基于多标签的自动人脸年龄估计分析-analysis of automatic face age estimation based on multiple tags.docx

下载文档 降价啦

7
0
约3.74万字
约 56页
2018-06-05 发布于上海
举报
版权申诉
保障服务

基于多标签的自动人脸年龄估计分析-analysis of automatic face age estimation based on multiple tags.docx

1、本文档共56页，可阅读全部内容。
2、有哪些信誉好的足球投注网站（book118）网站文档一经付费（服务费），不意味着购买了该文档的版权，仅供个人/单位学习、研究之用，不得用于商业用途，未经授权，严禁复制、发行、汇编、翻译或者网络传播等，侵权必究。
3、本站所有内容均由合作方或网友上传，本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺！文档内容仅供研究参考，付费前请自行鉴别。如您付费，意味着您自己接受本站规则且自行承担风险，本站不退款、不进行额外附加服务；查看《如何避免下载的几个坑》。如果您已付费下载过本站文档，您可以点击这里二次下载。
4、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等，请点击“版权申诉”（推荐），也可以打举报电话：400-050-0827(电话支持时间：9:00-18:30)。

基于多标签的自动人脸年龄估计分析-analysis of automatic face age estimation based on multiple tags

AbstractAgeestimationisanovelandsignificantresearchdirectionintheareaofbiometricidentification,andthereisplentyofageconcernedinformationcontainedinfacialimages.However,becauseofthedifficultiesingettingenoughtrainingsamplesaswellasthespecialcharacteristicsofagingprocess,theestimationresultsbasedonfacialimagesstillneedtobeimproved.Asaresult,motivatedbytheslowandcontinuousagingprocess,multi-labelbasedmodel,whichisfitterforagingcharacteristics,isintroducedtorepresentthefacialimagesinageestimation.Inmostofthepreviouswork,ageestimationisgenerallyformulatedasasingle-labelbasedproblem.However,sinceagingisagradualprocessandpeoplearealwaysintransitionperiodbetweenages,labelingafacialexamplewithanexactageisadifficultproblem.Meanwhile,sufficienttrainingdataislackformanyages.Inthispaper,toimprovetheaccuracyofageestimation,anovelapproachbyapplyingMulti-LabelLearningtotheagefeaturesisproposed.Inthatapproach,eachfacialimageistreatedasanexampleassociatedwiththeoriginlabelaswellasitsneighboringages,whichmakesthedatamorereliableandsufficient.Themotivationcomesfromtheobservationthat,withagechangesslowlyandsmoothly,peoplewouldlookquitelikethemselvesbeforeandafterseveralyears.Andinthispaper,inordertocapturecomprehensiveinformationinfacialimagesandfurtherbetterthepreviousapproach,multi-instancemodelisalsointroducedintherepresentationoffacialimages.Experimentsshowthattheproposedmulti-labelapproachoutperformsthetraditionalageestimationapproaches.Thoughthemulti-instanceimprovementdoesntachievebetterresults,analysisaswellasthefutureworkisshowninthispaper.Keywords:AgeEstimation,Multi-LabelLearning,Multi-InstanceMulti-LabelLearning目录摘要...............................................................................................................IAbstract............................................................................................................II1绪论1.1研究的背景及意义...............................................................................(1)1.2国内外研究现状..............................