Computer Science Honours Dissertation 499 Building MPEG4 Compliant 2D Cartoon Talking Heads

1
Computer Science Honours Dissertation
499Building MPEG-4 Compliant 2D Cartoon
Talking Heads

• Tonny Wijaya
• Department of Computing
• Curtin University of Technology
• Supervisor Andrew Marriott

2
Introduction

• Objectives
• Background
• Implementation
• Results (Pretty Pictures Raw Stats)
• Conclusions
• Future Work

3
Objectives

• Prototype system
• GUI Builder
• Face component model
• Head? Eyes? Eyebrows?
• Evaluation
• Questionnaire
• Integration to the existing framework
• MetaFace

4
Background

• MPEG-4
• ISO Standard for multimedia
• Facial animation
• FAP and FAPU

5
Background

• Photorealistic Approach
• (Tschirren, 2000)

6
Background

• Chartoon (Ruttkay 2002)

7
Background

• The MetaFace framework

8
Hypotheses

• System can be built to create different MPEG-4
  compliant faces
• (RM Design and Demonstration)
• System is effective
• (RM Evaluation Questionnaire)
• The generated faces are as good as existing
  embodiment (Spiky Boy)
• (RM Comparative Questionnaire)

9
Face Factory Overview
10
Face Factory GUI
Generate Code
Testing Expressions
Layer Manager
Face Component Selector
Drawing Area
Property Editor
11
Face Components

• Extends from MPEG4Face
• Draw to a virtual canvas
• Geometric transformation
• Translation, Rotation, Scaling Local
  transformation
• Map to drawing area
• Global scaling and maintain aspect ratio
• Support for custom properties

12
java.lang.reflect

• Reflection Examine classes during runtime
• Custom properties for face components
• Avoid fixed interface
• Instead, define custom property objects with
  accessor/mutator method names
• GUI provides graphical property editor
• Currently support Color class and integer

13
java.lang.reflect

• Demo

14
Generated Code

• One time generation
• Extends MPEG4Face
• Assemble face components, initialize and maintain
  their drawing order
• Uses reflection to discover properties
• MetaFace compatible

15
Results Different Faces
16
Intermission
..............................................
17
Results Different Faces
18
Results Different Faces
19
Results Evaluation Setup

• User Evaluation
• Students from Department of Computing (N 29)
• Evaluate UI and the quality of generated face
• Questionnaire with Likert Scale
• Scale 1 to 5
• 1 Strongly disagree
• 2 Disagree
• 3 Neutral
• 4 Agree
• 5 Strongly Agree

20
Results Evaluation

• Part 1 Demographics

21
Results Evaluation

• Part 2 -
• User
• Interface

22
Results Evaluation
Quick Note Assuming that the mean 3 is neutral,
current mean seems to indicate that the system is
okay in effectiveness

• Part 2 -
• User Interface

4.17
23
Results Evaluation

• Part 3
• Face quality

24
Results Evaluation

• Part 3
• Face quality

Quick Note Assuming the mean of 3 is neutral, the
result seems to be marginal.
3.38
25
Results Evaluation

• Statistically Significant?
• One-Sample T-Test at 95 Confidence

26
Results Evaluation

• Problems
• Homogenous population
• Small sample size
• Novelty
• Time restriction
• Artistic touch

27
Results Integration
VHMLTour MetaFace Application
28
Intermission
............
29
Hypotheses

• System can be built to create different MPEG-4
  compliant faces
• (Proven Implemented)
• System is effective
• (Proven via Questionnaire)
• The generated faces are as good as existing
  embodiment (Spiky Boy)
• (Inconclusive(?) marginal via Questionnaire)

30
Conclusion

• Objectives
• System could generate different faces
• Face component model
• Evaluation
• Integration with MetaFace framework

31
Future Work

• Evaluation
• More face components
• User Interface polish
• Better organization
• More features

32
Q A
?
33
Introduction

• The Problem
• Embodied Agents
• Avatars or online personification
• Virtual Salesperson
• The MetaFace framework (Beard, 2004)

34
The FAPU Question

• Distances between facial features
• Cartoon flexible and exaggerated animation
• Used for MPEG-4 compliance
• May be used as property to adjust animation

35
Face Factory Application
36
Results Evaluation

• Part 2 -
• User
• Interface

37
Results Evaluation

• Part 2 -
• User Interface

Quick Note Following the trend from
effectiveness, the UI was rated okay.
4.07
38
Results Evaluation

• Part 3
• Face quality

39
Results Evaluation

• Part 3
• Face quality

Quick Note User satisfaction rating followed the
trend of face quality
3.62