检索结果-内蒙古大学图书馆

STIT: Spatio-Temporal Interaction Transformers for Human-Object Interaction recognition in Videos 26

STIT: Spatio-Temporal Interaction Transformers for Human-Obj...

26th international Conference on pattern recognition / 8th international workshop on Image Mining - theory and Applications (IMTA)

作者： Almushyti, Muna Li, Frederick W. B. Univ Durham Dept Comp Sci Durham England Qassim Univ Educ Serv Buraydah Saudi Arabia

ISBN: (数字)9781665490627

ISBN: (纸本)9781665490627

Recognizing human-object interactions is challenging due to their spatio-temporal changes. We propose the Spatio-Temporal Interaction Transformer-based (STIT) network to reason such changes. Specifically, spatial transformers learn humans and objects context at specific frame time. Temporal transformer then learns the relations at a higher level between spatial context representations at different time steps, capturing long-term dependencies across frames. We further investigate multiple hierarchy designs in learning human interactions. We achieved superior performance on Charades, Something-Something v1 and CAD-120 datasets, comparing to baseline models without learning human-object relations, or with prior graph-based networks. We also achieved state-of-the-art accuracy of 95.93% on CAD-120 dataset [1] by employing RGB data only.

关键词： Deep learning Transformers pattern recognition Videos Context modeling

来源：评论

学校读者我要写书评

暂无评论

Quadratic Kernel Learning for Interpolation Kernel Machine based graph Classification 13th

Quadratic Kernel Learning for Interpolation Kernel Machine ...

引用

13th IAPR-TC-15 international workshop on graph-based representations in pattern recognition, GbRPR 2023

作者： Zhang, Jiaqi Liu, Cheng-Lin Jiang, Xiaoyi Faculty of Mathematics and Computer Science University of Münster Einsteinstrasse 62 Münster48149 Germany National Laboratory of Pattern Recognition Institute of Automation of Chinese Academy of Sciences Beijing100190 China School of Artificial Intelligence University of Chinese Academy of Sciences Beijing10049 China

ISBN: (纸本)9783031427947

Interpolating classifiers interpolate all the training data and thus have zero training error. Recent research shows their fundamental importance for high-performance ensemble techniques. Interpolation kernel machines belong to the class of interpolating classifiers and do generalize well. they have been demonstrated to be a good alternative to support vector machine for graph classification. In this work we further improve their performance by considering multiple kernel learning. We establish a general scheme for achieving this goal. the current experimental work is done using quadratic kernel combination. Our experimental results demonstrate the performance boosting potential of our approach against the use of individual graph kernels. © 2023, the Author(s), under exclusive license to Springer Nature Switzerland AG.

关键词： Support vector machines

来源：评论

学校读者我要写书评

暂无评论

Reducing the Computational Complexity of the Eccentricity Transform of a Tree 13th

Reducing the Computational Complexity of the Eccentricity...

引用

13th IAPR-TC-15 international workshop on graph-based representations in pattern recognition, GbRPR 2023

作者： Banaeyan, Majid Kropatsch, Walter G. Pattern Recognition and Image Processing Group TU Wien Vienna Austria

ISBN: (纸本)9783031427947

this paper proposes a novel approach to reduce the computational complexity of the eccentricity transform (ECC) for graph-based representation and analysis of shapes. the ECC assigns to each point within a shape its geodesic distance to the furthest point, providing essential information about the shape’s geometry, connectivity, and topology. Although the ECC has proven valuable in numerous applications, its computation using traditional methods involves heavy computational complexity. To overcome this limitation, we present a method that computes the ECC of a tree, significantly reducing the computational complexity from O(n2log(n) ) to O(b), where n and b are the numbers of vertices and branching points in the tree, respectively. Our method begins by computing the ECC for tree structures, which are simpler representations of shapes. Subsequently, we introduce the concept of a 3D curve that corresponds to a smooth shape without holes, enabling the computation of the ECC for more complex shapes. By leveraging the 3D curve representation, our method provides an upper-bound approximation of the ECC, which can be effectively utilized in various applications. the proposed approach not only preserves the valuable properties of the ECC but also significantly reduces the computational burden, making it a more efficient and practical solution for graph-based representation and analysis of shapes in both 2D and 3D contexts. © 2023, the Author(s), under exclusive license to Springer Nature Switzerland AG.

