检索结果-内蒙古大学图书馆

MSCA:A few-shot semantic segmentation framework driven by multi-scale cross-attention and information extraction

Computer Vision and Image Understanding 2025年

作者： Zhihao Ren Shengning Lu Xinhua Wang Yaoming Liu Yong Liang Key Laboratory of Advanced Manufacturing and Automation Technology Guilin University of Technology Education Department of Guangxi Zhuang Autonomous Region Guilin 541006 China College of Mechanical and Control Engineering Guilin University of Technology Guilin 541006 China

Few-Shot Semantic Segmentation (FSS) aims to achieve precise pixel-level segmentation of target objects in query images using only a small number of annotated support images. The main challenge lies in effectively capturing and transferring critical information from support samples while establishing fine-grained semantic associations between query and support images to improve segmentation accuracy. However, existing methods struggle with spatial alignment issues caused by intra-class variations and inter-class visual similarities, and they fail to fully integrate high-level and low-level decoder *** address these limitations, we propose a novel framework based on cross-scale interactive attention mechanisms. This framework employs a hybrid mask-guided multi-scale feature fusion strategy, constructing a cross-scale attention network that spans from local details to global context. It dynamically enhances target region representation and alleviates spatial misalignment issues. Furthermore, we design a hierarchical multi-axis decoding architecture that progressively integrates multi-resolution feature pathways, enabling the model to focus on semantic associations within foreground *** results show that our Multi-Scale Cross-Attention (MSCA) model performs exceptionally well on the PASCAL-5i and COCO-20i benchmark datasets, achieving highly competitive results. Notably, the model contains only 1.86 million learnable parameters, demonstrating its efficiency and practical applicability.

关键词：

来源：评论

学校读者我要写书评

暂无评论

CoF: Coarse to Fine-Grained Image Understanding for Multi-modal Large Language Models

arXiv

引用

arXiv 2024年

作者： Wang, Yeyuan Gao, Dehong Li, Bin Long, Rujiao Yi, Lei Cai, Xiaoyan Yang, Libin Zhang, Jinxia Yu, Shanqing Xuan, Qi School of Automation Northwestern Polytechnical University Shaanxi Xi’an China School of Cybersecurity Northwestern Polytechnical University Shaanxi Xi’an China Alibaba Group Zhejiang Hangzhou China The Key Laboratory of Measurement and Control of CSE Ministry of Education School of Automation Southeast University Nanjing210096 China Advanced Ocean Institute Southeast University Nantong226010 China Zhejiang University of Technology Zhejiang Hangzhou China Binjiang Institute of Artificial Intelligence Zhejiang Hangzhou China

The impressive performance of Large Language Model (LLM) has prompted researchers to develop Multi-modal LLM (MLLM), which has shown great potential for various multi-modal tasks. However, current MLLM often struggles to effectively address fine-grained multi-modal challenges. We argue that this limitation is closely linked to the models’ visual grounding capabilities. The restricted spatial awareness and perceptual acuity of visual encoders frequently lead to interference from irrelevant background information in images, causing the models to overlook subtle but crucial details. As a result, achieving fine-grained regional visual comprehension becomes difficult. In this paper, we break down multi-modal understanding into two stages, from Coarse to Fine (CoF). In the first stage, we prompt the MLLM to locate the approximate area of the answer. In the second stage, we further enhance the model’s focus on relevant areas within the image through visual prompt engineering, adjusting attention weights of pertinent regions. This, in turn, improves both visual grounding and overall performance in downstream tasks. Our experiments show that this approach significantly boosts the performance of baseline models, demonstrating notable generalization and effectiveness. Our CoF approach is available online at https://***/Gavin001201/CoF. © 2024, CC BY-NC-ND.

