Efficient modeling of entangled details for natural scenes Eric - - PowerPoint PPT Presentation

Efficient modeling of entangled details for natural scenes

Eric Guérin, Eric Galin, François Grosbellet Adrien Peytavie, Jean-David Génevaux LIRIS – CNRS – France

1

PG 2016

Introduction

Introduction Method Results Conclusion

2

PG 2016

Context/problem

Introduction Method Results Conclusion

• Natural scenes
• Numerous details
• Entangled
• Different kinds

⇒Tedious authoring

Stones Grass tufts Twigs

3

PG 2016

Related work

Simulations

Realistic Limited user control Does not scale + – –

Procedural

Control Specific Interpenetrations + – –

[Peytavie 2009] [Desbenoit 2006] [Hsu 2010] [Emilien 2015] [Grosbellet 2016]

Interactive editing

[Alsweis 2006]

Efficient Specific Memory + – –

Introduction Method Results Conclusion

4

PG 2016

Our approach

Realistic Efficient Not object-specific Light in memory Scalable Controllable + + + + + +

Introduction Method Results Conclusion

• Split the process into

two steps

1. Pre-compute collisions in a very dense tile 2. Fast Instantiation

• Multiple control types

5

• Key observation: if not regular, repetitions are not visible

PG 2016

The method

Introduction Method Results Conclusion

6

PG 2016

Pipeline in 2 steps

Input objects Ghost Tile Density fields

Collision graph Candidates

• 1. Ghost Tile

Generation

• 2. Instantiation

Introduction Method Results Conclusion

7

PG 2016

The method

Step 1 – Ghost tile construction

Introduction Method Results Conclusion

8

PG 2016

Ghost tile

Objects O Frames F Collision graph G Ghost Tile T

O

1 1

O ⊕ −y

1 1 2

O

2

O

2 1

O1 O 2

Candidates

O = F (O )

j j i i i

F

1 1

F

2 2

F

1 2

O ⊕ −x

2 2

Introduction Method Results Conclusion

9

PG 2016

Ghost tile construction

Introduction Method Results Conclusion

Algorithm

• 1. Pick a random frame in the tile
• 2. Compute intersections

inside the same tile in the neighbor tiles

• 3. If intersection, add two reciprocal arcs in the graph

⇒ Repeat (and use a spatial acceleration)

10

PG 2016

Distance between unions of spheres is easy

Collision detection

Leaf Stone

Introduction Method Results Conclusion

• Volume approximated by spheres
• Automatic or manual according to the

context

11

PG 2016

The method

Step 2 : Instantiation

Introduction Method Results Conclusion

12

PG 2016

Density description

Introduction Method Results Conclusion

Density functions

13

fi : R3 → R

Stones density function Twigs density functionfi

fj

PG 2016

Culling step

T ⊕ x Ghost Tile T f > 0 f = 0 f (a ) = 0

k j

ak

i

ak

j

f (a + t) > 0

k i

Introduction Method Results Conclusion

• Remove candidates whose density

vanishes at anchor point(s)

14

PG 2016

Instantiation step

Ghost Tile T T ⊕ −x O

3 1

T ⊕ x O ⊕ −x

4 1

O

2 1

O

1 1

O

4 1

Introduction Method Results Conclusion

• Select the highest priority candidate (green)
• Discard colliding candidates (orange)

15

PG 2016

Priorities

Introduction Method Results Conclusion

16

Random Altitude Distance to the boundary Distance to the boundary + partial filling

4.3k instances 4.5k instances 4.5k instances 3.8k instances

PG 2016

Results

Introduction Method Results Conclusion

17

PG 2016

Volumetric objects

Introduction Method Results Conclusion

Method that accounts for volumetric objects

18

PG 2016

Control over density

19k instances 18k instances 23k instances

Introduction Method Results Conclusion

Density functions to control the relative density of each object type

19

PG 2016

Complex scenes - Borie

Introduction Method Results Conclusion

63k flat stones Instantiation time 17s

20

PG 2016

Complex scenes - Field

Introduction Method Results Conclusion

4.3M straw instances Instantiation time 54.6s

21

PG 2016

Complex scenes - Meadow

Introduction Method Results Conclusion

Interactive authoring Standard stroke 1k instances in 1.5s

22

PG 2016

Conclusion

Introduction Method Results Conclusion

23

PG 2016

Conclusion

Introduction Method Results Conclusion

• Limitations
• No structure
• No animation
• General framework to model entangled

details

• Two steps
• 1. Offline pre-computation
• 2. Instantiation
• Efficient
• Handle interpenetrations
• Several user controls

24

PG 2016

Thank you for your attention!

See video and more on: http://liris.cnrs.fr/eric.guerin/efficient-modeling-of-entangled-details-for-natural-scenes/

25