CAMPLAN: A Camera Planning Agent

1
CAMPLAN A Camera Planning Agent

• By Nicolas Halper, Patrick Olivier, AAAI Workshop
  on Smart Graphics, Stanford, March 20-22, 2000

2
Outline

• Introduction
• GA Algorithm
• Selection of Shot Properties
• Evaluation of Shot Properties
• Image Generation
• System Evaluation
• QA

3
Introduction

• States of CAMPLAN System
• Selection of Shot Objectives
• Evaluation of Objectives
• Acquisition of a Camera State
• CAMPLAN uses GA algorithm

4
Outline

• Introduction
• GA Algorithm
• Selection of Shot Properties
• Evaluation of Shot Properties
• Image Generation
• System Evaluation
• QA

5
GA Algorithm

• Parameters Encoding
• Fitness Function
• Selection Function
• Reproduction
• Crossover
• Mutation

6
Outline

• Introduction
• GA Algorithm
• Selection of Shot Properties
• Evaluation of Shot Properties
• Image Generation
• System Evaluation
• QA

7
Selection of Shot Properties

• Reference Frame
• Viewplane the surface which extents infinitely
  in all directions perpendicular to the view
  direction of the camera
• Viewport subset of the viewplane with a defined
  bounded region
• Viewpoint the location of the camera

8
Selection of Shot Properties

• Shot Types
• Position the location an element of the scene,
  with respect to the reference frame
• Size how large an object is to appear.
• Orientation enable the user to view certain
  sides of an object
• Visibility occlusion relationship between objects

9
Outline

• Introduction
• GA Algorithm
• Selection of Shot Properties
• Evaluation of Shot Properties
• Image Generation
• System Evaluation
• QA

10
Evaluation of Shot Properties

• Projected Extents
• Visibility

11
Outline

• Introduction
• GA Algorithm
• Selection of Shot Properties
• Evaluation of Shot Properties
• Image Generation
• System Evaluation
• QA

12
Image Generation

• Chromosomes position/3, angles/3 and field of
  view
• The top 90 of the population survive to the next
  generation according to evaluation of shot
  properties
• The remaining 10 are re-generated by random
  crossover and/or mutation of chromosomes

13
Outline

• Introduction
• GA Algorithm
• Selection of Shot Properties
• Evaluation of Shot Properties
• Image Generation
• System Evaluation
• QA

14
System Evaluation
15
System Evaluation
16
System Evaluation

• CenterShot "roof"
• Unoccluded "roof" 100 tolerance 0
• SpanXorY "roof" 2 tolerance 1.2
• PositionRelativeX "pillar01" before before "roof"
• PositionRelativeX "pillar03" after after "roof"
  "grid"
• OccludedInViewportBy "pillar01" 100 tolerance 99

17
System Evaluation

• CenterShot "Escher building"
• EntirelyInViewport "Escher building"
• OccludedInViewport "step high" 0 80
• OccludedInViewport "step low" 0 80 "step high"
• NotOccludedBy "step low" "step low"
• NotOccludedBy "step high"

18
System Evaluation
Start Position
End Position
19
System Evaluation
20
QA

• THANKS FOR YOUR LISTENING!