关键词： Computational complexity

来源：评论

学校读者我要写书评

暂无评论

Find More Accurate Text Boundary for Scene Text Detection 26

Find More Accurate Text Boundary for Scene Text Detection

引用

26th international Conference on pattern recognition / 8th international workshop on Image Mining - theory and Applications (IMTA)

作者： Pan, Runqiu Li, Zezhou Zhu, Anna Wuhan Univ Technol Sch Comp Sci & Artificial Intelligence Wuhan Peoples R China

ISBN: (数字)9781665490627

ISBN: (纸本)9781665490627

Arbitrary shape text detection is still an open problem due to several challenges, e.g., dense text adhesion, various sizes, and noises. In this work, we propose a novel scene text detection network by extracting and connecting a set of points on the boundary of each text instance. Our method mainly consists of three branches, including text center line (TCL) prediction, text orientation (TO) prediction, and text boundary offset (TBO) prediction to get the text boundary point proposals. Utilize these proposals, and the final refinement results can be obtained by point sampling and graph attention network(GAT). the detector can overcome the text instance sticking problem with these text boundary representations. Additionally, we propose distance-based dice loss and instance-aware L1 loss to remove false positives and get over various text sizes on text boundary offset prediction respectively. In this way, our method can directly and efficiently generate accurate text boundaries without any post-processing. Extensive experiments on publicly available datasets show the effectiveness of our design and training strategy, which also demonstrates our method's state-of-the-art performance for arbitrary shape text detection.

关键词： Training Shape Adhesives Detectors pattern recognition Proposals

来源：评论

学校读者我要写书评

暂无评论

Memegraphs: Linking Memes to Knowledge graphs 17th

MemeGraphs: Linking Memes to Knowledge Graphs

引用

17th international Conference on Document Analysis and recognition (ICDAR)

作者： Kougia, Vasiliki Fetzel, Simon Kirchmair, thomas Cano, Erion Baharlou, Sina Moayed Sharifzadeh, Sahand Roth, Benjamin Univ Vienna Fac Comp Sci Vienna Austria Boston Univ Dept Elect & Comp Engn Boston MA 02215 USA Ludwig Maximilians Univ Munchen Fac Comp Sci Munich Germany

ISBN: (纸本)9783031416750;9783031416767

Memes are a popular form of communicating trends and ideas in social media and on the internet in general, combining the modalities of images and text. they can express humor and sarcasm but can also have offensive content. Analyzing and classifying memes automatically is challenging since their interpretation relies on the understanding of visual elements, language, and background knowledge. thus, it is important to meaningfully represent these sources and the interaction between them in order to classify a meme as a whole. In this work, we propose to use scene graphs, that express images in terms of objects and their visual relations, and knowledge graphs as structured representations for meme classification with a Transformer-based architecture. We compare our approach with ImgBERT, a multimodal model that uses only learned (instead of structured) representations of the meme, and observe consistent improvements. We further provide a dataset with human graph annotations that we compare to automatically generated graphs and entity linking. Analysis shows that automatic methods link more entities than human annotators and that automatically generated graphs are better suited for hatefulness classification in memes.

关键词： hate speech internet memes knowledge graphs multimodal representations

来源：评论

学校读者我要写书评

暂无评论

Trends in graph -based representations for pattern recognition

引用

pattern recognition LETTERS 2020年 134卷 3-9页

作者： Brun, Luc Foggia, Pasquale Vento, Mario Univ Salerno Dept Informat & Elect Engn & Appl Math Fisciano Italy GreyC Ecole Natl Super Ingn Caen ENSICAEN Caen France

In this paper we try to examine recent trends on the use of graph-based representations in pattern recognition, using as a vantage point the 11th IAPR-TC15 workshop GbR2017, dedicated to this topic. A survey of the paper presented at GbR2017 will give us the opportunity to reflect on the directions where the interest of the research community working on this subject is moving. (C) 2018 Elsevier B.V. All rights reserved.

关键词： graph-based representations graph matching graph edit distance graph kernels

来源：评论

学校读者我要写书评

暂无评论

Point pattern Feature Matching Method based on Polar Coordinate 2

Point Pattern Feature Matching Method Based on Polar Coordin...

引用

2nd IEEE international Conference on Software Engineering and Artificial Intelligence (SEAI) / 7th international workshop on pattern recognition (IWPR)

作者： Xu, Qin Zhang, Lei City Univ Macau Inst Data Sci Macau Peoples R China Tongji Univ Dept Traff Informat & Control Engn Shanghai Peoples R China

ISBN: (数字)9781665482233

ISBN: (纸本)9781665482233

Infrared label point feature matching is one of the basic problems in the application direction of computer vision, pattern recognition and robot navigation and positioning. First, the center of the label point set and the weighted distance center are obtained in Cartesian coordinates, then the Cartesian coordinates are converted into polar coordinates to construct the position information feature vector of the label point set. Finally, using the method of spatial geometric constraints to compare each point in the label code to the feature vector, the matching result of the label code point set is obtained. the contributions of this paper are as follows: 1. A feature vector of label point set based on polar coordinate transformation is constructed, which makes the feature vector invariant to rotation, translation and illumination. 2. Using the combination of feature vector and spatial geometric constraints to screen label point pairs can effectively overcome the adverse effects of outlier points and noise. the final experiments demonstrate the effectiveness and robustness of the polar coordinate feature matching of label codes.

