检索结果-内蒙古大学图书馆

Generalized Closed-Form Formulae for Feature-Based Subpixel Alignment in Patch-Based Matching

International Journal of computer Vision 2025年 1-18页

作者： Jospin, Laurent Valentin Laga, Hamid Boussaid, Farid Bennamoun, Mohammed Faculté ENAC École Polytechnique Fédérale de Lausanne Station 15 VD Lausanne CH-1015 Switzerland School of Computer Science and Software Engineering University of Western Australia 35 Stirling Highway Perth 6009 WA Australia School of Information Technology Murdoch University 90 South Street Murdoch 6150 WA Australia School of Electrical Engineering University of Western Australia 35 Stirling Highway Perth 6009 WA Australia

Patch-based matching is a technique meant to measure the disparity between pixels in a source and target image and is at the core of various methods in computer vision. When the subpixel disparity between the source and target images is required, the cost function or the target image has to be interpolated. While cost-based interpolation is easier to implement, multiple works have shown that image-based interpolation can increase the accuracy of the disparity estimate. In this paper we review closed-form formulae for subpixel disparity computation for one dimensional matching, e.g., rectified stereo matching, for the standard cost functions used in patch-based matching. We then propose new formulae to generalize to high-dimensional search spaces, which is necessary for unrectified stereo matching and optical flow. We also compare the image-based interpolation formulae with traditional cost-based formulae, and show that image-based interpolation brings a significant improvement over the cost-based interpolation methods for two dimensional search spaces, and small improvement in the case of one dimensional search spaces. The zero-mean normalized cross correlation cost function is found to be preferable for subpixel alignment. A new error model, based on very broad assumptions is outlined in the Supplementary Material to demonstrate why these image-based interpolation formulae outperform their cost-based counterparts and why the zero-mean normalized cross correlation function is preferable for subpixel alignement. © The Author(s) 2025.

关键词： Patch-based matching Predictive interpolation Stereo Subpixel

来源：评论

学校读者我要写书评

暂无评论

software Bug Count Prediction Using Abstract Syntax Trees (ASTs) 24

Software Bug Count Prediction Using Abstract Syntax Trees (A...

引用

Proceedings of the 2024 13th International Conference on software and Information engineering

作者： Rahmeh Fawaz Ibrahim Abdallah Qusef School of Computer Science Princess Sumaya University for Technology (PSUT)Jordan Amman Jordan School of Software Engineering Princess Sumaya University for Technology (PSUT)Jordan Amman Jordan

来源：评论

学校读者我要写书评

暂无评论

SeaCap: Multi-Sight Embedding and Alignment for One-Stage Image Captioner

引用

IEEE Transactions on Multimedia 2025年

作者： Wang, Bo Zhang, Zhao Zhao, Mingbo Jin, Xiaojie Xu, Mingliang Wang, Meng Hefei University of Technology School of Computer Science and Information Engineering Hefei230601 China Yunnan Key Laboratory of Software Engineering Yunan China Donghua University School of Computer Science and Engineering Shanghai201620 China ByteDance Culver CityCA90230 United States Zhengzhou University School of Computer and Artificial Intelligence Zhengzhou450001 China

Recent mainstream image captioning methods usually adopt two-stage captioners, i.e., calculating the object features of the given image by a pre-trained detector and then feeding them into a language model to generate the descriptive sentences. However, such a two-stage procedure will lead to a task-based information gap that decreases the performance of the captioners, because the object features learned from the detection task are suboptimal representations and cannot provide all the necessary information for subsequent sentence generation. Besides, the object features are usually represented by the last pooling features of the detector that lose the local details of images. In this paper, we propose a novel One-Stage Image Captioner using dynamic multi-sight embedding and alignment, called SeaCap, which directly transforms input images into descriptive sentences in one stage to eliminate the information gap. Specifically, to obtain rich features, we use the Swin Transformer to capture multi-level features, followed by a sights alignment module to alleviate the vision confusion, and then feed them into a novel dynamic multi-sight embedding module to exploit both the global structure and local texture of input images. To enhance the global modeling capacity of the visual encoder, we propose a new dual-dimensional refining module to non-locally model the interaction of the embedded features. As a result, SeaCap can obtain rich and useful information to improve the performance of the captioner. Extensive comparisons on the benchmark MS-COCO, Flickr8K and Flickr30 K datasets verified the superior performance of our method. © 1999-2012 IEEE.

