CAP 6412 Advanced Computer Vision - CS Departmentbgong/CAP6412/lec1.pdfEmail --- the best way to...

Preview:

Citation preview

CAP6412AdvancedComputerVision

Website:http://www.cs.ucf.edu/~bgong/CAP6412.html

Jan12,2016

Introductions

• Instructor:Boqing Gongbgong@crcv.ucf.edu

• Officehours:Tuesday4:30—5:30pmatHEC214

Email--- thebestwaytoreachme

• Put[CAP6412] insubjectline• Summarizemessageinsubjectline• Ex:[CAP6412]Meetingrequest:Thursday(Jan14)4:30pm?

Today

• Courseoverview• Logistics,requirements

Whatiscomputervision?

• Computerprogramstorecognizeimages• Understandimagesequencesorvideos• Visualscences aroundus• Imitatehumancapabilitiestounderstandscenes• Perception,visualperceivingofthe3Dworld

Whatiscomputervision?

• Duplicatethecapabilitiesofhumanvisiononmachines

• “Acquire,process,analyze,andunderstandimages,videos,3Ddata,andothertypesofhigh-dimensionaldataofthereal(visual)world”--- Wikipedia.org

Researchoncomputervision

• TopicsSceneundestandingMedicalimageanalysisHuman-centricCV

Humanactionrecognition,Face3DScenereconstructionDocument/characterrecognitionLow-levelcomputervision

• Methods,tools,modelsGraphicalmodels(HMM,GMM,etc.),Neuralnetworks,OpenCV

• Relateddisciplines• Graphics,Machinelearning,HCI,Imageprocessing,Medicine,Statistics

Researchoncomputervision

• Topics3DmodelingandreconstructionStereo,structurefrommotionBiomedicalimageanalysisComputationalphotographyFaceandgesturesLow-levelvision(edge,color,etc.)MotionandtrackingOptimizationStatisticallearning,interactivelearningRecognition(ofobjects,attributes,scenes)SegmentationVideo:action,tracking,surveillanceVisionforX

• Relateddisciplines

Disseminatingcomputervisionresearch

• Conferences • Journals,transactions

Disseminatingcomputervisionresearch

• Conferences- CVPR,ICCV,ECCV,NIPS- BMVC,WACV,ACCV,etc.

• Journals,transactions- IEEET-PAMI,IJCV- IEEET-IP,T-MM,CVIU,etc.

• Submission,double-blind review,rebuttal,decision• Morenoisyreviews• Publicationinabout6months

• Submission,single-blind review,response,review,response,…...,decision• Publicationin>1years

arXiv.org:e-prints,self-archived

Digestingcomputervisionresearch

• #1courseobjective- Familiarwithfrontierproblems,researchgroups,andresearchers- Understandstate-of-the-arttechniquesandtools- Familiarwithcomputervisiondatasets(benchmarks)- Understandevaluationmethodsandmetrics

- Improvecriticalanalysiscapabilities

Howtoreadresearchpapers?(1)

• Askquestionsbeforeyoustart

Howtoreadresearchpapers?(1)

• Askquestionsbeforeyoustart

- Whatarethemotivationsforthispaper?- Whatproblemisbeingsolved?- Whatistheproposedsolution?- Whatexperimentsaredesignedtotestthesolution?- Whataretheevaluationmethodsandmetrics

- Whatarethecontributions?- Whatarethefuturedirections?- Howisthepaperrelatedtowhatyoupreviouslyknew?- Howisthepaperrelatedtootherworks?

Howtoreadresearchpapers?(2)

• Repeat:decreasingmarginalgainuntilconvergence?• Getthegeneralideafirst

• Whatproblemisbeingsolved?• Whatarethemainstepsinthesolution?• Whatisthehigh-levelideaofthesolution?

