检索结果-内蒙古大学图书馆

31st International Conference on Computational Linguistics, COLING 2025

作者： Yang, Li Xiao, Zhiding Huang, Wenxin Zhong, Xian Hubei Key Laboratory of Big Data Intelligent Analysis and Application School of Computer Science and Information Engineering Hubei University China Hubei Key Laboratory of Transportation Internet of Things School of Computer Science and Artificial Intelligence Wuhan University of Technology China

ISBN: (纸本)9798891761964

The rapid development of multi-modal large language models (MLLMs) has positioned visual storytelling as a crucial area in content creation. However, existing models often struggle to maintain temporal, spatial, and narrative coherence across image sequences and frequently lack the depth and engagement of human-authored stories. To address these challenges, we propose Story with Large Language and Vision Assistant (StoryLLaVA), a novel framework for enhancing visual storytelling. Our approach introduces a Topic-Driven Narrative Optimizer (TDNO) that improves both the training data and MLLM models by integrating image descriptions, topic generation, and GPT-4-based refinements. Furthermore, we employ a preference-based ranked story sampling method that aligns model outputs with human storytelling preferences through positive-negative pairing. These two phases of the framework differ in their training methods: the former uses supervised fine-tuning, while the latter incorporates reinforcement learning with positive and negative sample pairs. Experimental results demonstrate that StoryLLaVA outperforms current models in visual relevance, coherence, and fluency, with LLM-based evaluations confirming the generation of richer and more engaging narratives. The enhanced dataset and model are available at https://***/XxxZzD/StoryLLaVA. © 2025 Association for Computational Linguistics.

关键词：

来源：评论

学校读者我要写书评

暂无评论

A Fine-Grained Network for Joint Multimodal Entity-Relation Extraction

引用

IEEE TRANSACTIONS ON KNOWLEDGE AND data ENGINEERING 2025年第1期37卷 1-14页

作者： Yuan, Li Cai, Yi Xu, Jingyu Li, Qing Wang, Tao South China Univ Technol Minist Educ Sch Software Engn Key Lab Big Data & Intelligent Robot Guangzhou 510641 Peoples R China Hong Kong Polytech Univ Dept Comp Hong Kong Peoples R China Kings Coll London Dept Biostat & Hlth Informat London WC2R 2LS England

Joint multimodal entity-relation extraction (JMERE) is a challenging task that involves two joint subtasks, i.e., named entity recognition and relation extraction, from multimodal data such as text sentences with associated images. Previous JMERE methods have primarily employed 1) pipeline models, which apply pre-trained unimodal models separately and ignore the interaction between tasks, or 2) word-pair relation tagging methods, which neglect neighboring word pairs. To address these limitations, we propose a fine-grained network for JMERE. Specifically, we introduce a fine-grained alignment module that utilizes a phrase-patch to establish connections between text phrases and visual objects. This module can learn consistent multimodal representations from multimodal data. Furthermore, we address the task-irrelevant image information issue by proposing a gate fusion module, which mitigates the impact of image noise and ensures a balanced representation between image objects and text representations. Furthermore, we design a multi-word decoder that enables ensemble prediction of tags for each word pair. This approach leverages the predicted results of neighboring word pairs, improving the ability to extract multi-word entities. Evaluation results from a series of experiments demonstrate the superiority of our proposed model over state-of-the-art models in JMERE.

关键词： Visualization data mining Contrastive learning Feature extraction Brain modeling Tagging Semantics Predictive models Pipelines Logic gates entity-relation extraction multimodal pre-trained model

来源：评论

学校读者我要写书评

暂无评论

A Knowledge-Enhanced Contrastive Learning Network for Weakly-Supervised Aspect Detection

IEEE Transactions on Audio, Speech and Language Processing

引用

IEEE Transactions on Audio, Speech and Language Processing 2025年 33卷 506-517页

作者： Zhuoming Zheng Yi Cai Liuwu Li School of Software Engineering South China University of Technology Guangzhou China Key Laboratory of Big Data and Intelligent Robot Ministry of Education South China University of Technology Guangzhou China

Aspect detection, aiming at identifying the aspects of review segments, is a fundamental task in opinion mining and aspect-based sentiment analysis. Due to the high cost and time consuming of human-annotation for massive reviews, several unsupervised and weakly-supervised methods are proposed recently. However, existing weakly-supervised models are mostly seed-driven methods based on co-occurrence of words, which suffer from lacking the ability of detecting the aspects with infrequent aspect terms and identifying Misc aspect. To tackle these problems, we leverage external knowledge to enhance the representation of aspects and segments by a weakly-supervised method. In this paper, we propose an aspect knowledge-enhanced contrastive learning (AKECL) network with two powerful knowledge-enhanced encoders for aspects and reivew segments to enhance weakly-supervised aspect detection task. Experiments in seven different domains show that AKECL outperforms the competitive baselines, and demonstrate the effectiveness of our proposed method, as well as the improvement by introducing external knowledge.