关键词： Embeddings

来源：评论

学校读者我要写书评

暂无评论

Forecasting land use changes in crop classification and drought using remote sensing

引用

Journal of Arid Land 2025年第5期17卷 575-589页

作者： Mashael MAASHI Nada ALZABEN Noha NEGM Venkatesan VEERAMANI Sabarunisha Sheik BEGUM Geetha PALANIAPPAN Department of Software Engineering College of Computer and Information SciencesKing Saud UniversityRiyadh 11451Saudi Arabia Department of Computer Sciences College of Computer and Information SciencesPrincess Nourah bint Abdulrahman UniversityRiyadh 11671Saudi Arabia Department of Computer Science Applied College at MahayilKing Khalid UniversityAbha 61421Saudi Arabia Department of Civil Engineering University College of EngineeringAnna UniversityAriyalur 621731India Department of Biotechnology P.S.R.Engineering CollegeSivakasi 626140India Department of Electronics and Communication Engineering School of Engineering and TechnologyDhanalakshmi Srinivasan UniversitySamayapuram 621112India

Challenges in land use and land cover(LULC)include rapid urbanization encroaching on agricultural land,leading to fragmentation and loss of natural ***,the effects of urbanization on LULC of different crop types are less *** study assessed the impacts of LULC changes on agriculture and drought vulnerability in the Aguascalientes region,Mexico,from 1994 to 2024,and predicted the LULC in 2034 using remote sensing data,with the goals of sustainable land management and climate resilience *** increasing urbanization and drought,the integration of satellite imagery and machine learning models in LULC analysis has been underutilized in this *** Landsat imagery,we assessed crop attributes through indices such as normalized difference vegetation index(NDVI),normalized difference water index(NDWI),normalized difference moisture index(NDMI),and vegetation condition index(VCI),alongside watershed delineation and spectral *** random forest model was applied to classify LULC,providing insights into both historical and future *** indicated a significant decline in vegetation cover(109.13 km^(2))from 1994 to 2024,accompanied by an increase in built-up land(75.11 km^(2))and bare land(67.13 km^(2)).Projections suggested a further decline in vegetation cover(41.51 km^(2))and continued urban land expansion by *** study found that paddy crops exhibited the highest values,while common bean and maize performed *** analysis revealed that mildly dry areas in 2004 became severely dry in 2024,highlighting the increasing vulnerability of agriculture to climate *** study concludes that sustainable land management,improved water resource practices,and advanced monitoring techniques are essential to mitigate the adverse effects of LULC changes on agricultural productivity and drought resilience in the *** findings contribute to the understanding of how remote sensing can be effectively used for long-t

关键词： land use and land cover(LULC) crop attributes drought vulnerability machine learning models remote sensing

来源：评论

学校读者我要写书评

暂无评论

FedVC: Virtual Clients for Federated Learning with Imbalanced Label Distribution 24

FedVC: Virtual Clients for Federated Learning with Imbalance...

引用

3rd International Conference on Algorithms, Data Mining, and Information technology, ADMIT 2024

作者： Chen, Zhouguo Chen, Hongda Liu, Shuai Liu, Yongdong Schoolof Computer Science and Engineering Southeast University Jiangsu Nanjing China The 30th Research Institute of China Electronics Technology Group Corporation Sichuan Chengdu China School of Software Engineering Xian Jiaotong University Shaanxi Xian China China Industrial Control Systems Cyber Emergency Response Team Beijing China

ISBN: (纸本)9798400718120

Federated Learning (FL) has gained considerable attention for collaborative training in big data analysis, particularly in terms of privacy and communication constraints. Despite its promising advantages, FL faces the challenge of dealing with imbalanced label distribution on edge devices, which complicates the learning task. In literature, numerous studies have attempted to tackle this issue by introducing additional hyperparameters, which can complicate and prolong the process of model training and tuning. This complexity may potentially lead to overfitting and limited performance on new datasets. In this paper, we present a new Federated Learning framework called FedVC, which introduces virtual clients to minimize hyperparameter reliance. FedVC aims to bridge the gap between balanced and imbalanced label distribution. Compared with traditional federated learning frameworks, the proposed FedVC model designs a dynamically balanced dataset by utilizing virtual clients, which is proposed for the first time in this work. These virtual clients determine which part of the data should be used in one single training round and manage backpropagation timing for local real clients. The balanced dataset promotes greater stability in local models, thereby enhancing the overall stability of the training process. The experimental results show that FedVC outperforms classical FedAvg and the compared baseline methods for the data with imbalanced label distribution. Especially, FedVC exhibits more stable and accurate results than alternative approaches in highly heterogeneous settings. The source code is available at https://***/aikedaerC/FedVC. © 2024 Copyright held by the owner/author(s).

