检索结果-南通市图书馆

Learning generalizable deep feature using triplet-batch-center loss for person re-identification

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2期64卷 58-59页

作者： Bin HU Jiwei XU Xinggang wang School of Electronic Information and Communications Huazhong University of Science and Technology

Dear editor,Metric learning loss functions are important for deep learning-based person re-identification(re-ID). Several loss functions, such as triplet loss, center loss, and sphere loss,have been proposed. However, there are a few discussions on the generalization ability of these loss functions.In this study, we proposed a novel metric-learning loss function, triplet-batch-center loss(TBCL), to learn more generalizable deep features than the previous loss functions.

关键词： Person re-ID triplet loss triplet-batch-center loss metric learning deep learning

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

Tiny Det: accurately detecting small objects within 1 GFLOPs

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) 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

Lignin characteristics in soil profiles in different plant communities in a subtropical mixed forest

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 Plant Ecology 2018年第4期11卷 560-568页

作者： Xinggang wang Qiuxiang Tian Qianxi Li Chang Liao Mei He Feng Liu Key Laboratory of Aquatic Botany and Watershed Ecology Wuhan Botanical GardenChinese Academy of SciencesWuhan 430074China University of Chinese Academy of Sciences Beijing 100049China

Aims Lignin is generally considered as an important indicator of soil organic carbon(SOC)storage and dynamics.to evaluate the effects of plant communities and soil depth on soil lignin is critical to better understand forest carbon cycling.Methods We compared lignin content and chemical signature in three soil depths of four major plant communities in a subtropical forest,which located in the north part of Wuling Mountains,China.Lignin was measured using CuO oxidation method.Important Findings Both lignin content and its biochemical signature in plant litter varied among communities.However,these differences were mostly no longer exist in the upper soil layers.Lignin chemistry in soils inherited some of the biochemical signature of lignin in litter,but in a diminished magnitude.these results suggest that different plant communities had similar decomposition process with vary-ing rates,caused diminished differences in lignin content and its biochemical signature.Lignin content decreased with soil depth,but the biochemical signature of lignin was not significantly dif-ferent among soil layers for all communities,which suggests that vertical movement of lignin within the soil profile is very likely a key process causing this similar biochemical signature.these results emphasized the important roles of lignin inputs and soil eluviation in shaping lignin characteristics and distribution in forest soils,which pinpoint the urgent need to consider hydrological processes in studying forest soil carbon cycling.

关键词： lignin content degradation status plant communities soil profile soil carbon

Efficient Pose: Efficient human pose estimation with neural architecture search

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

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

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

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

王兴刚雕塑作品

装饰 2009年第7期 143-143页

作者：王兴刚

关键词： cm

A Generalized Polymer Precursor Ink Design for 3D Printing of Functional Metal Oxides

在线全文

学校读者我要写书评

暂无评论

Nano-Micro Letters 2023年第10期15卷 433-448页

作者： Hehao Chen Jizhe wang Siying Peng Dongna Liu Wei Yan xinggang Shang Boyu Zhang Yuan Yao Yue Hui Nanjia Zhou Key Laboratory of 3D Micro/Nano Fabrication and Characterization of Zhejiang Province School of Engineering and Research Center for Industries of the FutureWestlake UniversityHangzhou 310030People’s Republic of China Institute of Advanced Technology Westlake Institute for Advanced StudyHangzhou 310024People’s Republic of China School of Materials Science and Engineering Zhejiang UniversityHangzhou 310027People’s Republic of China School of Chemical Engineering and Advanced Materials the University of AdelaideAdelaide 5005Australia

Three-dimensional-structured metal oxides have myriad applications for optoelectronic devices.Comparing to conventional lithography-based manufacturing methods which face significant challenges for 3D device architectures,additive manufacturing approaches such as direct ink writing offer convenient,on-demand manufacturing of 3D oxides with high resolutions down to sub-micrometer scales.However,the lack of a universal ink design strategy greatly limits the choices of printable oxides.Here,a universal,facile synthetic strategy is developed for direct ink writable polymer precursor inks based on metal-polymer coordination effect.Specifically,polyethyleneimine functionalized by ethylenediaminetetraacetic acid is employed as the polymer matrix for adsorbing targeted metal ions.Next,glucose is introduced as a crosslinker for endowing the polymer precursor inks with a thermosetting property required for 3D printing via the Maillard reaction.For demonstrations,binary(i.e.,ZnO,CuO,In_(2)O_(3),Ga_(2)O_(3),TiO_(2),and Y_(2)O_(3)) and ternary metal oxides(i.e.,BaTiO_(3) and SrTiO_(3)) are printed into 3D architectures with sub-micrometer resolution by extruding the inks through ultrafine nozzles.Upon thermal crosslinking and pyrolysis,the 3D microarchitectures with woodpile geometries exhibit strong light-matter coupling in the mid-infrared region.The design strategy for printable inks opens a new pathway toward 3D-printed optoelectronic devices based on functional oxides.

关键词： 3D printing Maillard reaction Polymer-assisted deposition Metal oxide Photonic crystal

由《侏罗纪世界3》观照科幻电影的风格延续与审美嬗变

在线全文

学校读者我要写书评

暂无评论

电影评介 2022年第14期 52-54页

作者：王兴刚内蒙古农业大学外国语言学院

电影《侏罗纪世界3》是“侏罗纪世界三部曲”的完结篇,在跨文化语境下延续了侏罗纪系列电影的叙事风格。影片承载了人们未来想象的审美流变,实现了人物塑造、文化想象与美学意义的有机融合。电影学者张卫认为科幻电影是对人类命运共同... 详细信息

电影《侏罗纪世界3》是“侏罗纪世界三部曲”的完结篇,在跨文化语境下延续了侏罗纪系列电影的叙事风格。影片承载了人们未来想象的审美流变,实现了人物塑造、文化想象与美学意义的有机融合。电影学者张卫认为科幻电影是对人类命运共同发展故事的讲述。(1)由环球影业出品,科林·特雷沃罗执导,克里斯·帕拉特和山姆·尼尔等人主演的科幻类型片《侏罗纪世界3》讲述了侏罗纪公园纳布拉尔岛火山爆发四年后,园内恐龙四散在世界各地,与人类共存。

关键词：科幻电影系列电影类型片侏罗纪公园火山爆发人类命运