sampling conditions and topological guarantees for shape reconstruction algorithms

44
Sampling conditions and topological guarantees for shape reconstruction algorithms Andre Lieutier, Dassault Sytemes Thanks to Dominique Attali for some slides (the nice ones) Thanks to Dominique Attali, Fréderic Chazal, David Cohen-Steiner for joint work

Upload: mandy

Post on 20-Jan-2016

31 views

Category:

Documents


0 download

DESCRIPTION

Sampling conditions and topological guarantees for shape reconstruction algorithms. Andre Lieutier , Dassault Sytemes Thanks to Dominique Attali for some slides (the nice ones ) Thanks to Dominique Attali , Fréderic Chazal , David Cohen-Steiner for joint work. Shape Reconstruction. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Sampling conditions and topological guarantees for shape reconstruction algorithms

Sampling conditions and topological guarantees for shape reconstruction algorithms

Andre Lieutier, Dassault Sytemes

Thanks to Dominique Attali for some slides (the nice ones)Thanks to Dominique Attali, Fréderic Chazal, David Cohen-Steiner for joint work

Page 2: Sampling conditions and topological guarantees for shape reconstruction algorithms

Shape Reconstruction

INPUT OUTPUT

Surface ofphysical object

UNKNOWN

• geometrically accurate• topologically correct

TriangulationSample in R3

Page 3: Sampling conditions and topological guarantees for shape reconstruction algorithms

INPUT OUTPUT

Unordered sequenceof images varyingin pose and lighting

Low-dimensional complex

Shape Reconstruction(or manifold learning)

Page 4: Sampling conditions and topological guarantees for shape reconstruction algorithms

Shape Reconstruction(or manifold learning)

INPUT OUTPUT

Space with smallintrinsec dimension

Sample in Rd

UNKNOWN

• geometrically accurate• topologically correct

Simplicial complex

Page 5: Sampling conditions and topological guarantees for shape reconstruction algorithms

Algorithms in 2DAlgorithms in 2D

heuristics to select a subsetof the Delaunay triangulation

Page 6: Sampling conditions and topological guarantees for shape reconstruction algorithms

Algorithms in 3DAlgorithms in 3D

heuristics to select a subsetof the Delaunay triangulation

Page 7: Sampling conditions and topological guarantees for shape reconstruction algorithms

A Simple Algorithm

INPUT OUTPUT

-offset = union of ballswith radius centeredon the sample

Shape Sample

UNKNOWN

Page 8: Sampling conditions and topological guarantees for shape reconstruction algorithms

A Simple Algorithm

INPUT OUTPUT

Shape Sample

-offset -complex

UNKNOWN

From Nerve Theorem:

Page 9: Sampling conditions and topological guarantees for shape reconstruction algorithms

A Simple AlgorithmOUTPUTShape Sample

Page 10: Sampling conditions and topological guarantees for shape reconstruction algorithms
Page 11: Sampling conditions and topological guarantees for shape reconstruction algorithms

Reconstruction theorem

Page 12: Sampling conditions and topological guarantees for shape reconstruction algorithms

Reconstruction theorem

[Niyogi Smale Weinberger 2004]

Sampling conditions

Page 13: Sampling conditions and topological guarantees for shape reconstruction algorithms
Page 14: Sampling conditions and topological guarantees for shape reconstruction algorithms

Beyond the reach : WFS and -reach

Page 15: Sampling conditions and topological guarantees for shape reconstruction algorithms

1

reach -reach wfs

wfs

Beyond the reach : WFS and -reach

Page 16: Sampling conditions and topological guarantees for shape reconstruction algorithms

Previous best known result for faithful reconstruction of set with positive m-reach

(Chazal, Cohen-Steiner,Lieutier 2006)

Page 17: Sampling conditions and topological guarantees for shape reconstruction algorithms

Previous best known result for faithful reconstruction of set with positive m-reach

(Chazal, Cohen-Steiner,Lieutier 2006)

