Harmonic 3D Shape Matching Michael Kazhdan Thomas Funkhouser Princeton University Michael Kazhdan Thomas Funkhouser Princeton University.

Slides:



Advertisements
Similar presentations
Distinctive Image Features from Scale-Invariant Keypoints David Lowe.
Advertisements

Shape Analysis and Retrieval D2 Shape Distributions Notes courtesy of Funk et al., SIGGRAPH 2004.
Wavelets Fast Multiresolution Image Querying Jacobs et.al. SIGGRAPH95.
Three things everyone should know to improve object retrieval
Evaluating Sketch Query Interfaces for a 3D Model Search Engine Patrick Min Joyce Chen, Tom Funkhouser Princeton Workshop on Shape-Based Retrieval and.
Extended Gaussian Images
Mesh modeling and processing M. Ramanathan STTP CAD 2011Mesh modeling and processing.
3D Shape Histograms for Similarity Search and Classification in Spatial Databases. Mihael Ankerst,Gabi Kastenmuller, Hans-Peter-Kriegel,Thomas Seidl Univ.
Chapter 8 Content-Based Image Retrieval. Query By Keyword: Some textual attributes (keywords) should be maintained for each image. The image can be indexed.
By: Bryan Bonvallet Nikolla Griffin Advisor: Dr. Jia Li
Xianfeng Gu, Yaling Wang, Tony Chan, Paul Thompson, Shing-Tung Yau
Recognizing Objects in Range Data Using Regional Point Descriptors A. Frome, D. Huber, R. Kolluri, T. Bulow, and J. Malik. Proceedings of the European.
Probabilistic Fingerprints for Shapes Niloy J. MitraLeonidas Guibas Joachim GiesenMark Pauly Stanford University MPII SaarbrückenETH Zurich.
Shape Analysis and Retrieval ( ) (Michael) Misha Kazhdan.
Reverse Engineering Niloy J. Mitra.
Rotation Invariant Spherical Harmonic Representation of 3D Shape Descriptors Michael Kazhdan Thomas Funkhouser Szymon Rusinkiewicz Princeton University.
Robert Osada, Tom Funkhouser Bernard Chazelle, and David Dobkin Princeton University Matching 3D Models With Shape Distributions.
Matching and Recognition in 3D. Moving from 2D to 3D – Some Things are Easier No occlusion (but sometimes missing data instead)No occlusion (but sometimes.
Reflective Symmetry Detection in 3 Dimensions
Selecting Distinctive 3D Shape Descriptors for Similarity Retrieval Philip Shilane and Thomas Funkhouser.
The Princeton Shape Benchmark Philip Shilane, Patrick Min, Michael Kazhdan, and Thomas Funkhouser.
Shape Descriptors I Thomas Funkhouser CS597D, Fall 2003 Princeton University Thomas Funkhouser CS597D, Fall 2003 Princeton University.
Video Google: Text Retrieval Approach to Object Matching in Videos Authors: Josef Sivic and Andrew Zisserman University of Oxford ICCV 2003.
Matching and Recognition in 3D. Moving from 2D to 3D Some things harderSome things harder – Rigid transform has 6 degrees of freedom vs. 3 – No natural.
Scale Invariant Feature Transform (SIFT)
Shape Matching and Anisotropy Michael Kazhdan, Thomas Funkhouser, and Szymon Rusinkiewicz Princeton University Michael Kazhdan, Thomas Funkhouser, and.
The Planar-Reflective Symmetry Transform Princeton University.
Multiple Object Class Detection with a Generative Model K. Mikolajczyk, B. Leibe and B. Schiele Carolina Galleguillos.
Spatial and Temporal Databases Efficiently Time Series Matching by Wavelets (ICDE 98) Kin-pong Chan and Ada Wai-chee Fu.
A Search Engine for 3D Models THOMAS FUNKHOUSER, PATRICK MIN, MICHAEL KAZHDAN, JOYCE CHEN, ALEX HALDERMAN, and DAVID DOBKIN Princeton University and DAVID.
Introduction --Classification Shape ContourRegion Structural Syntactic Graph Tree Model-driven Data-driven Perimeter Compactness Eccentricity.
AdvisorStudent Dr. Jia Li Shaojun Liu Dept. of Computer Science and Engineering, Oakland University 3D Shape Classification Using Conformal Mapping In.