关键词： Reviews Contrastive learning Vectors Feature extraction Knowledge graphs Speech processing Sentiment analysis Knowledge engineering Hardware Touch sensitive screens

来源：评论

学校读者我要写书评

暂无评论

RGMEF: a lightweight multi-exposure image fusion network based on retinex and gradient guidance

引用

MACHINE VISION AND APPLICATIONS 2025年第3期36卷 1-18页

作者： Feng, Yunfei Zhang, Mei Zhu, Jinhui South China Univ Technol Sch Automat Sci & Engn Guangzhou 510641 Peoples R China South China Univ Technol Sch Software Engn Guangzhou 510641 Peoples R China South China Univ Technol Minist Educ Key Lab Big Data & Intelligent Robot Guangzhou 510641 Peoples R China

The goal of the multi-exposure image fusion (MEF) is to create a high dynamic range (HDR) image with balanced exposures and more details by extracting and enhancing valid information from multiple low dynamic range images with varied exposures. In recent years, deep learning has provided many solutions for MEF, and they have achieved remarkable effects. However, because of the high network complexity, these approaches often have low efficiency and limit the potential improvement of the effect. In order to fuse valuable information from different images, we introduce a lightweight but robust MEF network, which focus on the image internal characteristics rather than long-distance information. In this study, we first design a decomposed network inspired by retinex theory to extract the illumination map from an image and obtain the gradient map through Sobel operator. Subsequently, we introduce two lightweight feature extraction modules to extract image features with the guidance of illumination and gradient maps respectively, and then fuse these features to produce an HDR result through three fusion blocks. Furthermore, we design a loss function to obtain a more suitable brightness range for the current scene. Numerous comparative experiments demonstrate the superiority of our method in terms of image texture, pixel intensity, and color retention. Moreover, some ablation studies indicate that our approach has a significant adaptability to source images.

关键词： Multi-exposure fusion Image guidance High dynamic range Retinex theory

来源：评论

学校读者我要写书评

暂无评论

Device-edge collaborative occluded face recognition method based on cross-domain feature fusion

引用

Digital Communications and Networks 2025年第2期11卷 482-492页

作者： Zhang, Puning Tan, Lei Yang, Zhigang Huang, Fengyi Sun, Lijun Peng, Haiying School of Communications and Information Engineering Chongqing University of Posts and Telecommunications Chongqing400065 China Advanced Network and Intelligent Connection Technology Key Laboratory of Chongqing Education Commission of China Chongqing400065 China Chongqing Key Laboratory of Ubiquitous Sensing and Networking Chongqing400065 China Chongqing Innovation Center of Industrial Big-Data Co. Lt Chongqing400700 China

The lack of facial features caused by wearing masks degrades the performance of facial recognition systems. Traditional occluded face recognition methods cannot integrate the computational resources of the edge layer and the device layer. Besides, previous research fails to consider the facial characteristics including occluded and unoccluded parts. To solve the above problems, we put forward a device-edge collaborative occluded face recognition method based on cross-domain feature fusion. Specifically, the device-edge collaborative face recognition architecture gets the utmost out of maximizes device and edge resources for real-time occluded face recognition. Then, a cross-domain facial feature fusion method is presented which combines both the explicit domain and the implicit domain facial. Furthermore, a delay-optimized edge recognition task scheduling method is developed that comprehensively considers the task load, computational power, bandwidth, and delay tolerance constraints of the edge. This method can dynamically schedule face recognition tasks and minimize recognition delay while ensuring recognition accuracy. The experimental results show that the proposed method achieves an average gain of about 21% in recognition latency, while the accuracy of the face recognition task is basically the same compared to the baseline method. © 2024

关键词： Face recognition

来源：评论

学校读者我要写书评

暂无评论

GoPIM: GCN-Oriented Pipeline Optimization for PIM Accelerators 31

GoPIM: GCN-Oriented Pipeline Optimization for PIM Accelerato...