Page 18: Sampling conditions and topological guarantees for shape reconstruction algorithms

Best known result for faithful reconstruction of set with positive m-reach

Page 19: Sampling conditions and topological guarantees for shape reconstruction algorithms

Under the conditions of the theorem, a simple offset of the sample is a faithful reconsruction

Previous best known result for faithful reconstruction of set with positive m-reach

(Chazal, Cohen-Steiner,Lieutier 2006)

Page 20: Sampling conditions and topological guarantees for shape reconstruction algorithms

Recovering homology(Cohen-Steiner,Edelsbrunner,Harer 2006)

(Chazal, Lieutier 2006)

Page 21: Sampling conditions and topological guarantees for shape reconstruction algorithms

Recovering homology(Cohen-Steiner,Edelsbrunner,Harer 2006)

(Chazal, Lieutier 2006)

Page 22: Sampling conditions and topological guarantees for shape reconstruction algorithms

Recovering homology(Cohen-Steiner,Edelsbrunner,Harer 2006)

(Chazal, Lieutier 2006)

Page 23: Sampling conditions and topological guarantees for shape reconstruction algorithms

Recovering homology(Cohen-Steiner,Edelsbrunner,Harer 2006)

(Chazal, Lieutier 2006)

Page 24: Sampling conditions and topological guarantees for shape reconstruction algorithms

Recovering homology(Cohen-Steiner,Edelsbrunner,Harer 2006)

(Chazal, Lieutier 2006)

Page 25: Sampling conditions and topological guarantees for shape reconstruction algorithms

Rips Complex

Page 26: Sampling conditions and topological guarantees for shape reconstruction algorithms

Samplng condition for Cech and Rips(D. Attali, A. Lieutier 2011)

[CCL06]

[NSW04]

Page 27: Sampling conditions and topological guarantees for shape reconstruction algorithms

Questions ?

Page 28: Sampling conditions and topological guarantees for shape reconstruction algorithms
Page 29: Sampling conditions and topological guarantees for shape reconstruction algorithms

Cech / Rips

Rips and Cech complexes generally don’t share the same the same topology, topology, but ...but ...

Rips and Cech complexes generally don’t share the same the same topology, topology, but ...but ...

Page 30: Sampling conditions and topological guarantees for shape reconstruction algorithms

Cech / Rips

Page 31: Sampling conditions and topological guarantees for shape reconstruction algorithms

Cech / Rips

Possesses a spirituous cyclethat we want to kill !

Possesses a spirituous cyclethat we want to kill !

Page 32: Sampling conditions and topological guarantees for shape reconstruction algorithms

Cech / Rips

Had there been a point close to the center,it would have distroy spirituous cycles appearingin the Rips, without changing the Cech.

Page 33: Sampling conditions and topological guarantees for shape reconstruction algorithms

Convexity defects function

Page 34: Sampling conditions and topological guarantees for shape reconstruction algorithms

Large -reach => small convexity defect functions

Page 35: Sampling conditions and topological guarantees for shape reconstruction algorithms

Density authorized

[CCL06]

[NSW04]

Page 36: Sampling conditions and topological guarantees for shape reconstruction algorithms

Questions ?

Page 37: Sampling conditions and topological guarantees for shape reconstruction algorithms

Cech Complex

Page 38: Sampling conditions and topological guarantees for shape reconstruction algorithms

Nerve Theorem

Page 39: Sampling conditions and topological guarantees for shape reconstruction algorithms

Persistent homology

Page 40: Sampling conditions and topological guarantees for shape reconstruction algorithms

Persistent homology

Page 41: Sampling conditions and topological guarantees for shape reconstruction algorithms

Persistent homology

Page 42: Sampling conditions and topological guarantees for shape reconstruction algorithms

Persistent homology

Page 43: Sampling conditions and topological guarantees for shape reconstruction algorithms

Persistent homology

Page 44: Sampling conditions and topological guarantees for shape reconstruction algorithms

Persistent homology