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

A Semantic and Intelligent Focused Crawler based on BERT Semantic Vector Space Model and Hybrid Algorithm (October 2024)

引用

IEEE Access 2025年

作者： Huang, Wenhao Zhang, Jiahao Li, Xin Zhou, Xiao Qi, Deyu Xi, Jianqing Liu, Wenjun School of Software Engineering South China University of Technology Guangzhou510006 China School of Computer and Software Engineering XiHua University Chengdu610039 China Guangdong University of Foreign Studies South China Business College Guangzhou510545 China Sichuan Provincial Engineering Research Center of Hydroelectric Energy Power Equipment Technology Chengdu610039 China

The goal of a focused crawler is to selectively fetch pages that are relevant to a given topic. Previous crawlers use text content to determine text topic relevance and manually determined weighting factors to predict the priority of unvisited URLs. However, there are still some problems in the above focused crawler methods, the calculation formula of semantic similarity between words is flawed. The weighting factor for the priority of unvisited URLs is determined arbitrarily. In order to solve the above problems, this paper proposes a semantic and intelligent focused crawler based on BERT semantic vector space model and hybrid algorithm. This method used BERT semantic vector space model to calculate the topic relevance of documents, and used a hybrid algorithm to optimize the weighting factor of unvisited URL priority. The experimental results show that the proposed BSVSM-HA crawler can obtain better evaluation indicators compared with the other three crawlers including Word2vec crawler, ELMO crawler and BSVSM crawler. In conclusion, the semantic and intelligent crawler proposed in this paper makes the semantic similarity between terms more accurate, and improves the topic relevance of the text, and the optimized weighting factor makes the priority evaluation of unvisited URLs more accurate. © 2025 Institute of Electrical and Electronics Engineers Inc.. All rights reserved.

关键词： Vectors

来源：评论

学校读者我要写书评

暂无评论

Exploratory Data Analysis and Predicting Indian Crime Against Women Using Machine Learning Approaches

Exploratory Data Analysis and Predicting Indian Crime Agains...

引用

Interdisciplinary Approaches in technology and Management for Social Innovation (IATMSI), IEEE Conference on

作者： N. Srinivasa Gupta B. Valarmathi K. Santhi T.Chellatamilan Sameeksha Nanda Department of Manufacturing School of Mechanical Engineering Vellore Institute of Technology Vellore Tamil Nadu India Department of Software and Systems Engineering School of Computer Science Engineering and Information Systems Vellore Institute of Technology Vellore Tamil Nadu India Department of Analytics School of Computer Science and Engineering Vellore Institute of Technology Vellore Tamil Nadu India Department of Information Technology School of Computer Science Engineering and Information Systems Vellore Institute of Technology Vellore Tamil Nadu India

ISBN: (数字)9798331521691

ISBN: (纸本)9798331521707

Crime forecasting is a critical endeavor aimed at preventing future criminal actions, specifically the rising rate of violence against women in India. This study proposes the model to predict crimes committed against women using data from the National Crime Records Bureau's (NCRB) monthly crime report. The dataset encompasses district and state-wise crime activities from 2001 to 2014. The proposed work objectives are identifying factors influencing these crimes and developing predictive models using machine learning (ML). The study employs Exploratory Data Analysis (EDA), to enhance the predictive power of the models. The suggested study uses XGBoost, Random Forest (RF), and Neural Network (NN). The neural network algorithm outperforms the Random Forest and XGBoost techniques among these three. The Random Forest algorithm has the worst performance. After comparing the accuracies of various methods, such as XGBoost (92%), and Random Forest (89.84%) classifiers using imbalanced datasets (without employing Synthetic Minority Oversampling Technique (SMOTE) technique), it was found that the neural network algorithm (95.81%) delivers the highest accuracy. This study looks at model accuracies and finds that, after applying the SMOTE algorithm, neural network algorithms performed best, with an accuracy of 96.76%, outperforming the accuracies of the remaining models, which included the Random Forest (93.08%) and XG-Boost (95%).

关键词： Radio frequency Technological innovation Accuracy Machine learning algorithms Data analysis Predictive models Prediction algorithms Data models Classification algorithms Random forests

来源：评论

学校读者我要写书评

暂无评论

QCACNN: A Quantum Convolutional Neural Network Algorithm for Traffic Sign Recognition in Carbon-Intelligent Electric Vehicles

引用

IEEE Internet of Things Journal 2025年

作者： Qu, Zhiguo Xia, Yichen Sun, Le Liu, Wenjie Muhammad, Ghulam Nanjing University of Information Science and Technology Engineering Research Center of Digital Forensics Ministry of Education School of Computer Science Nanjing210044 China Nanjing University of Information Science and Technology School of Computer Science Nanjing210044 China Nanjing University of Information Science and Technology School of Software Nanjing210044 China King Saud University College of Computer and Information Sciences Department of Computer Engineering Riyadh Saudi Arabia