Description of 3D-Shape Using a Complex Function on the Sphere Dejan Vranić and Dietmar Saupe Slides prepared by Nat for CS
Alignment Introduction Notes courtesy of Funk et al., SIGGRAPH 2004.
Orthogonal moments Motivation for using OG moments Stable calculation by recurrent relations Easier and stable image reconstruction - set of orthogonal.
1 Faculty of Information Technology Generic Fourier Descriptor for Shape-based Image Retrieval Dengsheng Zhang, Guojun Lu Gippsland School of Comp. & Info.
FRIP: A Region-Based Image Retrieval Tool Using Automatic Image Segmentation and Stepwise Boolean AND Matching IEEE TRANSACTIONS ON MULTIMEDIA, VOL. 7,
TEMPLATE BASED SHAPE DESCRIPTOR Raif Rustamov Department of Mathematics and Computer Science Drew University, Madison, NJ, USA.
Shape Matching for Model Alignment 3D Scan Matching and Registration, Part I ICCV 2005 Short Course Michael Kazhdan Johns Hopkins University.
Alignment and Matching
Shape Matching with Occlusion in Image Databases Aristeidis Diplaros Euripides G.M. Petrakis Evangelos Milios Technical University of Crete.
Shape Based Image Retrieval Using Fourier Descriptors Dengsheng Zhang and Guojun Lu Gippsland School of Computing and Information Technology Monash University.
Shape Analysis and Retrieval Statistical Shape Descriptors Notes courtesy of Funk et al., SIGGRAPH 2004.
Shape Analysis and Retrieval Structural Shape Descriptors Notes courtesy of Funk et al., SIGGRAPH 2004.
Axial Flip Invariance and Fast Exhaustive Searching with Wavelets Matthew Bolitho.
Signal Processing and Representation Theory Lecture 4.
Some Shape Descriptors for 3D Visual Objects
Shape Descriptors Thomas Funkhouser and Michael Kazhdan Princeton University Thomas Funkhouser and Michael Kazhdan Princeton University.
Signal Processing and Representation Theory Lecture 2.
Signal Processing and Representation Theory Lecture 3.
Introduction --Classification Shape ContourRegion Structural Syntactic Graph Tree Model-driven Data-driven Perimeter Compactness Eccentricity.
Dengsheng Zhang and Melissa Chen Yi Lim
Local invariant features Cordelia Schmid INRIA, Grenoble.
Event retrieval in large video collections with circulant temporal encoding CVPR 2013 Oral.
Reconstruction of Solid Models from Oriented Point Sets Misha Kazhdan Johns Hopkins University.
The Planar-Reflective Symmetry Transform Princeton University.
2004/03/03Sheun-Huei Guan, CML, NTU1 3D Model Retrieval After Shape Distributions.
Methods for 3D Shape Matching and Retrieval
Interactive Discovery and Semantic Labeling of Patterns in Spatial Data David Blei, Christiane Fellbaum, Adam Finkelstein, and Thomas Funkhouser Princeton.
Partial Shape Matching. Outline: Motivation Sum of Squared Distances.
Spatial Range Querying for Gaussian-Based Imprecise Query Objects Yoshiharu Ishikawa, Yuichi Iijima Nagoya University Jeffrey Xu Yu The Chinese University.
Skeleton Extraction of 3D Objects by Radial Basis Functions for Content-based Retrieval in MPEG-7 Ming Ouhyoung Fu-Che Wu, Wan-Chun Ma, Communication and.
1 Faculty of Information Technology Enhanced Generic Fourier Descriptor for Object-Based Image Retrieval Dengsheng Zhang, Guojun Lu Gippsland School of.
Spherical Extent Functions. Spherical Extent Function.
Hough Transform Elegant method for direct object recognition
3D MODELLING PART-4 Wikitechy.
Tutorial 2 Biological shape descriptors
Benchmarking CAD Search Techniques
Shape Analysis and Retrieval
Improving Retrieval Performance of Zernike Moment Descriptor on Affined Shapes Dengsheng Zhang, Guojun Lu Gippsland School of Comp. & Info Tech Monash.
Presentation transcript:

Harmonic 3D Shape Matching Michael Kazhdan Thomas Funkhouser Princeton University Michael Kazhdan Thomas Funkhouser Princeton University

Motivation Large databases of 3D models Mechanical CAD (National Design Repository) Molecular Biology (Audrey Sanderson) Computer Graphics (Princeton 3D Search Engine)

Goal Find 3D models with similar shapes 3D ModelShape Descriptor Model Database Nearest Neighbor

Research Challenge Need shape descriptor that is: DiscriminatingDiscriminating Concise to storeConcise to store Quick to computeQuick to compute Efficient to matchEfficient to match Need shape descriptor that is: DiscriminatingDiscriminating Concise to storeConcise to store Quick to computeQuick to compute Efficient to matchEfficient to match 3D ModelShape Descriptor Model Database Nearest Neighbor

Research Challenge Finding a 3D shape descriptor that is: DiscriminatingDiscriminating Concise to storeConcise to store Quick to computeQuick to compute Efficient to matchEfficient to match Finding a 3D shape descriptor that is: DiscriminatingDiscriminating Concise to storeConcise to store Quick to computeQuick to compute Efficient to matchEfficient to match 3D ModelShape Descriptor Nearest Neighbor Model Database

Research Challenge Finding a 3D shape descriptor that is: DiscriminatingDiscriminating Concise to storeConcise to store Quick to computeQuick to compute Efficient to matchEfficient to match Finding a 3D shape descriptor that is: DiscriminatingDiscriminating Concise to storeConcise to store Quick to computeQuick to compute Efficient to matchEfficient to match Model Database Nearest Neighbor Shape Descriptor 3D Model

Research Challenge Finding a 3D shape descriptor that is: DiscriminatingDiscriminating Concise to storeConcise to store Quick to computeQuick to compute Efficient to matchEfficient to match Finding a 3D shape descriptor that is: DiscriminatingDiscriminating Concise to storeConcise to store Quick to computeQuick to compute Efficient to matchEfficient to match 3D Model Nearest Neighbor Shape Descriptor Model Database

Research Challenge Finding a 3D shape descriptor that is: DiscriminatingDiscriminating Concise to storeConcise to store Quick to computeQuick to compute Efficient to matchEfficient to match Finding a 3D shape descriptor that is: DiscriminatingDiscriminating Concise to storeConcise to store Quick to computeQuick to compute Efficient to matchEfficient to match 3D Model Nearest Neighbor Shape Descriptor Model Database Many possible alignments

3D Model Matching Approaches Search over all possible alignments Too slow for large databaseToo slow for large database Search over all possible alignments Too slow for large databaseToo slow for large database min

3D Model Matching Approaches Search over all possible alignments Too slow for large databaseToo slow for large database Normalize alignment (e.g., with moments) OK for translation and scale, not for rotationOK for translation and scale, not for rotation Search over all possible alignments Too slow for large databaseToo slow for large database Normalize alignment (e.g., with moments) OK for translation and scale, not for rotationOK for translation and scale, not for rotation PCA Aligned Models

3D Model Matching Approaches Search over all possible alignments Too slow for large databaseToo slow for large database Normalize alignment (e.g., with moments) OK for translation and scale, not for rotationOK for translation and scale, not for rotation Build alignment invariance into descriptor Previous methods not very discriminatingPrevious methods not very discriminating Search over all possible alignments Too slow for large databaseToo slow for large database Normalize alignment (e.g., with moments) OK for translation and scale, not for rotationOK for translation and scale, not for rotation Build alignment invariance into descriptor Previous methods not very discriminatingPrevious methods not very discriminating Shape Histograms [Ankerst et al., 1999] Shape Histograms [Ankerst et al., 1999]

Outline IntroductionApproachImplementation Experimental Results Conclusion and Future Work IntroductionApproachImplementation Experimental Results Conclusion and Future Work

Our Approach Harmonic 3D shape descriptor Decompose 3D shapes into irreducible set of rotation independent componentsDecompose 3D shapes into irreducible set of rotation independent components Store “how much” of the model resides in each componentStore “how much” of the model resides in each component Harmonic 3D shape descriptor Decompose 3D shapes into irreducible set of rotation independent componentsDecompose 3D shapes into irreducible set of rotation independent components Store “how much” of the model resides in each componentStore “how much” of the model resides in each component 3D Model Shape Descriptor Rotation Independent Components

Our Approach Harmonic 3D shape descriptor Decompose 3D shapes into irreducible set of rotation independent componentsDecompose 3D shapes into irreducible set of rotation independent components Store “how much” of the model resides in each componentStore “how much” of the model resides in each component Harmonic 3D shape descriptor Decompose 3D shapes into irreducible set of rotation independent componentsDecompose 3D shapes into irreducible set of rotation independent components Store “how much” of the model resides in each componentStore “how much” of the model resides in each component Concentric Spheres 3D Model Shape Descriptor Rotation Independent Components

