检索结果-内蒙古大学图书馆

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

IEEE Transactions on Multimedia 2025年 27卷 3411-3425页

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 650051 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

M-Tuning: Prompt Tuning With Mitigated Label Bias in Open-Set Scenarios

引用

IEEE Transactions on Circuits and Systems for Video technology 2025年第6期35卷 5885-5899页

作者： Liao, Ning Zhang, Xiaopeng Cao, Min Yan, Junchi Shanghai Jiao Tong University MoE Key Laboratory of Artificial Intelligence Shanghai200240 China Huawei Inc. Guangdong Shenzhen518129 China Soochow University School of Computer Science and Technology Suzhou215006 China Shanghai Jiao Tong University School of Artificial Intelligence Shanghai200240 China

In realistic open-set scenarios where labels of a part of testing data are totally unknown, when vision-language (VL) prompt learning methods encounter inputs related to unknown classes (i.e., not seen during training), they always predict them as one of the training classes. The exhibited label bias causes difficulty in open set recognition (OSR), in which an image should be correctly predicted as one of the known classes or the unknown one. To achieve this goal, we propose a vision-language prompt tuning method with mitigated label bias (M-Tuning). It introduces open words from the WordNet to extend the range of words forming the prompt texts from only closed-set label words to more, and thus prompts are tuned in a simulated open-set scenario. Besides, inspired by the observation that classifying directly on large datasets causes a much higher false positive rate than on small datasets, we propose a Combinatorial Tuning and Testing (CTT) strategy for improving performance. CTT decomposes M-Tuning on large datasets as multiple independent group-wise tuning on fewer classes, then makes accurate and comprehensive predictions by selecting the optimal sub-prompt. Finally, given the lack of VL-based OSR baselines in the literature, especially for prompt methods, we contribute new baselines for fair comparisons. Our method achieves the best performance on datasets with various scales, and extensive ablation studies also validate its effectiveness. © 1991-2012 IEEE.

关键词： Training Tuning Testing Circuits and systems Accuracy Target recognition Robustness Learning systems Image recognition Visualization

来源：评论

学校读者我要写书评

暂无评论

Skin Lesion Detection and Classification from Dermoscopic Images Using a Hybrid Network 15th

Skin Lesion Detection and Classification from Dermoscopic Im...

引用

15th International Conference on Soft Computing and Pattern Recognition, SoCPaR 2023 and 14th World Congress on Nature and Biologically Inspired Computing, NaBIC 2023

作者： Kumar, Naween Murmu, Anita Dash, Yajnaseni Abraham, Ajith School of Computer Science Engineering and Technology Bennett University UP Greater Noida India School of Artificial Intelligence Bennett University UP Greater Noida India School of Artificial Intelligence Sai University Chennai India

ISBN: (纸本)9783031810794

Classification of skin abnormalities in dermatology plays an important role in the detection and diagnosis of various skin diseases such as melanoma. However, accurately classifying skin cancer remains difficult due to imbalanced classes, variances within classes, lack of training data, and failure to focus on semantically significant lesion parts. To address these issues, the proposed novel Deep-Convolutional Network (DCN) with a Machine-Learning (ML)-based SVM technique has been used to categorize skin cancer. To prevent overfitting, the DCN-SVM method is used as a classifier. Furthermore, the classification model applies the expert framework for learning based on domain functions. The effectiveness of test results using a UNet-MobileNet with 10,000 training images (HAM10000) dataset in seven distinct human-machine classes. The results demonstrate the effectiveness of this method in classifying dermatological diseases, demonstrating its potential to improve dermatological care and diagnostic accuracy. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

An identification and analysis method on the abnormal electricity usage behavior of urban/building energy systems based on statistical method and domain knowledge

引用

Building Simulation 2025年第4期18卷 863-879页

作者： Ying Zhang Manjia Liu Zaixun Ling Wenjie Gang Xiuxia Hao School of Environmental Science&Engineering Huazhong University of Science and TechnologyWuhan430074China Institute of Artificial Intelligence Huazhong University of Science and TechnologyWuhan430074China State Grid Hubei Electric Power Research Institute Wuhan430015China Hubei Key Laboratory of Multi-media Pollution Cooperative Control in Yangtze Basin School of Environmental Science&EngineeringHuazhong University of Science and TechnologyWuhan430074China

