Schloss Dagstuhl, September 2014 Shape Representation Carlo H. Séquin University of California, Berkeley User-Guided Inverse 3D Modeling.

Slides:



Advertisements
Similar presentations
Kinematics & Grasping Need to know: Representing mechanism geometry Standard configurations Degrees of freedom Grippers and graspability conditions Goal.
Advertisements

Interactive Inverse 3D Modeling James Andrews Hailin Jin Carlo Séquin.
To the 4th Dimension – and beyond!
MATHIEU GAUTHIER PIERRE POULIN LIGUM, DEPT. I.R.O. UNIVERSITÉ DE MONTRÉAL GRAPHICS INTERFACE 2009 Preserving Sharp Edges in Geometry Images.
Interactive Inverse 3D Modeling
University of California, Berkeley
Surfaces Chiew-Lan Tai. Surfaces 2 Reading Required Hills Section Hearn & Baker, sections 8.11, 8.13 Recommended Sections 2.1.4, , 3D Computer.
Carlo H. Séquin u (Descriptive) Geometry – my love since high school.
James Hamlin and Carlo H. Séquin University of California, Berkeley Ribbed Surfaces for Art, Architecture, Visualization Computer-Aided Design and Applications.
CHS UCB CS285 Sculpture Generator I Carlo H. Séquin University of California, Berkeley In Collaboration with Brent Collins Gower, Missouri.
CHS UCB ZURICH, Aug ART -- MATH Carlo H. Séquin University of California, Berkeley.
CHS UCB BID 02/02/02 Parameterized Sculpture Design Carlo H. Séquin University of California, Berkeley.
Graduate Student Visit Day, 03/12/07 Aesthetic Engineering Artistic Geometry Carlo H. Séquin University of California, Berkeley.
ETH, July 21, 2008 Scherk’s 2 nd Minimal Surface, Snowsculpting, EuroGraphics Award Carlo H. Séquin University of California, Berkeley.
Modelling. Outline  Modelling methods  Editing models – adding detail  Polygonal models  Representing curves  Patched surfaces.
IMA Tutorial, Instantaneous Motions - Applications to Problems in Reverse Engineering and 3D Inspection H. Pottmann.
Carlo H. Séquin u (Descriptive) Geometry – my love since high school.
Leonardo Meeting, SETI Institute, Feb. 10, 2010
AAAS 2001, San Francisco Procedurally Defined Geometrical Sculptures Carlo H. Séquin University of California, Berkeley Brent Collins Gower, Missouri.
CHS UCB CS285 Designing Viae Globi (Roads on a Sphere) Carlo H. Séquin University of California, Berkeley Inspired by Brent Collins Gower, Missouri.
Some Material on Swept Solids and Surfaces of Revolution From Chapter 10 of Mortenson Sections 10.5 – 10.6 Geometric Modeling
HCI Séquin’s Professional Interests Art-Math CAD-CAMGraphics Design ArtGeometry.
OZONE Meeting, SIGGRAPH 2003 mathematics & art Carlo H. Séquin U.C. Berkeley.
CONTACT, March , 2003 Art, Math, Computers, and Creativity Carlo Séquin, University of California, Berkeley.
CS285 Designing Viae Globi (Roads on a Sphere) Carlo H. Séquin University of California, Berkeley Inspired by Brent Collins Gower, Missouri.
BINARY MORPHOLOGY and APPLICATIONS IN ROBOTICS. Applications of Minkowski Sum 1.Minkowski addition plays a central role in mathematical morphology 2.It.
USF -- February 2001 Art, Math, and Sculpture Connecting Computers and Creativity Carlo H. Séquin University of California, Berkeley.
CS 184, 5/11/2009 My Own Use of Computer Graphics Carlo H. Séquin University of California, Berkeley.
Southern Oregon University, May 2003 Surface Optimization and Aesthetic Engineering Carlo Séquin, University of California, Berkeley.
E92 -- October 2002 Art, Math, and Sculpture Connecting Computers and Creativity Carlo H. Séquin EECS Computer Science Division University of California,
Leonardo Meeting, San Francisco, May 12, 2008
BID Lab Opening, May 14, Design, Technology, and the Human Experience (the positive side) Carlo H. Séquin Computer Science Division.