关键词： Visual languages

来源：评论

学校读者我要写书评

暂无评论

Optimization design of extensor for improving locomotion efficiency of inchworm-like capsule robot

引用

Science China(Technological Sciences) 2019年第11期62卷 1930-1938页

作者： GAO JinYang YAN GuoZheng SHI YunBo CAO HuiLiang HUANG Kun LIU Jun Science and Technology on Electronic Test and Measurement Laboratory North University of ChinaTaiyuan 030051China Shanxi Key Laboratory of Advanced Manufacturing Technology North University of ChinaTaiyuan 030051China Department of Instrument Science and Engineering Shanghai Jiaotong UniversityShanghai 200240China

An inchworm-like capsule robot(ILCR) is a promising device for a minimally invasive diagnosis and treatment of colon diseases. It consists of two expanders and one extensor, the former provides a traction force by expanding the colon and the latter can elongate and retract to enable active locomotion. However, the locomotion efficiency of the ILCR can be seriously lowered by the complex colon environment featuring slippery, viscoelastic, and suspend properties, which has been a main obstacle to its clinical application. This paper aims at improving the locomotion efficiency of the ILCR by optimizing its extensor design. To do this, the locomotion resistance of the ILCR in the colon is analyzed, and complying with a requirement that the traction force must be larger than the locomotion resistance to avoid slipping, a restriction on the extensor design is obtained. Then under the restriction and with reference to the Hyperelastic model which correlates stress and strain of colon tissue, a model for analyzing the influence of the design parameters of the extensor on the locomotion efficiency of the ILCR is built. With this model, the extensor has been optimized and the optimized results have been used to guide the development of a novel extensor, which employs two pairs of lead-screws and nuts and is actuated by one motor. Ex-vivo experiment has shown that the novel extensor can improve the locomotion efficiency of an ILCR prototype by 57%, without changing its total length.

关键词： inchworm-like capsule robot extensor design locomotion resistance locomotion efficiency

来源：评论

学校读者我要写书评

暂无评论

Numerical investigation and optimization of valve plate/port plate interface of integrated pump-motor unit for SWRO desalination system

引用

Desalination 2025年 613卷

作者： Zhang, Yu Ma, Zhonghai Yin, Fanglong Nie, Songlin Ji, Hui Luo, Hao Research Center of Novel Hydraulic Transmission and Control College of Mechanical & Energy Engineering Beijing University of Technology Beijing100124 China Beijing Key Laboratory of Advanced Manufacturing Technology Beijing University of Technology Beijing100124 China

With a novel design compared to conventional piston-type devices in seawater reverse osmosis (SWRO) desalination systems, integrated seawater pump-motor energy recovery device (ISPM-ERD) has the advantages of better integrity and higher efficiency than other kinds of ERDs. However, the defects of major leakage and friction brought by the valve plate/port plate interface (VPI) restrict its overall efficiency, and also the introduction of polyether-ether-ketone (PEEK) valve plate would contribute to such limitations. Thus, to investigate the lubricating characteristics of VPI inside ISPM-ERD, a mixed elastohydrodynamic (MEHD) model for numerical analysis is developed, considering pressure pulsations, rough peaks contact, and valve plate deformation which are solved by finite volume method (FVM) and finite element method (FEM) under various operating conditions. Also, an enhanced port plate design featuring reservoir groove (RG) is proposed to improve the VPI's support performance, reduce wear and power loss. Numerical simulations and experimental results demonstrate that the RG-enhanced VPI effectively increases the mechanical efficiency and overall efficiency, while causing a slight decrease in volumetric efficiency. This study offers a feasible way of in-depth research of VPI inside ISPM-ERD and other similar pump-motor ERDs, also providing a possible method for performance enhancement of SWRO systems. © 2025 Elsevier B.V.

关键词： Desalination

来源：评论

学校读者我要写书评

暂无评论

Hierarchical Clustering-Based State Grouping Reinforcement Learning for Switching Decision of Autonomous Vehicles

Hierarchical Clustering-Based State Grouping Reinforcement L...

