Multi-CueOnboardPedestrianDetectionSVM算法Adboost算法.ppt

Multi-CueOnboard(1/4),UniversityofKarlsruhe,Germany,2000-2006VisitingStudentatMcGillUniversity,Montreal,Canada,2004-20052006-,puterScience,TUDResearchInterestObjectRecognition,SceneUnderstanding,V’08,2CVPR’09Authorinformation(2/4)StefanWalkEducationDiplomainPhysics,TechnischeUniversitätDarmstadt,Germany20072007-,puterScience,TUDResearchInterestPeopledetection,Detectingfromvideodata(utilizingmotioninformation)Papers1CVPR’09Authorinformation(3/4)BerntSchieleEducationPhD:Docteurdel'InstitutPolytechniquedeGrenoble,France,1997MSc's:Diplom-Informatiker,UniversityofKarlsruhe,Germany,1994,DEAdel'informatiquedel'ENSIMAG,France,1993ExperienceAssistantProfessor,ETHZurich,Switzerland,1999-2004PostdoctoralAssociateandVisitingAssistantProfessor,MITandCambridge,MA,USA,1997-2000VisitingresearcheratCMU,1994Authorinformation(4/4)puting,puterinterfacesExternalActivitiesAssociateEditorofPAMI,V’08,CVPR’09,ICCV’09,V2011,ProgramCoChairPapers2CVPR’09,2CVPR’08,V’08,1IJCV’08AbstractThispapersystematicallyevaluatesdifferentfeaturesandclassifiersinasliding--(1/7)最优线性分类面线性可分样本集合,线性分类面一般形式为最优分类面要求,距离分类面最近的样本距离(margin)尽可能大