Understanding the abnormal electricity usage behavior of buildings is essential to enhance the resilience,efficiency,and security of urban/building energy systems while safeguarding occupant ***,data reflecting such behavior are often considered as outliers,and removed or smoothed during preprocessing,limiting insights into their potential *** paper proposes an abnormal behavior analysis method that identifies outliers(considering data distribution)and anomalies(considering the physical context)based on the statistical principle and domain knowledge,assessing their effects on energy supply security.A 4-quadrant graph is proposed to quantify and categorize the impacts of buildings on urban energy *** method is illustrated by data from 1,451 buildings in a *** show that the proposed method can identify abnormal data *** in the primary industry have more outliers,while those in the tertiary industry have more *** buildings affecting both the security and economy of urban energy systems are *** outliers rise more frequently from 8:00 to 18:00,on weekdays and in the summer and winter ***,the anomaly distribution has a weak connection with ***,the abnormal electricity usage behavior positively correlates with outdoor air *** method provides a new perspective for identifying potential risks,managing energy usage behavior,and enhancing flexibility of the urban energy systems.

关键词： resilience urban energy systems efficiency statistical method abnormal electricity usage behavior abnormal behavior analysis domain knowledge abnormal behavior analysis method

来源：评论

学校读者我要写书评

暂无评论

Combining KNN with AutoEncoder for Outlier Detection

引用

Journal of computer science & technology 2024年第5期39卷 1153-1166页

作者： Shu-Zheng Liu Shuai Ma Han-Qing Chen Li-Zhen Cui Jie Ding State Key Laboratory of Software Development Environment Beihang UniversityBeijing 100191China School of Software Shandong UniversityJinan 250100China Joint SDU-NTU Centre for Artificial Intelligence Research Shandong UniversityJinan 250100China School of Computer Science Jiangsu University of Science and TechnologyZhenjiang 212003China

K-nearest neighbor(KNN)is one of the most fundamental methods for unsupervised outlier detection because of its various advantages,e.g.,ease of use and relatively high ***,most data analytic tasks need to deal with high-dimensional data,and the KNN-based methods often fail due to“the curse of dimensionality”.AutoEncoder-based methods have recently been introduced to use reconstruction errors for outlier detection on high-dimensional data,but the direct use of AutoEncoder typically does not preserve the data proximity relationships well for outlier *** this study,we propose to combine KNN with AutoEncoder for outlier ***,we propose the Nearest Neighbor AutoEncoder(NNAE)by persevering the original data proximity in a much lower dimension that is more suitable for performing ***,we propose the K-nearest reconstruction neighbors(K NRNs)by incorporating the reconstruction errors of NNAE with the K-distances of KNN to detect ***,we develop a method to automatically choose better parameters for optimizing the structure of ***,using five real-world datasets,we experimentally show that our proposed approach NNAE+K NRN is much better than existing methods,i.e.,KNN,Isolation Forest,a traditional AutoEncoder using reconstruction errors(AutoEncoder-RE),and Robust AutoEncoder.

关键词： outlier detection AutoEncoder K-nearest neighbor(KNN) unsupervised learning

来源：评论

学校读者我要写书评

暂无评论

Migrant Resettlement by Evolutionary Multiobjective Optimization

IEEE Transactions on Artificial Intelligence

引用

IEEE Transactions on artificial intelligence 2025年第1期6卷 51-65页

作者： Liu, Dan-Xuan Gu, Yu-Ran Qian, Chao Mu, Xin Tang, Ke Nanjing University National Key Laboratory for Novel Software Technology School of Artificial Intelligence Nanjing210023 China Peng Cheng Laboratory Shenzhen518000 China Southern University of Science and Technology Department of Computer Science and Engineering Shenzhen518055 China