引用

IEEE International Conference on Unmanned Systems (ICUS)

作者： Wenjie Ouyang Yiwen Jiao Yang Liu Yang Li Manjiang Hu Hongmao Qin College of Mechanical and Vehicle Engineering Hunan University Changsha China College of Computer Science and Electronic Engineering Hunan University Changsha China College of Electrical Mechanical and Computer Science Delft University of Technology Delft The Netherlands State Key Laboratory of Advanced Design and Manufacturing Technology for Vehicle College of Mechanical and Vehicle Engineering Hunan University Changsha China Wuxi Intelligent Control Research Institute of Hunan University Wuxi China

Reinforcement learning (RL) has gained wide attention, but its implementation in autonomous vehicles is still limited by insufficient sample efficiency and heavy training costs. The training efficiency of RL agents is influenced by the dimension of the state space, which can be partitioned to reduce the complexity of sampling and computation. This study proposes a hierarchical clustering-based state grouping reinforcement learning (HCSG-RL) method for the switching decision of autonomous vehicles. First, we partition the base state space into groups and generate a hierarchical tree of state space groups. Then, we train multiple sub-agents for each node in the hierarchical tree. Finally, we add these trained-well sub-model into master policy. This method allows us to fully explore all state spaces and improve the training efficiency of individual agents, which handles the “long-tail” issue and the curse of dimensionality issue. We conduct experiments in a simulation environment and results show that the proposed method has 16–72% reward improvement compared to the tree model in different road length.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Catalyst–Support Interaction in Polyaniline‑Supported Ni_(3)Fe Oxide to Boost Oxygen Evolution Activities for Rechargeable Zn‑Air Batteries

引用

Nano-Micro Letters 2025年第1期17卷 176-190页

作者： Xiaohong Zou Qian Lu Mingcong Tang Jie Wu Kouer Zhang Wenzhi Li Yunxia Hu Xiaomin Xu Xiao Zhang Zongping Shao Liang An Department of Mechanical Engineering The Hong Kong Polytechnic UniversityHung HomKowloonHong Kong SARPeople’s Republic of China Jiangsu Key Laboratory of Atmospheric Environment Monitoring and Pollution Control Jiangsu Collaborative Innovation Center of Atmospheric Environment and Equipment TechnologySchool of Environmental Science and TechnologyNanjing University of Information Science and TechnologyNanjing 210044People’s Republic of China Department of Chemistry The Chinese University of Hong KongMa Lin BuildingShatinHong Kong SAR 999077People’s Republic of China WA School of Mines:Minerals Energy and Chemical Engineering(WASM‑MECE)Curtin UniversityPerthWA 6102Australia Research Institute for Advanced Manufacturing The Hong Kong Polytechnic UniversityHung HomKowloonHong Kong SARPeople’s Republic of China Research Institute for Smart Energy The Hong Kong Polytechnic UniversityHung HomKowloonHong Kong SARPeople’s Republic of China

Catalyst–support interaction plays a crucial role in improving the catalytic activity of oxygen evolution reaction(OER).Here we modulate the catalyst–support interaction in polyaniline-supported Ni_(3)Fe oxide(Ni_(3)Fe oxide/PANI)with a robust hetero-interface,which significantly improves oxygen evolution activities with an overpotential of 270 mV at 10 mA cm^(-2)and specific activity of 2.08 mA cm_(ECSA)^(-2)at overpotential of 300 mV,3.84-fold that of Ni_(3)Fe *** is revealed that the catalyst–support interaction between Ni_(3)Fe oxide and PANI support enhances the Ni–O covalency via the interfacial Ni–N bond,thus promoting the charge and mass transfer on Ni_(3)Fe *** the excellent activity and stability,rechargeable Zn-air batteries with optimum Ni_(3)Fe oxide/PANI are assembled,delivering a low charge voltage of 1.95 V to cycle for 400 h at 10 mA cm^(-2).The regulation of the effect of catalyst–support interaction on catalytic activity provides new possibilities for the future design of highly efficient OER catalysts.