Traffic sign recognition is essential for autonomous driving, enhancing driving efficiency and ensuring road safety. Traffic signs utilize colors and patterns to relay critical information, with color accuracy being especially significant. Traditional models, reliant on extensive data and computing resources, struggle to meet the efficiency demands of electric vehicles with carbon-intelligent computing. Leveraging the physical properties of quantum superposition states, quantum computing offers a solution with its unique parallel computing capabilities, potentially enhancing recognition efficiency and facilitating real-time processing. Quantum Convolutional Neural Networks (QCNNs) show promise in processing large-scale image data with improved efficiency and accuracy. However, most of QCNN researches focus on grayscale image classification, with limited studies on multi-channel data. This paper introduces the Quantum Channel Attention Convolutional Neural Network (QCACNN), which incorporates a Quantum Channel Attention Layer (QCAL) to enhance multi-channel data classification. The experimental results demonstrate that QCACNNs surpasses traditional QCNNs in traffic sign recognition, performing comparably to conventional CNNs and SENet models with fewer computing resources. Detailed performance analysis and ablation studies validate the effectiveness of each component within this architecture. Quantum noise resistance tests confirm the robustness of QCACNN, making it a viable solution for electric vehicles with enhanced scalability. © 2014 IEEE.

关键词： Quantum channel

来源：评论

学校读者我要写书评

暂无评论

CAPAST: Content Affinity Preserved Arbitrary Style Transfer

CAPAST: Content Affinity Preserved Arbitrary Style Transfer

引用

International Conference on Acoustics, Speech, and Signal Processing (ICASSP)

作者： Xinyuan Zheng Xiaojie Li Canghong Shi Jia He Zhan Ao Huang Xian Zhang Imran Mumtaz School of Computer Science Chengdu University of Information Technology China School of Computer and Software Engineering Xihua University China Faculty of Engineering and Mathematical Sciences The University of Western Australia Australia Department of Computer Science University of Agriculture Faisalabad Pakistan

ISBN: (数字)9798350368741

ISBN: (纸本)9798350368758

Balancing the consistency of style and the integrity of content is the main challenge in arbitrary style transfer domain. Currently, local style details can be effectively captured by attention mechanism but easily produce distorted style patterns and inconsistent content structure. In this paper, we propose a Content Affinity Preserving Arbitrary Style Transfer (CAPAST) framework to ensure style features can be stably integrated into the content structure. Considering the local feature learning ability of CNN and the global feature representation advantage of transformer, a dual encoder is proposed to capture local and global features of images with the combination between transformer and CNN. In addition, a channel and spatially aligned attention (CSAA) is introduced to generate high-quality results by stably fusing style features and content features. In experiments, we demonstrated the superior performance of our method in preventing content structure distortion and maintaining consistency between style and content. Codes are available at https://***/miaopashi-zxy/CAPAST.

关键词： Representation learning Acoustic distortion Codes Attention mechanisms Fuses Transformers Acoustics Speech processing

来源：评论

学校读者我要写书评

暂无评论

STR: Spatio-temporal trajectory representation learning with dual-focus encoder for whole trajectory similarity computation

引用

Information Fusion 2025年 123卷

作者： Li, Mengqiu Niu, Xinzheng Zhu, Jiahui Fournier-Viger, Philippe Wu, Youxi School of Computer Science and Engineering University of Electronic Science and Technology of China Chengdu China school of Information and Software Engineering University of Electronic Science and Technology of China Chengdu China College of Computer Science and Software Engineering Shenzhen University Shenzhen China School of Artificial Intelligence Hebei University of Technology Tianjin China

In the field of trajectory data mining, trajectory similarity computation is a key issue associated with trajectory representation learning. Current methods of trajectory similarity computation often represent trajectory data of varying lengths as vectors with uniform dimension. However, existing trajectory representations fall short of comprehensively calculating trajectory similarity: first, they focus solely on spatial information, while overlooking the temporal information of the trajectory, and second, they represent the temporal and spatial information separately. To address these limitations, we propose a novel Spatio-temporal Trajectory Representation method, named STR, aimed at capturing both the features of individual trajectories and the relationships between them in a three-dimensional spatio-temporal region from the perspectives of both trajectory points and whole trajectories, and improving the accuracy of trajectory similarity metrics. STR consists of two main modules: (i) a trajectory feature tokenization module, which considers the uneven density distribution of trajectory points in the three-dimensional space to build a hierarchical structure that considers space and time as a whole, and extracts spatio-temporal features to construct a token sequence;(ii) a dual-focus trajectory encoder learns the features of a given trajectory and those of similar trajectories, thereby enabling the consideration of relationships between whole trajectories. Moreover, a density-based whole trajectory similarity metric is designed to find similar trajectories. Experiments were conducted on three public datasets, and the accuracy of the trajectory representations obtained with STR is significantly greater than for existing state-of-the-art baselines on a variety of distance metrics. © 2025 Elsevier B.V.

关键词： Dual-focus trajectory encoder Representation learning Spatio-temporal trajectory Trajectory similarity

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：