Migration has been a universal phenomenon, which brings opportunities as well as challenges for global development. As the number of migrants (e.g., refugees) increases rapidly, a key challenge faced by each country is the problem of migrant resettlement. This problem has attracted scientific research attention, from the perspective of maximizing the employment rate. Previous works mainly formulated migrant resettlement as an approximately submodular optimization problem subject to multiple matroid constraints and employed the greedy algorithm, whose performance, however, may be limited due to its greedy nature. In this article, we propose a new framework called migrant resettlement by evolutionary multiobjective optimization (MR-EMO), which reformulates migrant resettlement as a biobjective optimization problem that maximizes the expected number of employed migrants and minimizes the number of dispatched migrants simultaneously, and employs a multiobjective evolutionary algorithm (MOEA) to solve the biobjective problem. We implement MR-EMO using three MOEAs: the popular nondominated sorting genetic algorithm II (NSGA-II), MOEA based on decomposition (MOEA/D) as well as the theoretically grounded global simple evolutionary multiobjective optimizer (GSEMO). To further improve the performance of MR-EMO, we propose a specific MOEA, called GSEMO using matrix-swap mutation and repair mechanism (GSEMO-SR), which has a better ability to search for feasible solutions. We prove that MR-EMO using either GSEMO or GSEMO-SR can achieve better theoretical guarantees than the previous greedy algorithm. Experimental results under the interview and coordination migration models clearly show the superiority of MR-EMO (with either NSGA-II, MOEA/D, GSEMO or GSEMO-SR) over previous algorithms, and that using GSEMO-SR leads to the best performance of MR-EMO. © 2024 IEEE.

关键词： Multiobjective optimization

来源：评论

学校读者我要写书评

暂无评论

Asymmetric interaction preference induces cooperation in human–agent hybrid game

引用

science China(Information sciences) 2025年

作者： Danyang JIA Xiangfeng DAI Junliang XING Pin TAO Yuanchun SHI Zhen WANG Department of Computer Science and Technology Tsinghua University School of Artificial Intelligence Optics and Electronics (iOPEN)Northwestern Polytechnical University

With the advancement of artificial intelligence, human interest in human–agent collaboration has grown, raising a series of challenges regarding the relationship between agents and humans, such as trust and cooperation. This leads to the inevitable consideration of the inherent human traits of subjective interaction preferences for different groups, particularly in human–agent hybrid systems where human–human, agent–agent, and human–agent interactions coexist. However,understanding how individual interaction preferences influence cooperation within such systems remains a major *** address this, this study proposes a human–agent hybrid prisoner's dilemma game system within the framework of evolutionary game theory. In spatial networks, the primary distinction between agents and humans lies in their decision-making flexibility: humans possess higher adaptive capabilities, follow link dynamics, and employ free decision-making rules, which allows them to select different strategies for different neighbors. In contrast, agents follow node dynamics, applying uniform decision rules and using the same strategy across all neighbors. We define subjective preferences for individuals in various groups, including interaction preferences between homogeneous and heterogeneous groups. The simulation results demonstrate that both humans and agents display asymmetric interaction preferences toward groups with different identities,which significantly enhances cooperative behavior in the system. In the hybrid system, human groups exhibit more stable prosocial behavior, whereas agent groups form highly cooperative clusters when there is a strong interaction preference for human groups. Additionally, endowing agents with the ability to identify their opponents effectively mitigates the interaction dilemma among agents.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Cloud Service Providers: Evaluating Scalability, Security, and Cost for Business Growth 15th

Cloud Service Providers: Evaluating Scalability, Security, a...