关键词： pattern recognition robot vision computer vision

来源：评论

学校读者我要写书评

暂无评论

11th IAPR-TC-15 international workshop on graph-based representations in pattern recognition, GbRPR 2017

11th IAPR-TC-15 International Workshop on Graph-Based Repres...

引用

11th IAPR-TC-15 international workshop on graph-based representations in pattern recognition, GbRPR 2017

ISBN: (纸本)9783319589602

the proceedings contain 26 papers. the special focus in this conference is on graph-based representations in pattern recognition. the topics include: Saliency detection via a graph based diffusion model;shape simplification through graph sparsification;reeb graphs of piecewise linear functions;learning graph matching with a graph-based perceptron in a classification context;a nested alignment graph kernel through the dynamic time warping framework;speeding-up graph-based keyword spotting in historical handwritten documents;error-tolerant coarse-to-fine matching model for hierarchical graphs;a Hungarian algorithm for error-correcting graph matching;a new algorithm for subgraph isomorphism;node matching computation between two large graphs in linear computational cost;measuring vertex centrality using the holevo quantity;on the interplay between strong regularity and graph densification;density normalization in density peak based clustering;fast nearest neighbors search in graph space based on a branch-and-bound strategy;exact computation of graph edit distance for uniform and non-uniform metric edit costs;improved graph edit distance approximation with simulated annealing;an edit distance between graph correspondences;a survey on applications of bipartite graph edit distance;minimising entropy changes in dynamic network evolution;synchronization over the birkhoff polytope for multi-graph matching and adaptive feature selection based on the most informative graph-based features.

关键词：

来源：评论

学校读者我要写书评

暂无评论

IWCLUL 2021 - 7th international workshop on Computational Linguistics of Uralic Languages, Proceedings

IWCLUL 2021 - 7th International Workshop on Computational Li...

引用

7th international workshop on Computational Linguistics of Uralic Languages, IWCLUL 2021

ISBN: (纸本)9781954085824

the proceedings contain 8 papers. the topics discussed include: the current state of Finnish NLP;integrating pattern- and fact-based fake news detection via model preference learning;a second pandemic? analysis of fake news about COVID-19 vaccines in Qatar;WRENCH: a comprehensive benchmark for weak supervision;cluster-based mention typing for named entity disambiguation;named entity recognition and classification on historical documents: a survey;an algorithm for generating gap-fill multiple choice questions of an expert system;and automated fact-checking: a survey.

关键词：

来源：评论

学校读者我要写书评

暂无评论

A Spatial Small Object Detection Algorithm based on the Improved YOLOv8

A Spatial Small Object Detection Algorithm Based on the Impr...

引用

pattern recognition and Artificial Intelligence (PRAI), international Conference on

作者： Meijuan Zhou Ning Wang Zhongxi Wang Xi Liu Zhong Ma Xi'an Microelectronics Technology Institute Xi'an China

ISBN: (数字)9798350350890

ISBN: (纸本)9798350350906

In spatial object detection tasks, due to the far observation distance, objects often exist in the form of point objects. Due to the extremely small size of the object, the number of detectable objects is very limited, which can easily lead to missed detection and false positives in object detection. To solve these issues, we proposes a spatial object detection algorithm based on the improved YOLOv8. this algorithm enhances the network's feature extraction ability by adding GAM Attention to the Neck part of the YOLOv8 object detection algorithm, reduces information reduction, and magnifies the global interactive representations to improve the performance of deep neural networks and reduce the interference of background noise in small object detection tasks in complex environments. It introduces 3D permutation and multilayer perceptron to achieve channel attention and a convolutional spatial attention submodule to better capture the detail information of small objects. the experimental results show that the improved YOLOv8 algorithm has improved the $\boldsymbol{mAP_{50}}$ by 1.3%, and $\boldsymbol{mAP_{50-95}}$ by 1.9% on the ISS dataset. In conclusion, the improved YOLOv8 can enhance the performance of small object detection.

关键词： three-dimensional displays Object detection Interference Artificial neural networks Multilayer perceptrons Feature extraction Neck pattern recognition Background noise Artificial intelligence

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：