CS285 Designing Viae Globi (Roads on a Sphere) Carlo H. Séquin University of California, Berkeley Inspired by Brent Collins Gower, Missouri.
3-D Modeling Concepts V part 2.
Modeling and representation 1 – comparative review and polygon mesh models 2.1 Introduction 2.2 Polygonal representation of three-dimensional objects 2.3.
Laplacian Surface Editing
Virtual reality. Tasks 3D digital model from planes 3D digital model of existing objects Office work Field observations Solid modeling Photogrammetry.
1/20 Obtaining Shape from Scanning Electron Microscope Using Hopfield Neural Network Yuji Iwahori 1, Haruki Kawanaka 1, Shinji Fukui 2 and Kenji Funahashi.
COMPUTER AIDED DESIGN -(CAD)-3
Art & Science UCB DeCal, September 17, 2013 Carlo H. Séquin University of California, Berkeley.
Model Construction 김 성 남. Contents What is Modeling? Model Construction - Interactive modeling tools - Scanning tools - Computer vision -
Non-Linear Models. Non-Linear Growth models many models cannot be transformed into a linear model The Mechanistic Growth Model Equation: or (ignoring.
Lecture 5: Curves And Surfaces, and Geometric Modeling
Schloss Dagstuhl, September 2014 Shape Representation Carlo H. Séquin University of California, Berkeley Slicing Less than Perfect B-Reps and the Winding-Number.
PREPARED BY: SAMERA BINTI SAMSUDDIN SAH SEM /2012 (NOV 2011)
Adobe Retreat, November 12, 2013 Weak Links in the Chain from Concept to Construction Carlo H. Séquin University of California, Berkeley.
Geometric Modeling. Volumetric o Collection device obtains regular grid of measurement values Examples: CT, MRI, PET, Ultrasound o Values are interpreted/visualized.
Art-in-Science (and Science-in-Art) Feb. 27, 2014 Carlo H. Séquin University of California, Berkeley Art of Minimal Energy (and of Maximal Beauty?)
BID Lunch, March 5, 2013 Last Week at MWSU EECS Computer Science Division University of California, Berkeley Carlo H. Séquin.
URAP, September 16, 2013 Carlo H. Séquin University of California, Berkeley The Beauty of Knots.
Simons Center, July 30, 2012 Carlo H. Séquin University of California, Berkeley Artistic Geometry -- The Math Behind the Art.
Schloss Dagstuhl, September 2014 Shape Representation Carlo H. Séquin University of California, Berkeley “LEGO Knot” and an Optimization Problem in a High-Dimensional.
EECS Computer Science Division University of California, Berkeley
University of California, Berkeley
Non-Linear Models. Non-Linear Growth models many models cannot be transformed into a linear model The Mechanistic Growth Model Equation: or (ignoring.
12/24/2015 A.Aruna/Assistant professor/IT/SNSCE 1.
Written by Changhyun, SON Chapter 5. Introduction to Design Optimization - 1 PART II Design Optimization.
By: David Gelbendorf, Hila Ben-Moshe Supervisor : Alon Zvirin
Application: Multiresolution Curves Jyun-Ming Chen Spring 2001.
LASER, December 10, 2014 Carlo H. Séquin University of California, Berkeley "Hyperbolic Hexagon" to "Evolving Trefoil“ my 20-year collaboration with Brent.
1 Sheet Metal Design (III) Hong RuJin September, 10, 2000.
CDS 301 Fall, 2008 Domain-Modeling Techniques Chap. 8 November 04, 2008 Jie Zhang Copyright ©
Creation of the Eurographics Award Trophy Carlo H. Séquin EECS Computer Science Division University of California, Berkeley.
COMPUTER AIDED ENGINEERING Integrierte Optimierung mit ANSA, LS-OPT und META – Infotag Morphing with.
3-D Modeling Concepts V part 2.
3-D Modeling Concepts V part 2.
Ribbed Surfaces for Art, Architecture, Visualization
3-D Modeling Concepts V part 2.
Procedurally Defined Geometrical Sculptures
“Procedural Solid Modeling” or
Presentation transcript:

