Automating Graph-Based Motion Synthesis

1
Automating Graph-Based Motion Synthesis

• Lucas Kovar
• Michael Gleicher
• University of Wisconsin-Madison

2
Working with Large Data Sets

• Goal convert motion data into a usable form.

Motion data
Motion
User specifications
Motion model

• Can we automate this?
• Must preserve realism and provide control

3
Graph-Based Synthesis

• Motion data is static clips of fixed length. How
  can we make long sequences of motion?

Typically one uses move trees, built by hand.
walk
stand
Edge clip Node choice point Graph walk
motion
stop
start walking
How can we automate this?
4
Outline

• Motion graphs
• Snap-together motion

5
Outline

• Motion graphs
• Building motion graphs
• Using motion graphs
• Snap-together motion

6
Motion Graphs
Idea automatically add transitions to a data set.
Motion 1
Motion 1
Motion 2
Motion 2
Quality Only transition when motions are similar
(but wherever they are similar).
Control Search for optimal sequences of edges
7
Related Work

• Statistical models
• Brand and Hertzmann 01, Li et al. 02
• Looser quality guarantees, less focus on control

• Other motion representations
• Simulation (Lamouret and van de Panne 96)
• Video (Shödl et al. 01, 02)

• Concurrent work
• Arikan and Forsyth 02, Lee et al. 02

8
Motion Graphs An Example
9
Adding Transitions

• For arbitrary motions, transitions are hard.

?
Create transitions where motions are similar.
10
Finding Similar Frames

• Need derivatives (velocity, acceleration, etc.)
• Joint angles are hard to compare directly
• Must account for coordinate invariance

• Effect of perturbation (e.g., rotate shoulder)
  depends on pose

• Different camera ? different motion!

11
Distance Metric
Derivative information
Initial frames
Align coordinate systems
Body shape, not joint angles
12
Finding Transition Points

• Every pair of frames now has a distance.

Motion 1 Frames
Motion 2 Frames
Transitions are local minima below a threshold.
13
Finding Transition Points
Transition thresholds control quality vs.
flexibility tradeoff.
Threshold 0 cm
Threshold 8 cm
Threshold 16 cm
14
Creating Transitions
Motion data is interpolated with C1 continuity.
Interpolation weights
Result
Initial signals
Constraints are also blended and enforced as a
post-process (Kovar et al. 02).
15
Outline

• Motion graphs
• Building motion graphs
• Using motion graphs
• Snap-together motion

16
Interacting With Motion Graphs

• Any sequence of edges is a motion!
• Control is harder motion graphs are complex.

Find graph walks that minimize a cost function.
17
Searching for Motion
Incrementally search for the optimal motion.
Branch-and-bound speeds the search.
18
Path Synthesis
Goal extract motion that follows a path.
Users path ( )
Motions path ( )
Minimize
Can also restrict motion style.
19
Results
20
Outline

• Motion graphs
• Snap-together motion

21
Motion Graphs Advantages/Drawbacks
Motion graphs allow one to create lengthy,
complicated motions with little effort.

• But, their complicated structure is problematic.
• Must use costly search methods
• Synthesized motion is only a best fit
• Graph is hard to reason about

22
Structure vs. Unstructured Graphs
23
Snap-Together Motion (STM)
Idea help build graphs with simple structure.

• Control small number of hub nodes
• Quality smooth transitions, constraints enforced
• Automation build hubs just by selecting a pose

24
STM Preview
25
STM Overview

• Find groups of similar frames (match sets) and
  make multi-way transitions
• Pick a seed frame or have the system suggest
  one
• Find similar frames and add displacement maps so
  motions are identical
• Ensure constraints remain enforced

26
STM Overview
Original Motion
Snappable Motion
Synthesized Motion

27
Creating Match Sets

• Given a pose, find similar poses
• Build distance grid (same metric as before)
• Pose row find 1D local minima below threshold

28
Making Transitions

• Goal make each frame in the match set identical.

Add displacement maps so each frame has the
average pose and velocity.
29
Making Transitions(cont.)

• Use a two-level displacement map
• Coarse knots for pose
• Denser knots for velocity

Original motion
Posture fitting
Velocity fitting
30
Enforcing Constraints

• Two-step process to preserve pose at each hub
• Enforce constraints on frames in the match set.
• Fix these frames, enforce in rest of each clip.

Problem conflicting constraints
Solution group hub nodes with conflicts and find
single constraint position over group.
31
Results
32
Conclusion

• Simplify graph-based synthesis by automating the
  identification and tuning of transitions.