• Drawanalogiestowhatyouknowwell• Readinorder:

• Abstract,conclusion,problemstatementifitexists,intro,approach,experiments

• Divideandconquer• HowtoreadresearchpapersbyDr.MubarakShah:http://crcv.ucf.edu/people/faculty/HowToRead.html

• Thethree-passapproachbyDr.SrinivasanKeshav:http://ccr.sigcomm.org/online/files/p83-keshavA.pdf

Howtoreviewresearchpapers?(1)

• Writeareviewonthepaper--- #2courseobjective- Summarizeatake-homemessage- Recallthemainstrengthsofthepaper

- Novelty- Technicalcorrectness- Clarity- Experimentalevaluation

Howtoreviewresearchpapers?(2)

• Writeareviewonthepaper--- #2courseobjective- Summarizeatake-homemessage- Recallthemainstrengthsofthepaper- Bepicky:weaknessofthepaper

- Lackofnovelty:comparingtopriorwork- Lackofclarity:language,organization,presentation- Technicalerrors:rare- Mismatchedexperimentdesign- Insufficientexperiments- Unfaircomparisonwithothermethods- Justifyyourcomment

Howtoreviewresearchpapers?(3)

• Writeareviewonthepaper--- #2courseobjective- Summarizeatake-homemessage- Recallthemainstrengthsofthepaper- Bepicky:weaknessofthepaper- Overallrating(adaptedfromNIPSreviewerinstructions)

- 0: Top10%ofthepapersIhaveread,anexcellentpaper,astrongaccept.- Iwillfightforacceptance.IwillconsidernotreviewingpapersforXXifthisisrejected.

- 1: Top50%ofacceptedNIPSpapers,averygoodpaper,aclearaccept.- Ivoteandargueforacceptance.

- 2: Goodpaper,accept.- Ivoteforacceptance,althoughwouldnotbeupset ifitwererejected.

- 3:Marginallyabovetheacceptancethreshold.- Itendtovoteforacceptingit,but leavingitoutoftheprogramwouldbenogreatloss.

- 4:Marginallybelowtheacceptancethreshold.- Itendtovoteforrejectingit,buthavingitintheprogramwouldnotbethatbad.

- 5: AnOKpaper,butnotgoodenough.Arejection.- Ivoteforrejectingit,althoughwouldnotbeupset ifitwereaccepted.

Howtoreviewresearchpapers?(4)

• Writeareviewonthepaper--- #2courseobjective- Summarizeatake-homemessage- Recallmainstrengthsofthepaper- Bepicky:weaknessesofthepaper- Overallrating(0,1--5)- Explainyourrating:howyouweighthestrengthsandweaknesses- Brainstormfuturedirections

- ??

Howtoreviewresearchpapers?(4)

• Writeareviewonthepaper--- #2courseobjective- Summarizeatake-homemessage- Recallmainstrengthsofthepaper- Bepicky:weaknessesofthepaper- Overallrating(0,1--5)- Explainyourrating:howyouweighthestrengthsandweaknesses- Brainstormfuturedirections

- Newsolutions- Newsolutionsextendingorinspiredbythepaper’ssolution- Openproblems- Otherproblemsthatcouldbenefitfromthepaper

Howtopresentresearchpapers(inclass)?(1)

• Makegoodpresentations--- #3courseobjective- Knowyouraudience:fellowgraduatestudentswithgoodbackground- Beawareofyourtime:Tues,Thurs,3pm—4:15pm- Adaptthepresentationgoal:explainanddiscussthepaper- Assumenooneintheclasshasreadthepaperbefore

Howtopresentresearchpapers?(2)

• Makegoodpresentations--- #3courseobjective- Title,authors(fullname),authors’institutes,yournameandemail- Motivationoftheresearch(1—2slides)- Problemstatement

- Whatisbeingsolved?- Whyisitanimportantproblem?