Our Approach Harmonic 3D shape descriptor Decompose 3D shapes into irreducible set of rotation independent componentsDecompose 3D shapes into irreducible set of rotation independent components Store “how much” of the model resides in each componentStore “how much” of the model resides in each component Harmonic 3D shape descriptor Decompose 3D shapes into irreducible set of rotation independent componentsDecompose 3D shapes into irreducible set of rotation independent components Store “how much” of the model resides in each componentStore “how much” of the model resides in each component Frequency Decomposition 3D Model Shape Descriptor Rotation Independent Components

Our Approach Harmonic 3D shape descriptor Decompose 3D shapes into irreducible set of rotation independent componentsDecompose 3D shapes into irreducible set of rotation independent components Store “how much” of the model resides in each componentStore “how much” of the model resides in each component Harmonic 3D shape descriptor Decompose 3D shapes into irreducible set of rotation independent componentsDecompose 3D shapes into irreducible set of rotation independent components Store “how much” of the model resides in each componentStore “how much” of the model resides in each component 3D Model Shape Descriptor Rotation Independent Components Amplitudes

Outline IntroductionApproachImplementation Experimental Results Conclusion and Future Work IntroductionApproachImplementation Experimental Results Conclusion and Future Work

Voxelization Convert polygonal model to 3D voxel grid Rasterize surfaces (no solid reconstruction)Rasterize surfaces (no solid reconstruction) Normalize for translation and scaleNormalize for translation and scale Convert polygonal model to 3D voxel grid Rasterize surfaces (no solid reconstruction)Rasterize surfaces (no solid reconstruction) Normalize for translation and scaleNormalize for translation and scale 3D Model3D Voxel Grid

Spherical Decomposition Intersect with concentric spheres

Frequency Decomposition Represent each spherical function as a sum of different frequencies Spherical Functions Frequency Components

Fourier Analysis Circular Function

Fourier Analysis +++=+ … Cosine/Sine Decomposition Circular Function

Fourier Analysis = +++ Constant =+ … Frequency Decomposition Circular Function

Fourier Analysis = Constant1 st Order =+ … + Frequency Decomposition Circular Function

Fourier Analysis = Constant1 st Order2 nd Order =+ … + Frequency Decomposition Circular Function

Fourier Analysis = Constant1 st Order2 nd Order3 rd Order =+ … + … + Frequency Decomposition Circular Function

++++ … + Fourier Analysis = +++ Constant1 st Order2 nd Order3 rd Order + … Frequency Decomposition = Amplitudes invariant to rotation Circular Function

Harmonic Analysis Spherical Function

Harmonic Analysis +++=+ … Spherical Function Harmonic Decomposition

Harmonic Analysis = Constant1 st Order2 nd Order3 rd Order =+ … + … Spherical Function

Building Shape Descriptor Store “how much” (L 2 -norm) of the shape resides in each frequency of each sphere Frequency Decomposition Amplitudes Harmonic Shape Descriptor

Matching Model similarity defined as L 2 -distance between their descriptors Bounds proximity of voxel grids over all rotationsBounds proximity of voxel grids over all rotations Model similarity defined as L 2 -distance between their descriptors Bounds proximity of voxel grids over all rotationsBounds proximity of voxel grids over all rotations, = Sim

Outline IntroductionApproachImplementation Experimental Results Conclusion and Future Work IntroductionApproachImplementation Experimental Results Conclusion and Future Work

Princeton 3D Search Engine Query

Retrieval Experiment Viewpoint “household” database 1,890 models, 85 classes 153 dining chairs25 livingroom chairs16 beds12 dining tables 8 chests28 bottles39 vases36 end tables

Retrieval Results Precision-recall curve (mean for all queries) Recall Precision Random 3D Harmonics Query

Retrieval Results Precision versus recall (mean for all queries) Recall Precision 3D Harmonics (Our Method) D2 Shape Distributions [Osada et al., 2001] Shape Histograms [Ankerst, 1999] EGI [Horn, 1984] Moments [Elad et al., 2001] Random 1

Retrieval Results Precision versus recall (mean for all queries) Recall Precision 3D Harmonics (Our Method) D2 Shape Distributions [Osada et al., 2001] Shape Histograms [Ankerst, 1999] EGI [Horn, 1984] Moments [Elad et al., 2001] Random 1

