检索结果-内蒙古大学图书馆

International Workshop on Advanced Computational Intelligence (IWACI)

作者： Yiren Wang Hassan Dawood Qian Yin Ping Guo the Image Processing and Pattern Recognition Laboratory Beijing Normal University Beijing China the Image Processing and Pattern Recognition Laboratory Beijing Normal University Beijing China Computer Science Department Beijing Normal University Beijing China

Automatic image annotation and tagging is necessary for indexing and searching of images using querying a text. It is widely used in search engines like Google, Yahoo, Baidu, etc. Fast image Tagging (FastTag) algorithm is proposed to accelerate image annotation process, while keeping the precision of automatic image annotation results. Feature mapping is used to map image features vectors onto higher dimensional feature space. Feature mapping methods plays an important role in automatic image annotation. In this paper, we have compared 6 kernels, among which four kernels are used in homogeneous feature mapping and two kernels are used in discriminative tree based feature mapping, to investigate which feature mapping performs better for automatic image annotation. The performance of these methods has been analyzed by conducting intensive experiments on three different datasets as used by FastTag algorithm in their experiments. We have found that the homogeneous feature mapping with χ 2 kernel is more suitable when used in FastTag algorithm in terms of precision, recall, FI score and N+ measures, and with a relatively acceptable performance.

关键词： Phase change materials Search engines Kernel Indexing Histograms

来源：评论

学校读者我要写书评

暂无评论

Towards Better Data Exploitation in Self-Supervised Monocular Depth Estimation

arXiv

引用

arXiv 2023年

作者： Liu, Jinfeng Kong, Lingtong Yang, Jie Liu, Wei Institute of Image Processing and Pattern Recognition Department of Automation Shanghai Jiao Tong University Shanghai China

Depth estimation plays an important role in robotic perception systems. The self-supervised monocular paradigm has gained significant attention since it can free training from the reliance on depth annotations. Despite recent advancements, existing self-supervised methods still underutilize the available training data, limiting their generalization ability. In this paper, we take two data augmentation techniques, namely Resizing-Cropping and Splitting-Permuting, to fully exploit the potential of training datasets. Specifically, the original image and the generated two augmented images are fed into the training pipeline simultaneously and we leverage them to conduct self-distillation. Additionally, we introduce the detail-enhanced DepthNet with an extra full-scale branch in the encoder and a grid decoder to enhance the restoration of fine details in depth maps. Experimental results demonstrate our method can achieve state-of-the-art performance on the KITTI and Cityscapes datasets. Moreover, our KITTI models also show superior generalization performance when transferring to Make3D, NYUv2 and Cityscapes datasets. Our codes are available at https://***/LiuJF1226/BDEdepth. Copyright © 2023, The Authors. All rights reserved.

关键词： Distillation

来源：评论

学校读者我要写书评

暂无评论

Exponential image Enhancement in Daytime Fog Conditions

Exponential Image Enhancement in Daytime Fog Conditions

引用

17th International IEEE Conference on Intelligent Transportation Systems & The Asia-Pacific Council on Systems Engineering Conference 2014

作者： Mihai Negru Sergiu Nedevschi Radu Ioan Peter Image Processing and Pattern Recognition Group Computer Science DepartmentTechnical University of Cluj-Napoca head of the Image Processing and Pattern Recognition Group Computer Science DepartmentTechnical University of Cluj-Napoca Mathematics Department Technical University of Cluj-Napoca

The images captured in fog conditions have degraded contrast,that makes current image processing applications sensitive and error *** propose in this paper an efficient single image enhancement algorithm suitable for daytime fog conditions and based on an original mathematical model,for computing the atmospheric veil,that takes into account the variation in fog density to the *** model is inspired by the functions that appear in partition of unity in the differential geometry *** observing images captured in fog conditions,usually the fog has a very low density in front of the camera and this density has a non-linear increase with the distance,such that objects are no longer visible at greater *** using our mathematical model we are able to obtain superior reconstructions of the original fog-free image,when comparing to traditional *** advantage of our method is the ability to adapt the model in accordance to the density of the fog.A quantitative and qualitative evaluation is performed on both synthetic and real camera *** evaluation proves that our mathematical model is more suitable for image enhancement in both homogeneous and heterogeneous fog *** algorithm is able to perform image enhancement in real time for both color and gray scale images.

