工学部_研究紹介_2019_英語版

33/84

MaruyamaLabInMaruyamaLab.,ourresearchfocusesonimageunderstanding,contentbasedimagesearch,patternrecognition,3Drecoveryfromimages,andcomputergraphics.Especiallyweareveryinterestedintechniquestoimproveimageunderstandingandpatternrecognitionperformancebylearningfromexamples.We can easily recognize and understand the contents included in the image. This ability is improved through learning from our everyday experience. Our research goal is to providing such significant ability of learning to recognize various kinds of information to computer.Weareinvestigatingmethodsforimageunderstandingsuchasimagemodelingbasedonprobabilistictopicmodels,imagesegmentationviaenergyminimization,recoveryof3Dstructurefromimagesequence,Ourresearchinterestalsoincludecomputergraphicssuchasphoto-realisticrenderingalgorithms.Thesetechniquescanbeusedforvariouskindsofapplicationsincludingcontentbasedimagequeryontheweb,human-computerinteraction,MinoruMaruyamaProfessorSince 1996, he is with ShinshuUniv. Before joining ShinshuUniv., he was with Mitsubishi Electric Corp. 1990-1991, he was a visiting scientist of MIT AI Lab.Improving Image Understanding Performance by Learning from ExamplesOverhalfofthegraduatesgotograduateschool(masterprogramofShinshuUniversity).Graduatesofourlabareemployedbyelectronicscompanies,ITcompanies,includingvideogamecompaniesetc.Manygraduatesareworkingassoftware/ITengineers.We also study on computer graphics techniques such as photo-realistic rendering algorithms, semi-automatic character animation synthesis via reinforcement learning etc.Our work on the application of topic models to pattern classification received best paper award honorable mention at the International Workshop on Document Analysis Systems 2009 (DAS2009).After GraduationIn the FutureElectrical& Computer EngineeringYamamotoLabYamamotoLab.isstudyingondevelopingsearchalgorithmsforencrypteddataandXML(ExtensibleMarkupLanguage)data.Asecuresearchtechniqueforencrypteddataisoneofthemostimportanttechnologiesincloudcomputingbecauseitcanpreserveaprivacyandsensitivedata.AlsoasearchtechniqueforXMLdataisacriticalissuebecauseXMLisbecomingastandardlanguagesovertheInternet.Thereforeweaimatdevelopingefficientsearchalgorithmsfortheseissues.Inubiquitousnetworksociety,asecuresearchtechniqueforpreservingaprivacyandsensitivedataisoneofthemostimportanttechnologies.FurthermoreasearchtechniqueforXMLdataisalsoacriticaltechnology.YamamotoLab.isaimingatdevelopingnewmethodsfortheseissues.Hiroaki YamamotoProfessorIn 1988, he joined ShinshuUniversity. His research interests include pattern matching algorithms, automata theory, and information retrieval. GPU (Graphical Processing Unit): A GPU has many cores, and plays an important role in developing efficient algorithms.Study on Secure and Efficient SearchesSearchable encryption system: A user sends an encrypted query to a server. The server searches for documents using the encrypted query.Studentsgetmainlyjobsrelatedtotheinformationtechnology.After GraduationIn the FutureElectrical& Computer Engineering31