关键词： Catalyst-support interaction Supported catalysts Heterointerface Oxygen evolution reaction Zn-air batteries

来源：评论

学校读者我要写书评

暂无评论

Multiple magnetism-controlled topological states in EuAgAs

引用

Physical Review B 2021年第16期104卷 165424-165424页

作者： Yahui Jin Xu-Tao Zeng Xiaolong Feng Xin Du Weikang Wu Xian-Lei Sheng Zhi-Ming Yu Ziming Zhu Shengyuan A. Yang Key Laboratory of Low-Dimensional Quantum Structures and Quantum Control of Ministry of Education Department of Physics and Synergetic Innovation Center for Quantum Effects and Applications Hunan Normal University Changsha 410081 China School of Physics and Key Laboratory of Micro-nano Measurement-Manipulation and Physics Beihang University Beijing 100191 China Research Laboratory for Quantum Materials Singapore University of Technology and Design Singapore 487372 Singapore Key Lab of Advanced Optoelectronic Quantum Architecture and Measurement (MOE) Beijing Key Lab of Nanophotonics and Ultrafine Optoelectronic Systems and School of Physics Beijing Institute of Technology Beijing 100081 China

The interplay between magnetism and band topology is a focus of current research on magnetic topological systems. Based on first-principle calculations and symmetry analysis, we reveal that multiple intriguing topological states can be realized in a single system EuAgAs, controlled by the magnetic ordering. The material is a Dirac semimetal in the paramagnetic state, with a pair of accidental Dirac points. Under different magnetic configurations, the Dirac points can evolve into magnetic triply degenerate points, magnetic linear and double Weyl points, or being gapped out and making the system a topological mirror semimetal characterized by mirror Chern numbers. The change in bulk topology is also manifested in the surface states, including the surface Fermi arcs and surface Dirac cones. In addition, the antiferromagnetic states also feature a nontrivial Z4 index, implying a higher order topology. These results deepen our understanding of magnetic topological states and provide different perspectives for spintronic applications.

关键词： Antiferromagnetism Electronic structure First-principles calculations Topological materials

来源：评论

学校读者我要写书评

暂无评论

Analysis of Tool Wear and Wear Mechanism in Dry Forming Milling for Rail Milling Train

Analysis of Tool Wear and Wear Mechanism in Dry Forming Mill...

引用

作者： Pan, Chao Gu, Xiaoshan Wang, Mulan Wang, Baosheng China Railway Shanghai Administration Group Nanjing EMU Department Nanjing210012 China Jiangsu Key Laboratory of Advanced Numerical Control Technology NanjingJiangsu211167 China Jiangsu Provincial Engineering Laboratory of Intelligent Manufacturing Equipment Nanjing211167 China

Milling tools of the rail milling train have to face severe weariness due to the uneven surface hardness, inconsistent cutting rate and uncertain status of the rail itself as well as the vibration caused by the suspension system of the rail milling train and its characteristic hunting motion. This study performed a milling tool wearing status analysis and conducted relative research on wearing mechanisms with electron scanning microscope based on the typical milling tools and specific working conditions of rail milling train. Results indicate that the wearing status of rail milling train consists mainly of crater wear, notch wear, cutting edge chipping, surface exfoliation, milling sintering and tool break. The tool wearing conditions of rail milling train, compared with ordinary milling process, are significantly severe, and associated with multiple types of wearing status of which certain relevance exists. The research is of great significance for the improvement of the milling tool, the optimization of the process and the improvement of the mobile milling quality of the rail. © Springer Nature Singapore Pte Ltd. 2020.