关键词： image Exponential image Enhancement in Daytime Fog Conditions

来源：评论

学校读者我要写书评

暂无评论

ArtDiff: Artwork Generation via Conditional Diffusion Models 23

ArtDiff: Artwork Generation via Conditional Diffusion Models

引用

15th International Conference on Digital image processing, ICDIP 2023

作者： Liu, Wenhui Liu, Bin Lin, Xiao Li, Bo He, Zhifen Wang, Kang School of Mathematics and Information Science Key Laboratory of Jiangxi Province for Image Processing and Pattern Recognition Nanchang Hangkong University Jiangxi Nanchang China Department of Supercomputing Management Jiangxi Science and Technology Infrastructure Center Jiangxi Nanchang China

ISBN: (纸本)9798400708237

Artwork Generation is an important research area of computer vision. Recently, kinds of generative models have achieved great success in natural image generation. However, artwork generation has rarely been studied due to the unfixed structure of artworks. Combined with prevailing diffusion models, we propose a simple yet effective framework, named as ArtDiff, for artwork generation. Given a name of an artist, we can generate diverse and novel artworks which reflecting the style of the artist. To the best of our knowledge, ArtDiff is the first diffusion model that generate artworks with the guidance of artist's name. Experimental results demonstrate that ArtDiff is able to recognize the artist's preference and generate artworks with reasonable structures and fine-grained details. © 2023 ACM.

关键词： Diffusion

来源：评论

学校读者我要写书评

暂无评论

FG-UAP: Feature-Gathering Universal Adversarial Perturbation

FG-UAP: Feature-Gathering Universal Adversarial Perturbation

引用

International Joint Conference on Neural Networks (IJCNN)

作者： Zhixing Ye Xinwen Cheng Xiaolin Huang Department of Automation Institute of Image Processing and Pattern Recognition Shanghai Jiao Tong University Shanghai China

Deep Neural Networks (DNNs) are susceptible to elaborately designed perturbations, whether such perturbations are dependent or independent of images. The latter one, called Universal Adversarial Perturbation (UAP), is very attractive for model robustness analysis, since its independence of input reveals the intrinsic characteristics of the model. Relatively, another interesting observation is Neural Collapse (NC), which means the feature variability may collapse during the terminal phase of training. Motivated by this, we propose to generate UAP by attacking the layer where NC phenomenon happens. Because of NC, the proposed attack could gather all the natural images’ features to its surrounding, which is hence called Feature-Gathering UAP (FG-UAP). We evaluate the effectiveness our proposed algorithm on abundant experiments, including untargeted and targeted universal attacks, attacks under limited dataset, and transfer-based black-box attacks among different architectures including Vision Transformers, which are believed to be more robust. After that, we empirically verify the effectiveness of NC's conclusion on UAP by attacking on only 10% of the dataset while keeping comparable performance. Finally, we investigate FG-UAP in the view of NC by analyzing the labels and extracted features of adversarial examples, finding that collapse phenomenon becomes stronger after the model is corrupted. Codes for the project are available at https://***/yzx1213/FG-UAP.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Layout and analysis: Finding text, titles, and photos in digital images of newspaper pages

Layout and analysis: Finding text, titles, and photos in dig...

引用

International Conference on Document Analysis and recognition

作者： J. Wieser A. Pinz Department for Pattern Recognition and Image Processing Inst. for Automation Technical University of of Vienna Wien Austria

An important step in the analysis of printed documents is the segmentation and classification of blocks into categories such as photographs, titles, paragraphs, etc. The authors present an approach to enhance and combine two commonly used methods, a merging bottom-up approach and a cutting top down approach, to segment pages of a newspaper. The implementation of a layout analysis system as a preprocessing module for a commercial product is described.< >

关键词： image analysis Digital images image segmentation Merging pattern analysis image converters Text analysis Optical character recognition software Particle separators Size measurement

来源：评论

学校读者我要写书评

暂无评论

A Modified Dense Displacement Field Estimation Algorithm Combined with Block-matching Method

A Modified Dense Displacement Field Estimation Algorithm Com...

