FinalProject776:“ComputerVision”,Fall2015Duedate:Nov30,2015,12:00pm(bye-mailtojmf@cs.unc.edu)Data:http://cvlabwww.epfl.ch/data/multiview/denseMVS.htmlWhatshouldyouturnin:Asinglepdffilewithallsolutionsandadiscussionofyoursolution.Alongwiththepdffileyoushouldprovidealinkforyourresultsandcode.In addition to the above you should prepare a short (~5 min) presentationdescribingyourwork.1)ImageDataAssociationThe goal of the assignment is to apply an unsupervised clustering algorithm tosegregateasetofimagesobservingdifferentscenes.

a) Download the datasets fountain-P11, Herz-Jesu-P8, entry-P10 from theStrechaMVSevaluationwebsiteandformanaggregatedataset.

b) Extract a global image descriptor for each image. You may choose any

available descriptor and are not required to develop descriptor code.Exampledescriptorsincludecolorhistograms,GIST,bagofvisualwords,etc.Alternatively, you may choose to engineer your own image descriptor ifdesired.

c) Implement amean-shift unsupervised clustering algorithm to partition theaggregateddatasetintodisjointimagesets.

d) ExtraCredit.DownloadtheCastle-P19imagedatasetandaggregatewiththeentry-P10 image dataset. Perform the same clustering procedure using thesameparameters.Pleasenote,thereissceneoverlapamongthesedatasets,makingcorrectsegregationmoreambiguous.

Includeinyourreporttheattainedclusters;discussyourchoiceofimagedescriptoraswellasthemechanismsusedforparametertuningofyourclusteringalgorithm.

2)StructureEstimationThegoaloftheassignmentistoimplementthemultiviewplanesweepalgorithmfordepthmap generation. To this end you shall use the provided camera pose andcalibrationinformationprovidedforeachimageintheaboveStrechadatasets.

a) Manually select fromeachof theclustergenerated for task1d)a referenceimagetogenerateadepthmap.

b) Use the remaining images in the cluster as matching images for themultiviewplanesweepalgorithm

Present in your report the attained depthmaps, explain the plane inducedhomography used for planesweeping and discuss your choices for photo-consistencymeasure,depthsamplingstrategyandmultiviewcostaggregation

3)SegmentationThegoalofthisassignmentistodevelopaframeworktoperformsegmentationofaspecificforegroundobject.TheobjectofinterestisaforegroundtractorfoundintheimagesoftheCastle-19dataset.

Thedesiredoutputisapixelwisesegmentationforeachoftheimagesinthedataset.

This is an open-ended problem for which we seek a solution that balancessegmentation qualitywith the level of supervision required to carry out the task.Moreover, in designing your solution you may leverage the shared informationamongdatasets(e.g. Castle-P19hascommonsceneregionswithentry-P10,whiletheavailableCastle-P30datasetalsodepictstheforegroundobjectandcanbeseenas a superset of Castle-P19), the available 3D ground truth data for each dataset(providedasadense3Dmesh),orevenelicituserinputaspartofasemi-automatedsegmentationframework.Includeinyourreporttheattainedsegmentationandprovideadetaileddiscussionofyoursolution.