引用

31st IEEE International Symposium on High Performance Computer Architecture, HPCA 2025

作者： Yang, Siling He, Shuibing Wang, Wenjiong Yin, Yanlong Wu, Tong Chen, Weijian Zhang, Xuechen Sun, Xian-He Feng, Dan The State Key Laboratory of Blockchain and Data Security Zhejiang University China Zhejiang Lab China Institute of Blockchain and Data Security China Zhejiang Key Laboratory of Big Data Intelligent Computing China Washington State University Vancouver United States Illinois Institute of Technology United States Huazhong University of Science and Technology China Wuhan National Laboratory for Optoelectronics China

ISBN: (纸本)9798331506476

Graph convolutional networks (GCNs) are popular for a variety of graph learning tasks. ReRAM-based processing-in-memory (PIM) accelerators are promising to expedite GCN training owing to their in-situ computing capability. However, existing accelerators can be severely underutilized even with pipelines, due to the oversight of the skewed execution times of various GCN stages and the ignorance of skewed degrees of graph vertices. In this work, we propose GOPIM, a GCN-oriented pipeline optimization for PIM accelerators to expedite GCN training. First, GOPIM proposes an ML-based scheme that allocates crossbar resources to the most needed stages to streamline the overall pipeline. Second, GOPIM utilizes a selective vertex updating technique that evenly distributes vertices on crossbars by interleaved mapping. These techniques collectively reduce the overall execution time without losing much accuracy. We also provide a practical architecture design for GOPIM. Our experimental results show that, GoPIM achieves up to 191 × speedup and 16.1 × energy saving, compared to the state-of-the-art work. © 2025 IEEE.

关键词： Graph neural networks

来源：评论

学校读者我要写书评

暂无评论

Numerical study of microorganisms swimming near a convex wall in a Giesekus fluid

引用

Physical Review E 2025年第1期111卷 015103-015103页

作者： Chenlin Zhu Fangyuan Peng Dingyi Pan Zhaosheng Yu Zhaowu Lin Key Laboratory of Intelligent Manufacturing Quality Big Data Tracing and Analysis of Zhejiang Province China Jiliang University Hangzhou 310018 China

The motion of microorganisms in complex fluids stands out as a prominent subject within fluid mechanics. In our study, we utilize the fictitious domain method to investigate the locomotion of squirmers along a convex wall in Giesekus viscoelastic fluids. This study examines the influence of fluid elasticity and wall curvature on squirmer particles, analyzing their movement patterns in detail. Near the convex wall, three distinct behavioral characteristics emerge: scattering, orbiting forward, and orbiting backward. The findings reveal that, compared with Newtonian fluids, squirmers exhibit a stronger tendency to be attracted toward the wall in viscoelastic fluids. This behavior is attributed to the elastic stress of the fluid, which generates a reverse torque on microbial particles, altering their movement direction and hindering their escape from the wall. Notably, as the wall curvature decreases, the likelihood of particles escaping diminishes.

关键词： Forward scattering

来源：评论

学校读者我要写书评

暂无评论

Multi-Quadrotor Cooperative Encirclement and Capture Approach in Obstacle Environments

Multi-Quadrotor Cooperative Encirclement and Capture Approac...

引用

2025 IEEE International Conference on Mechatronics, ICM 2025

作者： Li, Zhenyu Wu, Ruixin Wang, Jin Hu, Chuanqi Zheng, Zhi Meng, Jun Lu, Guodong Zhejiang University State Key Laboratory of Fluid Power and Mechatronic Systems Hangzhou310058 China Zhejiang University Zhejiang Key Laboratory of Industrial Big Data and Robot Intelligent Systems Hangzhou310058 China Robotics Research Center of Yuyao City Ningbo315400 China College of Electrical Engineering Zhejiang University Hangzhou310027 China Center for Data Mining and Systems Biology College of Electrical Engineering Zhejiang University Hangzhou310027 China

ISBN: (纸本)9798331533892

This article proposes an algorithm designed for autonomous quadrotors to capture target quadrotors in complex obstacle environments, enabling multiple pursuers to perform distributed capture while simultaneously avoiding obstacles and collisions. We first introduce a 3D obstacle space representation method based on safety-priority reachable region (SPRR) and implement real-time construction for general point cloud maps. Subsequently, leveraging SPRR, we employ an area minimization approach to ensure the effective capture of the evader within a limited time. The outcomes are then integrated into a MINCO-based trajectory planning framework to ensure their practical execution by real quadrotors. Relying on SPRR, the optimization problem is simplified and the solution efficiency is improved. In the scenario of simulating the dynamics of real quadrotors, we compare the proposed algorithm with state-of-the-art evasion and pursuit methods, validating its effectiveness and advancements. This work presents a practical solution for multi-quadrotor collaborative capture. © 2025 IEEE.

