1 / 27

Lecture 1 4 : Convolutional Neural Networks on Surfaces via Seamless Toric Covers

Lecture 1 4 : Convolutional Neural Networks on Surfaces via Seamless Toric Covers. Jiacheng Cheng Feb, 21, 2018. 1. Convolutional neural networks on surfaces via seamless toric covers.

noone
Télécharger la présentation

Lecture 1 4 : Convolutional Neural Networks on Surfaces via Seamless Toric Covers

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Lecture 14: Convolutional Neural Networks on Surfaces via Seamless Toric Covers Jiacheng Cheng Feb, 21, 2018

  2. 1 Convolutional neural networkson surfacesviaseamlesstoriccovers HaggaiMaron,MeiravGalun,NoamAigerman,MiriTrope,NadavDym,ErsinYumer, Vladimir G. Kim, YaronLipman

  3. Problemstatement Easy Hard

  4. DeepLearning Geometric Deep Learning

  5. WhatisrequiredtodefineaCNN?

  6. Translations • Two dimensional,commutative Isometries ofR^2 • Convolution • Linear • Translationinvariant • Pooling • Non-linear(max) • Sub-translation invariant

  7. DefiningCNNsonsurfaces

  8. Translations onsurfaces? • Translationonsurface≝locallyEuclideantranslation • Flowalongnon-vanishingvectorfields

  9. Flat torus! • Translations “modulo1” • Full translation invariance on the flattorus !

  10. Only thetorus! Index of vectorfield Euler characteristic • Poincaré-Hopf:Foracompactorientablesurface • Index–ameasureofthecomplexitynearavanishingpoint • Non-vanishingvectorfieldimpliesgenus1-torus

  11. 15 CNNonflattorus Cyclic padding

  12. 16 Recap • CNNiswell-definedoverflat-torus • RoadblocksforCNNonsphere-typesurfaces • Topological:NolocallyEuclideantranslationsonspheres • Geometrical:Theflattorusisflatandoursurfaceisnot

  13. 17 Solution: Map the surface to a flattorus

  14. Torus4-cover

  15. 19 MappingtheTorustotheflatTorus ! Aigerman and Lipman,2015

  16. The pullbacktranslation !

  17. 22 Pull-back Translations: pull-back Euclideantranslations Two dimensional,commutative Conformalmaps ! Pull-backconvolution Linear Theorem: Translationinvariance Pull-backpooling Non-linear(max) Sub-translationinvariant

  18. 24 Newlayers projection cyclicpadding

  19. Datageneration Inputimage Labels

  20. 26 Testphase • Aggregation from differenttriplets • “Magnifyingglass” • Scale factor asweights • + + + =

  21. 27 Human bodysegmentation Train: 370models FAUST, MIT, SCAPE,ADOBE Test: 18models SHREC07

  22. Easyfunctions Raw • Normals • Average geodesicdistance • Wave kernelsignature Complex

  23. Human bodysegmentation Train: 370models FAUST, MIT, SCAPE,ADOBE Test: 18models SHREC07

  24. CNNappliedtootherdata

  25. Biological landmarksdetection • Train: 73teeth from BOYER • Onlycurvatureandscalefactor Test: 8 teeth fromBOYER

  26. 32 Biologicallandmarks

  27. 35 Conclusion • CNN of sphere-typesurfaces • Wedefinedameaningfulconvolutiononsurfaces • Learns from rawfeatures • ReusingCNNsoftwareforimages • Limitationsandfuturework • Scope:Onlyspheretypesurfaces • Nocanonicalchoicefortriplets(andconvolutions) • Learn aggregationoperator

More Related