检索结果-南通市图书馆

Learning to focus: cascaded feature matching network for few-shot image recognition

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Science China(Information Sciences) 2021年第9期64卷 90-102页

作者： Mengting CHEN Xinggang WANG Heng LUO Yifeng GENG wenyu LIU School of Electronic Information and Communications Huazhong University of Science and Technology

Generally, deep networks learn to recognize a category of objects by training on a large number of annotated images accurately. However, a meta-learning problem known as a low-shot image recognition task occurs when a few images with annotations are available for learning a recognition model for a single category. Consequently, the objects in testing/query and training/support image datasets are likely to vary in terms of size, location, style, and so on. In this paper, we propose a method, cascaded feature matching network(CFMN), to solve this problem. We train the meta-learner to learn a more fine-grained and adaptive deep distance metric using feature matching block, which aligns associated features together and naturally ignores non-discriminative features. By applying the proposed feature matching block in different layers of the network, multi-scale information among the compared images is incorporated into the final cascaded matching feature, which boosts the recognition performance and generalizes better by learning on relationships. Moreover, the experiments for few-shot learning(FSL) using two standard datasets:miniImageNet and Omniglot, confirm the effectiveness of our proposed method. Besides, the multi-label fewshot task is first studied on a new data split of the COCO dataset, which further shows the superiority of the proposed feature matching network when performing the FSL in complex images.

关键词： few-shot learning image recognition feature matching self-attention

Application of Fair-faced Concrete Material in Architectural Design

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Journal of Landscape Research 2024年第2期16卷 77-80,86页

作者： liu wenyu North China University of Technology Beijing 100144China

With the development of the society,the function of the building also becomes rich from single refuge,and modern architects can’t satisfy the pursuit of traditional architectural form,and design goal is transformed into emphasizing the dialogue between architecture and people,architecture and emotions,and architecture and environment.As a kind of new building materials,fair-faced concrete expresses almost any emotion and meets people’s emotional needs with its excellent structural performance and good plasticity.In this paper,the characteristics,expressive force and application of fair-faced concrete are studied and analyzed,and the application and development prospect of fair-faced concrete in architectural design are summarized.

关键词： Fair-faced concrete Expressive force Architectural design Tadao Ando

Automatic extraction method of force chain information and its application in the flow photoelastic experiment of granular matter

维普期刊数据库评论

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Particuology 2023年第12期83卷 142-155页

作者： Qingfa Chen Enlin Long Wenjing Niu Jun liu wenyu Fan Hangyu Li School of Civil Engineering and Architecture Key Laboratory of Disaster Prevention and Structural Safety of Ministry of EducationGuangxi UniversityNanning530004China School of Resources Environment and MaterialsGuangxi UniversityNanning530004China

The force chain is the core of the multi-scale analysis of granular matter.Accurately extracting the force chain information among particles is of great significance to the study of particle mechanics and geological hazards caused by particle flow.However,in the photoelastic experiment,the precise identification of the branching points of force chains has not been effectively realized.Therefore,this study proposes an automatic extraction method of force chain key information.First,based on the Hough transform and the Euclidean distance,a particle geometric information identification model is established and geometric information such as particle circle center coordinates,radius,contact point location,and contact angle is extracted.Then,a particle contact force information identification model is established following the color gradient mean square method.The model realizes the rapid calibration and extraction of a large number of particle media contact force information.Next,combined with the force chain composition criterion and its quasilinear feature,an automatic extraction method of force chain information is established,which solves the problem of the accurate identification of the force chain branch points.Finally,in the photoelastic experiment of ore drawing from a single drawpoint,the automatic extraction method of force chain information is verified.The results show that the macroscopic distribution of force chains during ore drawing from a single drawpoint is left–right symmetrical.Strong force chains are mostly located on the two sides of the model but in small numbers and they mainly develop vertically.Additionally,the ends are mostly in a combination of Y and inverted Y shapes,while the middle is mostly quasilinear.Weak force chains are abundant and mostly distributed in the middle of the model,and develop in different directions.The proposed extraction method accurately extracts the force chain network from the photoelastic experiment images and dynamically characterizes the force chains of granular

关键词： Granular matter Photoelastic experiment Digital image processing Force chain extraction method

Tiny Det: accurately detecting small objects within 1 GFLOPs

维普期刊数据库评论

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Science China(Information Sciences) 2023年第1期66卷 318-319页

作者： Shaoyu CHEN Tianheng CHENG Jiemin FANG Qian ZHANG Yuan LI wenyu LIU Xinggang WANG School of Electronic Information and Communications Huazhong University of Science and Technology Institute of Artificial Intelligence Huazhong University of Science and Technology Horizon Robotics Inc. Google Inc.