关键词： Milling (machining)

来源：评论

学校读者我要写书评

暂无评论

Microstructure, Tensile Properties and Heat-Resistant Properties of Selective Laser Melted Almgsczr Alloy Under Long-Term Aging Treatment

SSRN

引用

SSRN 2022年

作者： Bi, Jiang Liu, Lei Chen, Guang Jia, Xiangdong Chen, Xi Xia, Hongbo Li, Xiaopeng Starostenkov, Mikhail Dmitrievich Han, Bing Dong, Guojiang Key Laboratory of Advanced Forging & Stamping Technology and Science of Ministry of Education Yanshan University Qinhuangdao066004 China College of Mechanical and Electrical Engineering Nanjing Forestry University Nanjing210037 China State Key Laboratory of Advanced Welding and Joining Harbin Institute of Technology Harbin150001 China School of Mechanical Engineering Yangzhou University Yangzhou225127 China College of Material Science and Engineering Nanjing University of Science and Technology Nanjing210094 China General Physics Department Altai State Technical University Leninst. 46 Barnaul656099 Russia Institute of Intelligent Manufacturing Guangdong Academy of Sciences Guangdong Key Laboratory of Modern Control Technology Guangzhou510070 China

Selective laser melting (SLM)-processed AlMgScZr alloy has aroused widespread interest in the aerospace industry due to its outstanding comprehensive performance and excellent heat resistant properties. The fine grain morphology and diffusely distributed thermally stable precipitates of this alloy are believed to play a critical role in its high-temperature performance. By modifying the alloy with Sc and Zr, nano-sized Al3(Sc, Zr) phases can act as heterogeneous nucleating particles to effectively promote grain refining and precipitate strengthening. In this paper, a Sc and Zr-modified Al-Mg alloy was printed by SLM, and the long-term aging treatments in the temperature range of 275-350 ℃ with a constant holding time of 144 h were performed to investigate the heat-resistant properties of this alloy. Under as-fabricated condition, the AlMgScZr alloy exhibited high tensile strengths of 487.7 MPa and 393.9 MPa at temperatures of 25 ℃ and 200 ℃, respectively. When the alloy was aged at 300 ℃ for 144 h, its grain morphology was almost unchanged, but its tensile strength and elongation at room temperature reached 501.3 MPa and 31.3 %, respectively. This SLM-fabricated AlMgScZr alloy is expected to replace some titanium alloy components for medium-temperature applications. © 2022, The Authors. All rights reserved.

关键词： Aluminum alloys

来源：评论

学校读者我要写书评

暂无评论

IR2Net: Information Restriction and Information Recovery for Accurate Binary Neural Networks

arXiv

引用

arXiv 2022年

作者： Xue, Ping Lu, Yang Chang, Jingfei Wei, Xing Wei, Zhen The School of Computer Science and Information Engineering Hefei University of Technology Hefei230009 China The School of Computer Science and Information Engineering Hefei University of Technology Hefei230009 China The Anhui Mine IOT and Security Monitoring Technology Key Laboratory Hefei230088 China The Engineering Research Center of Safety Critical Industrial Measurement and Control Technology Ministry of Education Hefei University of Technology Hefei230009 China The Intelligent Manufacturing Institute of HeFei University of Technology Hefei230009 China

Weight and activation binarization can efficiently compress deep neural networks and accelerate model inference, but cause severe accuracy degradation. Existing optimization methods for binary neural networks (BNNs) focus on fitting full-precision networks to reduce quantization errors, and suffer from the trade-off between accuracy and computational complexity. In contrast, considering the limited learning ability and information loss caused by the limited representational capability of BNNs, we propose IR2Net to stimulate the potential of BNNs and improve the network accuracy by restricting the input information and recovering the feature information, including: 1) information restriction: for a BNN, by evaluating the learning ability on the input information, discarding some of the information it cannot focus on, and limiting the amount of input information to match its learning ability;2) information recovery: due to the information loss in forward propagation, the output feature information of the network is not enough to support accurate classification. By selecting some shallow feature maps with richer information, and fusing them with the final feature maps to recover the feature information. In addition, the computational cost is reduced by streamlining the information recovery method to strike a better trade-off between accuracy and efficiency. Experimental results demonstrate that our approach still achieves comparable accuracy even with ∼10x floating-point operations (FLOPs) reduction for ResNet-18. The models and code are available at https://***/pingxue-hfut/IR2Net. Copyright © 2022, The Authors. All rights reserved.

关键词： Recovery

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：