引用

15th International Conference on Soft Computing and Pattern Recognition, SoCPaR 2023 and 14th World Congress on Nature and Biologically Inspired Computing, NaBIC 2023

作者： Kumar, Naween Dash, Yajnaseni Abraham, Ajith Saini, Vaibhav School of Computer Science Engineering and Technology Bennett University UP Greater Noida India School of Artificial Intelligence Bennett University UP Greater Noida India School of Artificial Intelligence Sai University Chennai India

ISBN: (纸本)9783031810855

Cloud computing brought a radical change in information technology by enabling infinite growth in terms of scalability and flexibility and by making new technologies available and easily accessible. Although the identification of Cloud Service Providers (CSPs) and their services is straightforward, there are difficulties for clients when they attempt to traverse the immense market share of CSPs. This paper empowers users by abstracting the leading CSPs such as Amazon, Microsoft, and Google, by developing a comparative view under the influence of various service parameters. When the features and offerings of various cloud providers are known, enabling the business to determine its suitability for the cloud, informed decisions are made. Assessing the range of services that each vendor provides and classifying them based on the type of services offered can be either under the IaaS, PaaS, or SaaS. It emphasizes evaluating key aspects of versatility to accommodate the business expansion, highly secure protection that meets legislative requirements, and affordable rates for expenditures. Based on this analysis, the paper recommends that businesses should make the right choices to work with the right cloud partner so it should promote a lot of growth and development. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Cloud Computing CSPs Expansion Flexibility Growth PaaS SaaS Security Service Delivery Models such as IaaS

来源：评论

学校读者我要写书评

暂无评论

Waste Classification Using Deep Neural Networks 15th

Waste Classification Using Deep Neural Networks

引用

15th International Conference on Soft Computing and Pattern Recognition, SoCPaR 2023 and 14th World Congress on Nature and Biologically Inspired Computing, NaBIC 2023

作者： Kumar, Naween Dash, Yajnaseni Abraham, Ajith Nayak, Akanksha Mehta, Ansh Tandon, Kushagra School of Computer Science Engineering and Technology Bennett University UP Greater Noida India School of Artificial Intelligence Bennett University UP Greater Noida India School of Artificial Intelligence Sai University Chennai India

ISBN: (纸本)9783031810824

Waste management has been rapidly increasing worldwide lately due to the skyrocketing waste production and environmental and health-related issues. However, solving this issue using conventional approaches is not feasible. Various efforts are being made to make it more accessible through better technology, including automation. In this research, the Deep Neural Networks (DNN) techniques are utilized to classify waste with the help of image data. The proposed method is tested using multiple experiments for different waste classification tasks, including separating recyclables and organics from the waste. Different models, such as CNN, InceptionV3, ResNet50, and VGG16 are trained for the classification task. The presented work contributes to understanding the advancement of technology use in one of our generation’s most critical challenges – waste management. The obtained findings indicate an effective utilization of DNNs to acquire the necessary knowledge to enhance our future with more sustainable technologies. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Waste treatment

来源：评论

学校读者我要写书评

暂无评论

Enhancing Respiratory Monitoring by CNN Using Mel Frequency Cepstral Coefficients 15th

Enhancing Respiratory Monitoring by CNN Using Mel Frequency ...

引用

15th International Conference on Soft Computing and Pattern Recognition, SoCPaR 2023 and 14th World Congress on Nature and Biologically Inspired Computing, NaBIC 2023

作者： Dash, Yajnaseni Abraham, Ajith Gupta, Shivam Rathore, Shaurya Vardhan Patil, Harsh School of Artificial Intelligence Bennett University UP Greater Noida India School of Computer Science Engineering and Technology Bennett University UP Greater Noida India School of Artificial Intelligence Sai University Chennai India

ISBN: (纸本)9783031810794

This study aims to provide a comprehensive analysis of the distribution of respiratory sounds using a large number of recordings. The data set that we used includes normal breathing sounds, wheezing sounds, and coughing sounds, all of which indicate a different breathing pattern. The method involved first removing audio data and making the data consistent, then extracting features using Mel Frequency Cepstral Coefficients (MFCC), Based on these features, breath sounds are classified using a convolutional neural networks (CNN) model. The CNN model has shown high accuracy in distinguishing all the various respiratory sounds, showing and evaluating that it could be useful in diagnosis. This work has demonstrated how well CNN can identify breath sounds and suggests ways to develop automatic, non-invasive respiratory monitoring tools to help in the field of respiratory problems. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Convolutional neural networks

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：