Retrieval Results Precision versus recall (mean for all queries) Recall Precision 3D Harmonics (Our Method) D2 Shape Distributions [Osada et al., 2001] Shape Histograms [Ankerst, 1999] EGI [Horn, 1984] Moments [Elad et al., 2001] Random 1

Retrieval Results Precision versus recall (mean for all queries) Recall Precision 3D Harmonics (Our Method) D2 Shape Distributions [Osada et al., 2001] Shape Histograms [Ankerst, 1999] EGI [Horn, 1984] Moments [Elad et al., 2001] Random 1

Summary and Conclusion Harmonic shape descriptor is a rotation invariant representation that is: Discriminating (46%-245% better than others tested)Discriminating (46%-245% better than others tested) Concise to store (2048 bytes)Concise to store (2048 bytes) Quick to compute (1-5 seconds)Quick to compute (1-5 seconds) Efficient to match (0.45 seconds: 20,000 model DB)Efficient to match (0.45 seconds: 20,000 model DB) Harmonic shape descriptor is a rotation invariant representation that is: Discriminating (46%-245% better than others tested)Discriminating (46%-245% better than others tested) Concise to store (2048 bytes)Concise to store (2048 bytes) Quick to compute (1-5 seconds)Quick to compute (1-5 seconds) Efficient to match (0.45 seconds: 20,000 model DB)Efficient to match (0.45 seconds: 20,000 model DB)

Future Work Extensions Partial object matching?Partial object matching? Other 3D shape functions?Other 3D shape functions? Other applications Molecular biologyMolecular biology MedicineMedicine PaleontologyPaleontology ForensicsForensicsExtensions Partial object matching?Partial object matching? Other 3D shape functions?Other 3D shape functions? Other applications Molecular biologyMolecular biology MedicineMedicine PaleontologyPaleontology ForensicsForensics Query

Thank You Funding National Science FoundationNational Science Foundation Sloan FoundationSloan FoundationPeople Bernard Chazelle, David Dobkin, David Jacobs, David Kazhdan, Allison Klein, Patrick Min, Szymon Rusinkiewicz, Peter Sarnak, Julianna TymoczkoBernard Chazelle, David Dobkin, David Jacobs, David Kazhdan, Allison Klein, Patrick Min, Szymon Rusinkiewicz, Peter Sarnak, Julianna TymoczkoFunding National Science FoundationNational Science Foundation Sloan FoundationSloan FoundationPeople Bernard Chazelle, David Dobkin, David Jacobs, David Kazhdan, Allison Klein, Patrick Min, Szymon Rusinkiewicz, Peter Sarnak, Julianna TymoczkoBernard Chazelle, David Dobkin, David Jacobs, David Kazhdan, Allison Klein, Patrick Min, Szymon Rusinkiewicz, Peter Sarnak, Julianna Tymoczko

Analysis The Harmonic Descriptor is not invertible Different spheres rotate independentlyDifferent spheres rotate independently Different orders rotate independentlyDifferent orders rotate independently Rotations are not transitive within an orderRotations are not transitive within an order The Harmonic Descriptor is not invertible Different spheres rotate independentlyDifferent spheres rotate independently Different orders rotate independentlyDifferent orders rotate independently Rotations are not transitive within an orderRotations are not transitive within an order

Analysis The Harmonic Descriptor is not invertible Different spheres rotate independentlyDifferent spheres rotate independently Different orders rotate independentlyDifferent orders rotate independently Rotations are not transitive within an orderRotations are not transitive within an order The Harmonic Descriptor is not invertible Different spheres rotate independentlyDifferent spheres rotate independently Different orders rotate independentlyDifferent orders rotate independently Rotations are not transitive within an orderRotations are not transitive within an order l=0 l=1 l=2 l=3

Inter-Radial Coherence Force same orders on different spheres to rotate together by setting up the rotation invariant matrix M k with: where f k,j is the k-th order component of the restriction of f to the j-th radius. (The diagonal is precisely the collection of L 2 -norms.) Force same orders on different spheres to rotate together by setting up the rotation invariant matrix M k with: where f k,j is the k-th order component of the restriction of f to the j-th radius. (The diagonal is precisely the collection of L 2 -norms.)

Polygon Rasterization Rasterize using the Euclidean Distance Transform to measure how much models miss by: Polygonal Model Voxel Model