Dear editor,Object detection plays an important role in computer vision,and has gradually become the technical foundation of many applications. However, the inference procedure of advanced detection models requires ma... 详细信息

关键词： Economic and social effects

3D-printed tissue repair patch combining mechanical support and magnetism for controlled skeletal muscle regeneration

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Bio-Design and Manufacturing 2022年第2期5卷 249-264页

作者： Xiaocheng Wang Ruibo Zhao Jian Wang Xinghuan Li Lijuan Jin wenyu LIU Lifang Yang Yonghua Zhu Zhikai Tan College of Biology Hunan UniversityChangsha 410082China Department of Orthopaedics Xiangya HospitalCentral South UniversityChangsha 410008China Key Laboratory of Carcinogenesis and Cancer Invasions of Ministry of Education School of Basic Medicine ScienceCancer Research InstituteCentral South UniversityChangsha 410008China

Physical forces,such as magnetic and mechanical stimulation,are known to play a significant role in the regulation of cell response.In the present study,a biomimetic regeneration patch was fabricated using E-jet 3 D printing,which integrates mechanical and magnetic stimulation in a biocompatible"one-pot reaction"strategy when combined with a static magnetic field(SMF).The magneto-based therapeutic regeneration patch induced myoblasts to form aligned and multinucleated myotubes,regulated the expression of myogenic-related genes,and activated the p38αmitogen-activated protein kinase pathway via the initiation of myogenic differentiation.To validate the efficiency of the proposed strategy,the regeneration patch was implanted into mice and exposed to a suitable SMF,which resulted in significantly enhanced in vivo skeletal muscle regeneration.The findings demonstrated that appropriate external physical stimulation provides a suitable biophysical microenvironment that is conducive to tissue regeneration.The method used in the present study represents a promising technique to induce the regeneration of damaged skeletal muscle tis sue.

关键词： Functional biomaterials Magnetic actuation Myogenic differentiation Skeletal muscle regeneration 3D printing

Identification and functional analysis of heterogeneous FOXP3^+ Treg cell subpopulations in human pancreatic ductal adenocarcinoma

在线全文

学校读者我要写书评

暂无评论

Science Bulletin 2018年第15期63卷 972-981页

作者： Gang Yi Shiwei Guo wenyu LIU Huan Wang Rendong liu Andy Tsun Gang Jin Bin Li Shanghai Key Laboratory of Bio-energy Crops School of Life Science Shanghai University Shanghai Institute of Immunology Shanghai Jiao Tong University School of Medicine Department of Immunology and Microbiology Shanghai Jiao Tong University School of Medicine CAS Center for Excellence in Molecular Cell Science Unit of Molecular Immunology Institute Pasteur of Shanghai Shanghai Institutes for Biological Sciences University of Chinese Academy of Sciences Medical School Chinese Academy of Sciences Department of General Surgery Changhai Hospital Second Military Medical University

CD4^+CD25^+regulatory T(Treg) cells express the transcription factor FOXP3 and play an essential role in preventing autoimmunity. Abundant Treg cell accumulation in tumors and tumor draining lymph nodes(TDLNs) has been reported to correlate with both poor and favorable prognosis in various cancers, which suggests that Tregs may have multiple effects on antitumor immunity. However, the heterogeneity of tumor^-and TDLN^-infiltrating Treg cells remains unclear. Here we provide heterogeneity analysis of tumor infiltrating human CD4^+Treg cells and their matched adjacent tissues and TDLNs. We defined three different subpopulations of tumor^-and TDLN^-infiltrating Treg cells by Helios and CCR8 expression in pancreatic ductal adenocarcinoma(PDAC) and confirmed their functional heterogeneity. Helios^+CCR8^+Treg cells with potent suppressor function and limited IL^-2 and IFN^-c secretion were identified in tumors and TDLNs. On the contrary, Helios^-CCR8^-Treg cells have impaired suppressive activity, and elevated expression of pro^-inflammatory cytokines. More advanced grades of PDAC have predominantly Helios^+CCR8^+Treg cells and few Helios^-CCR8^-Treg cells both in tumors and TDLNs that suggests poor prognosis. These data could help further define the role of Treg cells and their functional role in tumors and TDLNs.