Schloss Dagstuhl, September 2014 Shape Representation Carlo H. Séquin University of California, Berkeley User-Guided Inverse 3D Modeling

Dagstuhl Discussion Points: Shape representation issues at the start and conclusion of designing RP models u Focus on HCI difficulties and CAD problems, at the start and end of a design / modeling project: u How to get started?  How to get your ideas into the CAD system.  I like to obtain a high-level parameterized procedural description.

Background: 40 Years of Design CCD TV Camera (1973) Soda Hall (1992) RISC 1 Computer Chip (1984) Octa-Gear (CyberBuild,1999)

Recent Designs and Models “Evolving Trefoil” Double-Boy Klein Bottle 4-dimensional 120-Cell “Inverted double-sock” Klein bottle

Brent Collins “Hyperbolic Hexagon II”

“Hyperbolic Hexagon” by B. Collins u 6 saddles in a ring u = “wound up” 6-story Scherk tower u 6 holes passing through symmetry plane at ±45º u Discussion: What if … l we added more stories ? l or introduced a twist before closing the ring ?

Brent Collins’ Prototyping Process Armature for the Hyperbolic Heptagon Mockup for the Saddle Trefoil Time-consuming ! (1-3 weeks)

Sculpture Generator I, GUI

Captured Concept straight or twisted “Scherk Tower”“Scherk-Collins Toroids” Scherk’s 2 nd minimal surface

Base Geometry: One Scherk Story u Hyperbolic Slices ==> Triangle Strips u Pre-computed -- then warped into toroid

Shapes from Sculpture Generator I

A Simple Scherk-Collins Toroid Parameters: (genome) u branches = 2 u stories = 1 u height = 5.00 u flange = 1.00 u thickness = 0.10 u rim_bulge = 1.00 u warp = u twist = 90 u azimuth = 90 u textr_tiles = 3 u detail = 8

Generalizing the Above Approach  “User-guided Inverse 3D Modeling” u Start from some previously defined artifact that might be reverse engineered. u Create a high-level description with suitable parameterization. u Important: Include the designer in the reverse-engineering loop!

Possible Starting Points u A (re-)design rarely starts in a vacuum! u There may be available: l a physical artifact to be 3D-scanned, l a set of photographs of an inspirational object, l a wire-mesh (.STL file) of a previous design. u We want to obtain a high-level description, l suitably modularized (handles, body, feet…), l and parameterized (handle diameter; body profile).

Modular Reverse Engineering u Extract parameterized descriptions, module by module. u In each case choose a representation that best enables the intended re-design. u Use plausible, commonly used CAD constructs: l CSG l Quadrics l Extrusions l Rotational Sweeps l Progressive Sweeps

User-guided Inverse 3D Modeling ( Jimmy Andrews’ PhD thesis) u Let the user select a high-level model structure that is most useful for immediate re-design. Initial artifact Redesigns enabled by different imposed structure

Option 1: Varying Rotational Symmetry 3 fold4 fold20 fold Extract one sector; collapse/expand in polar coordinates.

Opt.2: Editing as Surface of Revolution Mesh is rotationally collapsed to yield a compound “cross-section”; This cross-section can then be edited, and this will affect the whole mesh.

Opt.3: Extraction as a Progressive Sweep 20-story Scherk chain Revised trefoil sweep path

User-Guided Fitting Modules (Andrews) u Stationary sweeps: (Surfaces of revolution, helices, etc) u Progressive sweeps: u Quadrics: u Freeform surfaces: u CSG modules: …

… then assume point or facet belongs to sweep. If normal is perpendicular to velocity field: (simple motion)(simple velocity field) Defined by a simple sweep motion, with a fixed axis (e.g. revolution, helix, spiral) Stationary Sweeps

Fitting Algorithm: u Find velocity field that fits marked data points: Minimize (subject to constraint): u Grow the region by adding more fitting points u Repeat (typically converges in 2-3 iterations) [ Pottmann, Lee, and Randrup, 98 ]

Interactive Surface Editing A rotational sweep around the z-axis is specified. A “thick profile” is extracted by collapsing φ-component. Portions of the “thick profile” can be selected and moved; the corresponding surface elements move radially: (a) the whole nose & cheeks area is enlarged; (b) only the nose tips are stretched. (a) (b)

Progressive Sweeps u More parameters: Make incremental local adjustments u Allow more complex cross-section transformations (translation, rotation, scaling) u User stroke provides initial guess u Fit by iteratively extending and optimizing

Progressive Sweep Fitting u Starting from user stroke, optimize cross section u Iteratively extend and re-optimize u Stop when best further extension would have high error [Andrews, Joshi, Sequin 2011]

Flexibility of Progressive Sweeps u Capture the parameterized procedural description that best fits the users re-design plans. Yellow strokes (#1) defines the start of a progressive sweep. An optional 2 nd stroke extends or restricts the sweep range.

Versatility of Progressive Sweeps Different starting strokes and different error tolerances result in a wide variety of possible extracted sweeps. Sweep path and profiles can be edited independently. Surface details with respect to the extracted sweep can be conserved and reapplied after any editing moves, or they can be ignored or smoothed out.

Preserving Surface Details Modifying the sweep path & scaling, while preserving surface details:

Freeform Surfaces: Laplacian Surface Editing This is the workhorse to handle “everything else”: - Smoothing out unwanted details (ear). - Keeping model elements smoothly connected (leg).

System Pipeline Unstructured mesh Photos 3D scans Input Data Editable Model Model Hierarchy & Re-fitting Clean User- guided fitting modules Nice rendering STL for RP OBJ for CAD Redesigned output

Salient Points u Designer knows best in what direction she wants to push a particular design. u That knowledge determines preferred representation. u Use a GUI where the designer can easily convey that information to the extraction system. u See results immediately and try to make desired modification (if it does not work as intended do a somewhat different extraction)  Instant feedback.  Your ideas and comments ??