体遮挡、数据冗余等难点,提出用双边网格的数据结构形式来实现视频前景分割.首先为视频数据构造一个高维的映射空间;然后根据标签数据为空间中每一个节点计算相应的分割值,并采用图割算法进行分割;最后把高维特征空间中的结果降维到当前对应的视频帧中,完成分割.采用DAVIS数据对算法进行实验结果的表明,通过对分割结果的定性分析与定量对比,证明该算法切实可行,并能有效地解决上述问题.

关键词：视频处理前景分割标签数据双边网格

维普期刊数据库

Build Gaussian Distribution Under Deep Features for Anomaly Detection and Localization

在线全文

学校读者我要写书评

暂无评论

Journal of New Media 2022年第4期4卷 179-190页

作者： Mei Wang Hao Xu yadang chen School of Computer&Software Nanjing University of Information Science and TechnologyNanjing210044China College of Computer Science and Engineering Chongqing University of TechnologyChongqing400054China

Anomaly detection in images has attracted a lot of attention in the field of computer *** aims at identifying images that deviate from the norm and segmenting the defect within ***,anomalous samples are difficult to collect comprehensively,and labeled data is costly to obtain in many practical *** proposes a simple framework for unsupervised anomaly ***,the proposed method directly employs CNN pre-trained on ImageNet to extract deep features from normal images and reduce dimensionality based on Principal Components Analysis(PCA),then build the distribution of normal features via the multivariate Gaussian(MVG),and determine whether the test image is an abnormal image according to Mahalanobis *** further investigate which features are most effective in detecting *** experiments on the MVTec anomaly detection dataset show that the proposed method achieves 98.6%AUROC in image-level anomaly detection and outperforms previous methods by a large margin.

关键词： Anomaly detection dimensionality reduction multivariate gaussian visual inspection