- Maincontributionsofthepaper- Studiedanewandimportantproblem- Proposedanovelapproach- Improvedorextendedexistingmethods- Comparedseveralpopularmethods- Exploredavarietyofusecases(manydatasetsofdifferentkinds)- Presentednewtheories- Introducednewmethodologiesortoolstocomputervision

Howtopresentresearchpapers?(3)

• Makegoodpresentations--- #3courseobjective- Title,authors(fullname),authors’institutes,yournameandemail- Motivationoftheresearch(1—2slides)- Problemstatement(1—2slides)- Maincontributionsofthepaper- Approachoutline(1slide)- Detailsoftheproposedapproach- Experiments

- Data,features,baselines,evaluationmetrics,results- Relatedwork(1—3slides)- Conclusion:take-homemessage(1—2slides)

Howtopresentresearchpapers?(4)

• Makegoodpresentations--- #3courseobjective- Title,authors(fullname),authors’institutes,yournameandemail- Motivationoftheresearch(1—2slides)- Problemstatement(1—2slides)- Maincontributionsofthepaper- Approachoutline(1slide)- Detailsoftheproposedapproach- Experiments- Relatedwork(1—3slides)- Conclusion:take-homemessage(1—2slides)- Strengths&weaknessesofthepaper(1—2slides)- Overallrating&why(howyouweighthestrengthsandweaknesses)(1slide)- Futuredirections(1—3slides)

Syllabus(1)

• Overarchingobjective- Introducethefrontiersandtrendsincomputervision- Preparestudentsforcomputervisionresearch

• Throughtakingthiscourse- Gainin-depthunderstandingofthestate-of-the-artsincomputervision- Writehigh-qualitypaperreviews- Improvepresentationskills- Sharpenprogrammingandcriticalanalysisskills- Identifynewresearchdirections

Syllabus(2)

• Topics- Convolutionalneuralnetworks:basicsandapplications- Visionandlanguage:imagecaptioning,visualquestionanswering,etc.- Representations:low-level,middle-level(attributes,parts),CNNlayers,etc.- Video:action,tracking,surveillance,etc.- Statisticalmodelsandlearning- Imageretrievalandmatching- Human-centeredcomputervision:face,pose,pedestrian,ego-centric,feedback,humansintheloop,activelearning,etc.

- Visualsaliency:saliency,gaze,objectproposal,etc.- Formoretopics,seeSyllabus andPapersoncoursehomepage:http://www.cs.ucf.edu/~bgong/CAP6412.html

SubmituptothreepreferredtopicsbyJan14th 2016(Thurs),1pm(UCFFinancialAid)

Syllabus(3)

• GradingpolicyReports(writingpaperreviews):25%

- Tworeviewsperweek(dueresp.Tuesday,Thursday,3pm)- DueonJan14th 2016(Thurs),1pm:uptothreepreferredtopics

PaperPresentations:20%- 1—2presentationseveryclass(volunteers?)

DiscussionandAttendance:20%ProgrammingProjects:25%

- Potentialworkshop&conferencesubmissions- Projectpresentation:10%

Finalprojectpresentation:1:00pm—3:50pm,April28th.Lettergrades:90—100:A,80—89:B,70—79:C,60—69:D,below60:F

Syllabus(4)

• Latehomeworkpolicy- Threelatedaysintotalforallreportsandprojects- Countingatthegranularityof12hours- Noadditionallatedays

Syllabus(5)

• ImportantdatesJan11th, 2016:ClassesbeginJan14th,2016:LastdaytodropandrequestfullrefundJan15th,2016:AdddeadlineonmyUCFMarch23rd,2016:WithdrawaldeadlineApril26th,2016:Classesend;LastdaytoremoveincompleteApril28th,2016:Finalprojectpresentation

Seehttp://calendar.ucf.edu/2016/spring foranyupdates

Syllabus(6)

• StatementonAcademicIntegrity:The UCF Golden Rule (http://goldenrule.sdes.ucf.edu/) will be observed inthe class. Plagiarism and Cheating of any kind on an examination, quiz, orassignment will result at least in an "F" for that assignment (and may,depending on the severity of the case, lead to an "F" for the entire course)and may be subject to appropriate referral to the Office of Student Conductfor further action. I will assume for this course that you will adhere to theacademic creed of this University and will maintain the highest standards ofacademic integrity. In other words, don't cheat by giving answers to othersor taking them from anyone else. I will also adhere to the highest standardsof academic integrity, so please do not ask me to change (or expect me tochange) your grade illegitimately or to bend or break rules for one personthat will not apply to everyone.

Questions?

Recommended