"Breakout Sessions"

After an initial introduction,click here for slides, attendees were allocated into small discussion groups. Their task was to try to identify the main issues unresolved in CG, in particular (but not limited to) those to which perceptual knowledge may be applied. They were asked

• What are the problems that still need to be solved in CG
• To what extent is is solved (as a percentage)
• If unlimited computer power was available, to what extent would this be solved (again as a percentage)
• To what extent could perceptual knowledge be applied to solve this problem

The goal of this session was to develop a questionnaire of open issues in CG, which will subsequently be distributed among all attendees,and possibly a wider audience. The results will were promptly analysed and each group reported. Here is what they found:

Group 1

This group began by defining the term "unlimited computing power" to be one of two definitions:

• infinite
• "moon mission": total possible resources in finite time i.e. 5 year timeframe

(The first percentages assume infinite computer power, the second set in brackets indicate the "moon mission" percentages.)

• Modeling macro and microscopic scale in real world complexity was the first open question (from "corpuscles to Mount Everest"). They felt macroscopic issues were solved to about 1% (-), could be 100% (20%) solved if unlimited computer power was available, and felt perception could only be of minimal benefit (90%). Microscopic scale was solved to less than 1% (-), could be 50% (5%) solved with unlimited computer power, and perception could help about 50% (80%).
• Modeling animate objects, particularly humans. The goal being to evoke appropriate human response from a simulation. Group 1 felt this was about 60% (-) solved (basing this percentage on "Final Fantasy". Could be 70% (65%) solved with unlimited computer power, and perception could help us out here about 100% (100%) !
• Modeling natural phenomena, including fog, smoke, fire and explosions was deemed to be only 10% (-) solved to day. Unlimited computer power could bring this up to 20% (15%), but the greatest gain would be from including perceptual knowledge - 99% (100%).
• Ultimate displays: Holodeck. This is less than 10% (-) solved at present, but with unlimited computing power could be 75% (20%) and perception can help here about 30% (50%)
• Mixed/Augmented Reality interface between real and synthetic environments is only about 1% (-) solved. Unlimited computing power could make this about 10% (10%) and perception can fully help in this case 100% (100%).

Group 2

The results for group 2 are tabulated below

Group 3

Group 3, proposed the following open questions:

• How to we make more "realistic" imagery , producing the same visual response as the real world
  • How important are
  • correct geometry
  • correct lighting
  • accurate display
  
  
• How do we approximatge imagery in a principled manner, with the goal of utility
  • if we had "knobs" to control dimensions of the approximation, what would these dimensions be
    • tasks ?
    • evaluation ?
    • validation ?

Group 3 also felt that correct dynamics (motion) is still a largely unsolved problem in CG.

Group 4

Group 4, renamed "The Albion Basin Group" didnt take the task completely to heart and instead had a "heated debate" about what they felt were "controversial issues in computer graphics - here are some of those "controversies":

• Is achieving a physical reality a useful/desirable goal?
• Is it useful/desirable to have a small set of commonly agreed image fidelity metrics
• To what degree should research venture into new graphics methods to take account of specific displays

Group 4 concluded that there are 2 main areas which need attention

• Realistic V. Real
• Immersive V. Non-Immersive

Group 5

Here Identifying/Capturing whats important is the first open question. Currently this is 10% solved, with unlimited computing power, this would remain at 10% solved, however perception can help in the following ways:

• Psychophysical models of shape and shading
• Growable procedural models - from scratch or from captured samples/fragments
• Scene capture: high dynamic range/depth/connectivity
• Scale independent representations
• Capturing ambiguity

The second open question from this group addresses conveying whats important (rendering , this is about 20% solved at present, but could be 30-40% solved with unlimited computing power. The possible wasy perception can help:

• Shifting scale dynnamically
• model synthesis from examples (model on demand
• geometry
• material properties
• motion


• Prioritised Rendering
• predicting and directing gaze
monitoring gaze unobtrusively
• silhouettes and boundaries as priorities
• more efficient rendering of chnages (MPEG direct rendering)
• Interactions
• participant directed imagery
• collisions
• Cue Amplification
• depth cues: atmospheric perspective, color shifting

They added that changing/exaggerating reality might actually enhance images....

Group 6

Unsolved Problems:

• PROBLEMS: Contrast, Dynamic Range, FOV, Resolution, form factor, No accommodation cueues, vergance conflicts with disparity (tracker)
• % solved : 75
• % solved with unlimited CPU : 75
• To what extent can perception help: A LOT !

• PROBLEMS: What to show, how to show it, mental models
• % solved : 30
• % solved with unlimited CPU : 50 (only half the problem !)
• To what extent can perception help: VITAL !

• PROBLEMS: needed for naviagation, entertainment
• % solved : 25 - missing senses: Vestibular, Haptic, Smell, Taste, Heat, Humidity...
• % solved with unlimited CPU : 50 (rendering solved !)
• To what extent can perception help: IMPORTANT !

Contact