返回
[CVPR 2011] Hierarchical Semantic Indexing for Large Scale Image Retrieval

[CVPR 2011] Hierarchical Semantic Indexing for Large Scale Image Retrieval

ID：40702230

大小：930.58 KB

页数：8页

时间：2019-08-06

[CVPR 2011] Hierarchical Semantic Indexing for Large Scale Image Retrieval_第1页
[CVPR 2011] Hierarchical Semantic Indexing for Large Scale Image Retrieval_第2页
[CVPR 2011] Hierarchical Semantic Indexing for Large Scale Image Retrieval_第3页
[CVPR 2011] Hierarchical Semantic Indexing for Large Scale Image Retrieval_第4页
[CVPR 2011] Hierarchical Semantic Indexing for Large Scale Image Retrieval_第5页
资源描述：

《[CVPR 2011] Hierarchical Semantic Indexing for Large Scale Image Retrieval》由会员上传分享，免费在线阅读，更多相关内容在学术论文-天天文库。

1、HierarchicalSemanticIndexingforLargeScaleImageRetrievalJiaDeng1,3AlexanderC.Berg2LiFei-Fei3PrincetonUniversity1StonyBrookUniversity2StanfordUniversity3AbstractQueryimageTop5retrievedimagesWithouthierarchyThispaperaddressestheproblemofsimilarimagere-trieval,especiallyinthesettingoflarge-sca

2、ledatasetswithmillionstobillionsofimages.Thecorenovelcontributionisanapproachthatcanexploitpriorknowledgeofaseman-Withhierarchytichierarchy.Whensemanticlabelsandahierarchyrelat-ingthemareavailableduringtraining,significantimprove-Withouthierarchymentsoverthestateoftheartinsimilarimageretrie

3、valareattained.Whilesomeofthisadvantagecomesfromtheabilitytouseadditionalinformation,experimentsexploringaspecialcasewherenoadditionaldataisprovided,showWithhierarchythenewapproachcanstilloutperformOASIS[6],thecur-rentstateoftheartforsimilaritylearning.Exploitinghi-Withouthierarchyerarchic

4、alrelationshipsismostimportantforlargerscaleproblems,wherescalabilitybecomescrucial.Theproposedlearningapproachisfundamentallyparallelizableandasaresultscalesmoreeasilythanpreviouswork.Anaddi-Withhierarchytionalcontributionisanovelhashingscheme(forbilinearsimilarityonvectorsofprobabilities

5、,optionallytakingintoFigure1.Imagesretrievedbyexploitinghierarchyversusthoseaccounthierarchy)thatisabletoreducethecomputationalwithoutconsideringhierarchy.Greenbarsshowgroundtruthsim-costofretrieval.ExperimentsareperformedonCaltech256ilaritytothequery,definedbasedonthecategoryhierarchy(seea

6、ndthelargerImageNetdataset.Sec.5.2).Longerbarsindicatemoresimilarity.thatanimagecontainingahorsewouldbemoresimilarto1.Introductiononecontainingadogthantoanimagecontainingawind-mill.ItisfeasibletospecifyahierarchicalstructureintermsThispaperaddressestheproblemofsimilarimageretrievalofsemant

7、icattributes,butmaybequitedifficulttodosodi-–givenaqueryimage,findsimilarimagesinalargeimagerectlyintermsoflowlevelfeatures.collection–asdepictedinfigure1.Asillustratedthere,re-sultsshowthatexploitinghierarchicalrelationshipscansig-Thecurrentstateoftheartforsimil

当前文档最多预览五页，下载文档查看全文

此文档下载收益归作者所有

当前文档最多预览五页，下载文档查看全文
温馨提示：
1. 部分包含数学公式或PPT动画的文件，查看预览时可能会显示错乱或异常，文件下载后无此问题，请放心下载。
2. 本文档由用户上传，版权归属用户，天天文库负责整理代发布。如果您对本文档版权有争议请及时联系客服。
3. 下载前请仔细阅读文档内容，确认文档内容符合您的需求后进行下载，若出现内容与标题不符可向本站投诉处理。
4. 下载文档时可能由于网络波动等原因无法下载或下载错误，付费完成后未能成功下载的用户请联系客服处理。