关键词：房间胰腺腺癌 cytokines 鉴定异构太阳神 IL-2

EAT-NAS: elastic architecture transfer for accelerating large-scale neural architecture search

在线全文

学校读者我要写书评

暂无评论

Science China(Information Sciences) 2021年第9期64卷 103-115页

作者： Jiemin FANG Yukang CHEN Xinbang ZHANG Qian ZHANG Chang HUANG Gaofeng MENG wenyu LIU Xinggang WANG Institute of Artificial Intelligence Huazhong University of Science and Technology School of Electronic Information and Communications Huazhong University of Science and Technology National Laboratory of Pattern Recognition Institute of Automation Chinese Academy of Sciences

Neural architecture search(NAS) methods have been proposed to relieve human experts from tedious architecture engineering. However, most current methods are constrained in small-scale search owing to the issue of huge computational resource consumption. Meanwhile, the direct application of architectures searched on small datasets to large datasets often bears no performance guarantee due to the discrepancy between different datasets. This limitation impedes the wide use of NAS on large-scale tasks. To overcome this obstacle, we propose an elastic architecture transfer mechanism for accelerating large-scale NAS(EATNAS).In our implementations, the architectures are first searched on a small dataset, e.g., CIFAR-10. The best one is chosen as the basic architecture. The search process on a large dataset, e.g., ImageNet, is initialized with the basic architecture as the seed. The large-scale search process is accelerated with the help of the basic architecture. We propose not only a NAS method but also a mechanism for architecture-level transfer learning. In our experiments, we obtain two final models EATNet-A and EATNet-B, which achieve competitive accuracies of 75.5% and 75.6%, respectively, on ImageNet. Both the models also surpass the models searched from scratch on ImageNet under the same settings. For the computational cost, EAT-NAS takes only fewer than 5 days using 8 TITAN X GPUs, which is significantly less than the computational consumption of the state-of-the-art large-scale NAS methods.

关键词： architecture transfer neural architecture search evolutionary algorithm large-scale dataset

Removal of Basic Nitrogen Compounds from Coker Diesel by Eutectic Ionic Liquid

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

China Petroleum Processing & Petrochemical Technology 2018年第3期20卷 48-54页

作者： liu Jie liu Jinbo Li Wenshen Long wenyu College of Chemistry Chemical Engineering and Environmental Engineering Liaoning Shihua University

The eutectic ionic liquid(EIL) tetraethyl ammonium bromide-malonic acid(TEAB-Mal) was synthesized, with its structure characterized by the FT-IR spectroscopy and the 1 H NMR spectrometry. The performance for removal of basic nitrogen compounds by EIL was studied using coker diesel as the feedstock. Experimental results showed that the EIL(TEAB-Mal) exhibited a good denitrogenation performance, leading to a 93.6% of basic N-removal efficiency under reaction conditions covering: a temperature of 30 °C, an EIL to oil mass ratio of 1:1, an extraction time of 30 min, and a settling time of 120 min, while the basic nitrogen content in diesel dropped from 580 μg/g to 37 μg/g. In addition, the efficiency for extraction of basic N-compounds could still reach 62.9% at an EIL/oil mass ratio of 1:7 after four recycles of the EIL.

关键词：移动效率混合物柴油机液体溶解奥尼氮红外光谱学

Efficient Pose: Efficient human pose estimation with neural architecture search

在线全文

学校读者我要写书评

暂无评论

Computational Visual Media 2021年第3期7卷 335-347页

作者： Wenqiang Zhang Jiemin Fang Xinggang Wang wenyu LIU School of EIC Huazhong University of Science and TechnologyWuhan 430074China Institute of Artificial Intelligence Huazhong University of Science and TechnologyWuhan 430074China

Human pose estimation from image and video is a key task in many multimedia applications.Previous methods achieve great performance but rarely take efficiency into consideration,which makes it difficult to implement the networks on lightweight devices.Nowadays,real-time multimedia applications call for more efficient models for better interaction.Moreover,most deep neural networks for pose estimation directly reuse networks designed for image classification as the backbone,which are not optimized for the pose estimation task.In this paper,we propose an efficient framework for human pose estimation with two parts,an efficient backbone and an efficient head.By implementing a differentiable neural architecture search method,we customize the backbone network design for pose estimation,and reduce computational cost with negligible accuracy degradation.For the efficient head,we slim the transposed convolutions and propose a spatial information correction module to promote the performance of the final prediction.In experiments,we evaluate our networks on the MPII and COCO datasets.Our smallest model requires only0.65 GFLOPs with 88.1%PCKh@0.5 on MPII and our large model needs only 2 GFLOPs while its accuracy is competitive with the state-of-the-art large model,HRNet,which takes 9.5 GFLOPs.

关键词： pose estimation neural architecture search efficient deep learning