关键词： Motion planning

来源：评论

学校读者我要写书评

暂无评论

Enhanced Fabric Defect Detection With Feature Contrast Interference Suppression

引用

IEEE TRANSACTIONS ON INSTRUMENTATION AND MEASUREMENT 2025年 74卷

作者： Wang, Jinbao Cheng, Jiayi Gao, Can Zhou, Jie Shen, Linlin Shenzhen Univ Natl Engn Lab Big Data Syst Comp Technol Shenzhen 518060 Peoples R China Shenzhen Univ Coll Comp Sci & Software Engn Shenzhen Peoples R China Shenzhen Univ Shenzhen Inst Artificial Intelligence & Robot Soc Shenzhen 518060 Peoples R China Guangdong Key Lab Intelligent Informat Proc Shenzhen 518060 Peoples R China

Detecting fabric defects is a critical task in the textile industry, as it involves identifying and localizing imperfections in fabrics. However, many existing unsupervised defect detection methods are sensitive to variations and disturbances in the normal background of fabric images, leading to a high false alarm rate and unsatisfactory performance. To address this issue, we propose a novel defect detection method called feature contrast interference suppression (FCIS) that aims to enlarge the feature distance between foreground and background textures while suppressing the distribution of background features to obtain a more robust background representation. Our approach consists of several key components: First, we introduce background-foreground contrastive learning (BFCL) to enhance feature representation between background and foreground elements at the feature level. Second, we use multiscale denoising knowledge distillation to aid the model in learning more accurate representations of normal background features. Finally, we design an attention decoder that refines and segments the defect mask generated through feature comparison, resulting in improved detection accuracy. Our proposed FCIS has been validated using five mainstream datasets. Experimental results demonstrate that our proposed method outperforms existing approaches, achieving improvements of 2.38% in image-level AUROC and 3.73% in pixel-level AUPRO, along with a 9.12% enhancement in average defect localization precision.

关键词： Feature extraction Fabrics Defect detection Image reconstruction Training Accuracy Noise reduction data models Decoding Attention segmentation fabric defect detection interference suppression knowledge distillation unsupervised contrastive learning unsupervised contrastive learning

来源：评论

学校读者我要写书评

暂无评论

Visual comparative analytics of multimodal transportation

引用

VISUAL INFORMATICS 2025年第1期9卷 18-30页

作者： Deng, Zikun Chen, Haoming Lu, Qing-Long Su, Zicheng Schreck, Tobias Bao, Jie Cai, Yi South China Univ Technol Sch Software Engn Guangzhou Peoples R China South China Univ Technol Key Lab Big Data & Intelligent Robot Guangzhou Peoples R China Natl Univ Singapore Dept Civil & Environm Engn Singapore Singapore Tongji Univ Key Lab Rd & Traff Engn Minist Educ Shanghai Peoples R China Graz Univ Technol Graz Austria JD Intelligent Cities Res Beijing Peoples R China

Contemporary urban transportation systems frequently depend on a variety of modes to provide residents with travel services. Understanding a multimodal transportation system is pivotal for devising well-informed planning;however, it is also inherently challenging for traffic analysts and planners. This challenge stems from the necessity of evaluating and contrasting the quality of transportation services across multiple modes. Existing methods are constrained in offering comprehensive insights into the system, primarily due to the inadequacy of multimodal traffic data necessary for fair comparisons and their inability to equip analysts and planners with the means for exploration and reasoned analysis within the urban spatial context. To this end, we first acquire sufficient multimodal trips leveraging well-established navigation platforms that can estimate the routes with the least travel time given an origin and a destination (an OD pair). We also propose TraDyssey, a visual analytics system that enables analysts and planners to evaluate and compare multiple modes by exploring acquired massive multimodal trips. TraDyssey follows a streamlined query-and-explore workflow supported by user-friendly and effective interactive visualizations. Specifically, a revisited difference-aware parallel coordinate plot (PCP) is designed for overall mode comparisons based on multimodal trips. Trip groups can be flexibly queried on the PCP based on differential features across modes. The queried trips are then organized and presented on a geographic map by OD pairs, forming a group-OD-trip hierarchy of visual exploration. Domain experts gained valuable insights into transportation planning through real-world case studies using TraDyssey. (c) 2025 The Authors. Published by Elsevier B.V. on behalf of Zhejiang University and Zhejiang University Press Co. Ltd. This is an open access article under the CC BY license (http://***/licenses/by/4.0/).

关键词： Traffic visualization Multimodal transportation Comparative analysis

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：