引用

2000年第16届世界计算机大会

作者： Guo Fengjun YangXin Wilfried Philips Shi Penfei Institute of Image Processing & Pattern Recognition Shanghai Jiao Tong University TELIN Department Ghent University

In this paper,we presented a displacement field estimation algorithm based on a relaxed smoothness constraint;this algorithmcan preserve discontinuities in the displacement field to some *** image data is irregular and the images are noisy,the method produces some big residual errors in the residual *** this paper we propose an improved displacement field estimation algorithm which uses the displacement information obtained using blockmatching to modify the matching *** results show,this leads to smaller residual error maps, without introducing block artefacts,as would happen in the case of simple block matching when there is much noise in the *** the displacement filed using this method is more consistent than using a method without additional block matching.

关键词： image matching displacement field discontinuity preservation elastic matching conjugate gradient method block-matching.

来源：评论

学校读者我要写书评

暂无评论

Multi-Class Segmentation for Traffic Scenarios at Over 50 FPS

Multi-Class Segmentation for Traffic Scenarios at Over 50 FP...

引用

IEEE Intelligent Vehicles Symposium

作者： Arthur D. Costea Sergiu Nedevschi Image Processing and Pattern Recognition Group Computer Science Department Technical University of Cluj-Napoca Romania

ISBN: (纸本)9781479936397

Multi-class segmentation assigns a class label to each pixel in an image. It represents a significant task for the semantic understanding of images and has received plentiful attention over the last years. The current state of art is dominated by conditional random field based approaches, defined over pixels or image segments. However, high accuracy segmentation comes at a high computational cost. The best performing methods can barely run at few frames per second and are far from real-time applications. Our goal is to bridge the gap between current state of the art segmentation approaches and real-time applications. In this paper we propose an efficient approach for individual pixel classification. Multiple local descriptors are computed densely and then quantized using visual codebooks. Joint boosting is used to classify each pixel based on the quantized local descriptors. We show that using careful design choices and GPU optimization we can achieve sate of the art segmentation results at over 50 FPS. We also propose a Conditional Random Field (CRF) model defined over superpixels that uses the proposed pixel classifier for the estimation of unary potentials. The CRF based multi-class segmentation can run at over 30 FPS. The proposed approach is validated on the MSRC21 and CamVid multi-class segmentation benchmarks, the former one consisting of urban traffic sequences.

关键词： choices GPU CRF

来源：评论

学校读者我要写书评

暂无评论

Moving rigid objects segmentation in 3D dynamic traffic scenes using a stereovision system

Moving rigid objects segmentation in 3D dynamic traffic scen...

引用

IEEE International Conference on Intelligent Computer Communication and processing (ICCP)

作者： Catalin Golban Sergiu Nedevschi Computer Science Department Technical University Image Processing and Pattern Recognition Group Cluj-Napoca Romania

This paper proposes a novel method for detecting the moving vehicles in dynamic urban traffic scenes using a stereo camera. Relying on the fact that a set of feature points on a rigid 3D scene object are staying in a rigid 3D configuration, we propose to compute the relative motion between the camera and a moving object with an algorithm that follows from the visual odometry based motion estimation methods. Subtracting the camera motion we obtain the absolute object motion. Additionally we create a compact representation of the scene using superpixels computed from intensity and depth information. A graph-like structure is built, having superpixels as nodes and indicating neighboring relationships between adjacent superpixels. Objects are segmented using a fast region growing algorithm that considers as seeds the features used to compute the object motion.

关键词： Three-dimensional displays Visualization Computer vision Cameras Feature extraction Stereo vision image motion analysis

来源：评论

学校读者我要写书评

暂无评论

Propagation of uncertainty in landmark based self-localization of autonomous mobile robots

Propagation of uncertainty in landmark based self-localizati...

引用

10th IAPR Conference on Machine Vision Applications, MVA 2007

作者： Bais, Abdul Sablatnig, Robert Khawaja, Yahya M. Novak, Gregor Institute of Computer Technology Vienna University of Technology Vienna Austria Pattern Recognition and Image Processing Group Institute of Computer Aided Automation ViennaUniversity of Technology Vienna Austria Department of Electrical Engineering NWFP University of Engineering and Technology Peshawar Pakistan

ISBN: (纸本)9784901122078

This paper presents uncertainty propagation in landmark based position estimation methods. Analysis of two methods has been carried out where robot position is estimated by detecting one or two globally distinct features using a pivoted stereo vision system. We make a basic assumption about error in estimating point features in camera images and propagate it into robot position estimate using first order approximation of non-linear functions. Simulation results illustrate the performance of the method.

关键词： Stereo image processing

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：