1. ERC Expressive Seminar
March 7th
Models and Intuitive Modeling
Loïc Barthe
VORTEX group
IRIT – Université de Toulouse

2. Plan Context and introduction Intuitive modeling
Modeling with meshes only
Other representations then meshes
Sketching
On an ideal system ? Well some insights ...
But what about highly complex topology and more general objects ?
Conclusion

3. Displays / 3D Printers => 3D content => Geometric modeling

4. What to avoid for a modeling session ?
Boring and fastidious
No undo and unpredictable
Complex and non-intuitive

5. Intuitive modeling ? Ideally
System fonctionnalities fast and easy to learn / understand
Shape corresponding to user expectations when responding to user input
Freedom in shape variety
Obvious construction of the modeling process
Accessible interaction
Fast object creation
Interactive feedbacks

6. Surface Properties Ideally Continuity : C0, C1, C2,
Oscillations only if desired
Planar or not
Different level of details
Fair / elegant
Arbitrary genus

7. Triangle meshes
Fast and easy rendering of 3D objects require the use of triangle meshes
Virtual environments are essentially composed of triangle meshes

8. Triangle meshes
Whatever the modeling process, at the end, we want the surface to be fairly represented by a mesh, preferably 2D Manifold

9. Surface = a 2D Manifold Mesh
What it is good for :
Fast rendering
Represent complex objects
Multi-resolution modeling
Subdivision
Simplification
Deformations
Control net
Smooth surface
Finite element
Parametrizable
Remeshing
Material

10. Surface = a 2D Manifold Mesh
What it is NOT good for :
Changes in object topology
In/out queries
Collisions
Inter-objects
self-intersections
CSG
Highly deformable objects
Fluids

11. Meshes and Geometric Modeling
Non dedicated software 
Total freedom
But :
Blender
3DS Max
Maya

12. Meshes and Geometric Modeling
Dedicated software 
Dedicated to a specific context
User friendly => Mostly planar surfaces
User friendly => Organic / Smooth surfaces
SketchUp
Meshmixer
Sculptris

13. Mesh from another representation
Dedicated software 
Dedicated to a specific context
User friendly => Organic / Smooth surfaces
Teddy - sketching
Z-Brush - Z-Sphere

14. Zoom : Z-Brush vs Sculptris
Prototyping :
Z-Spheres vs Large deformations
Intuitive modeling
Complete
Surface properties
Globally C2 continuous
Genus ?
Intuitive modeling
Shape thickness
Surface properties
Continuity C>0
Genus 0

15. Zoom : Z-Brush vs Sculptris
Detail modeling with brushes and mesh deformations :
Subdivision surfaces vs Local remeshing
Intuitive modeling
Becomes complex
Surface properties
Complete
Intuitive modeling
Becomes complex
Surface properties
Continuity C>0
Genus 0

16. Zoom : Z-Brush vs Sculptris
Detail modeling with brushes and mesh deformations :
Subdivision surfaces vs Local remeshing
Quads
Multiresolution
Sharp edges
Isotropic refinement
Triangles
Adaptive resolution
Sharp edges
Anisotropic refinement

17. And Sketching ? Surface reconstruction from contours
Teddy's process : contour => chordal axis => Mesh
Incrementally add new object parts
Prototypes, Smooth surfaces

18. Sketching + Implicit Surfaces
Surface reconstruction from contours
General process : contour => chordal axis => Implicit Surface => Mesh
Prototypes, Smooth surfaces
Advantages :
Arbitrary genus
Easier and more stable implementation
Composition operators
But
Oscillations hard to avoid
Surface deformations non trivial
=>
=>

19. Implicit Surfaces Different families of field functions
RBF = contour + additional points + normals
HRBF = contour / lines (unorganized) + normals
Soft Objects = contour + medial axis
Oscillations
Thickness hard to control
Convolution Surfaces = contour + medial axis
But for oscillations :
But for thickness :
Triangle skeletons
Radially varying radius

20. Implicit surfaces in general
What they are NOT good for :
Fast rendering
Represent complex objects
Multi-resolution modeling
Subdivision
Simplification
Deformations
Control net
Smooth surface
Finite element
Parametrizable
Remeshing
Material

21. Implicit surfaces in general
What they are good for :
Changes in object topology
In/out queries
Collisions
Inter-objects
self-intersections
CSG
Highly deformable objects
Fluids
Represent smooth surfaces
Automatic thickness

22. An ideal system ? Z-spheres vs sketching
The first is the user providing the beginning of the reconstruction for the second
Could work together ?
« Dog Labrador drawing » in google image :

23. An ideal system ? Accept more approximative reconstruction
Oscillations
Fairness
Deformations for shape adjustment
Deformations for pose adjustment
Skinning at joints
Contact deformations
Non centered skeleton
Use animation skeleton to enable pose modification and skinning

24. An ideal system ?
But also, what about
Blocks
Mindcraft

25. An ideal system ?
Blocks / skeletons of different forms (cubes, cylinders, spheres, others ?)
Implicit surfaces for reconstructing the surface with a correct topology
High quality approximation with a mesh

26. Conclusion An ideal solution : nothing obvious
Apparently, different models / interactions for different steps :
prototype / adjustment / details
Sketching is the solution : it is one, but I guess, not alone
Z-Spheres : not really explored => nothing else to do ?
Explore other ways for building a 3D object
Intuitive modeling is still an open problem

27. Questions

28. Implicit surfaces Can they compete with meshes ? In general : NO
Is there an interest to use them alone ?
Yes for volume manipulation
Yes for fluid rendering
Yes for representing highly dynamic objects changing topology
Yes for fast prototyping by sketching
Otherwise … No.
But why not mixing meshes and implicit surfaces ? Good suggestion but
It is a complex problem
Several moderatly successful attempts in the years around 2000
Recent works on skinning will maybe be a first step.

29. Implicit surfaces
What could I expect from an efficient and effective research ?
TO DO
Fast rendering
Fine details representation
Parametrization
Efficient dual mesh/implicit-surface representation
Would allow you to end up with
A robust functional representation
An efficient high level model for mesh generation, deformation and topology modifications

30. Parametric patches They will not help you :
They share the same limitations
They have aditional limitations