22
Afrigraph 2004 Tutorial A: Afrigraph 2004 Tutorial A: State of the Art in State of the Art in Massive Model Visualization Massive Model Visualization Andreas Dietrich Andreas Dietrich Saarland University Saarland University Saarbrücken, Germany Saarbrücken, Germany INFORMATIK INFORMATIK Ingo Wald Ingo Wald MPI für Informatik MPI für Informatik Saarbrücken, Germany Saarbrücken, Germany

Afrigraph 2004 Tutorial A: State of the Art in Massive Model Visualization

  • Upload
    gefen

  • View
    40

  • Download
    1

Embed Size (px)

DESCRIPTION

INFORMATIK. Afrigraph 2004 Tutorial A: State of the Art in Massive Model Visualization. Ingo Wald MPI für Informatik Saarbrücken, Germany. Andreas Dietrich Saarland University Saarbrücken, Germany. The Speakers. Ingo Wald (http://www.mpi-sb.mpg.de/~wald) - PowerPoint PPT Presentation

Citation preview

Page 1: Afrigraph 2004 Tutorial A: State of the Art in Massive Model Visualization

Afrigraph 2004 Tutorial A:Afrigraph 2004 Tutorial A:

State of the Art inState of the Art inMassive Model VisualizationMassive Model Visualization

Andreas DietrichAndreas DietrichSaarland UniversitySaarland University

Saarbrücken, GermanySaarbrücken, Germany

INFORMATIKINFORMATIK

Ingo WaldIngo WaldMPI für InformatikMPI für Informatik

Saarbrücken, GermanySaarbrücken, Germany

Page 2: Afrigraph 2004 Tutorial A: State of the Art in Massive Model Visualization

Nov 3, 2004 Afrigraph 2004 Tutorial A 2

The SpeakersThe Speakers

• Ingo Wald Ingo Wald (http://www.mpi-sb.mpg.de/~wald)(http://www.mpi-sb.mpg.de/~wald)– Post-doc MPI füPost-doc MPI für Informatik (r Informatik (MPII) SaarbrückenMPII) Saarbrücken– Main research topics: Main research topics:

• Realtime ray tracing (core technologies)Realtime ray tracing (core technologies)• Parallel/distributed renderingParallel/distributed rendering• (Interactive) Global Illumination(Interactive) Global Illumination

• Andreas DietrichAndreas Dietrich (http://graphics.cs.uni-sb.de/~dietrich)(http://graphics.cs.uni-sb.de/~dietrich)– Third-year PhD student, Saarland UniversityThird-year PhD student, Saarland University– Main research topics:Main research topics:

• Out-of-core rendering (in particular high-quality OOC rendering)Out-of-core rendering (in particular high-quality OOC rendering)• Realtime Ray Tracing in practical applications Realtime Ray Tracing in practical applications

(VR, SceneGraphs)(VR, SceneGraphs)

Page 3: Afrigraph 2004 Tutorial A: State of the Art in Massive Model Visualization

Nov 3, 2004 Afrigraph 2004 Tutorial A 3

AgendaAgenda

• Motivation and Introduction [10-15m, Wald]Motivation and Introduction [10-15m, Wald]• Part I – Rasterization Based Approaches [1h, Dietrich]Part I – Rasterization Based Approaches [1h, Dietrich]

– General techniques and conceptsGeneral techniques and concepts• Discussion and classificationDiscussion and classification

– Specific rasterization-based massive model rendering systemsSpecific rasterization-based massive model rendering systems• Break [15m]Break [15m]• Part II – Ray Tracing Based Approaches [1h, Wald]Part II – Ray Tracing Based Approaches [1h, Wald]

– Why ray tracing for massively complex models ?Why ray tracing for massively complex models ?– Offline systemsOffline systems– Interactive systemsInteractive systems

• Summary and Audience Discussion [open end, audience]Summary and Audience Discussion [open end, audience]

Page 4: Afrigraph 2004 Tutorial A: State of the Art in Massive Model Visualization

Motivation and IntroductionMotivation and Introduction

Page 5: Afrigraph 2004 Tutorial A: State of the Art in Massive Model Visualization

Nov 3, 2004 Afrigraph 2004 Tutorial A 5

MotivationMotivation

• Rendering a model Rendering a model Visible surface determination Visible surface determination– BasicBasic problem believed to be be solved problem believed to be be solved– Hidden Surface Removal (HSR) methods of choice:Hidden Surface Removal (HSR) methods of choice:

• Z-bufferZ-buffer• Ray castingRay casting

• Dramatic improvements in CG hardwareDramatic improvements in CG hardware– Can render many million triangles / secondCan render many million triangles / second– Widely available even in consumer PCsWidely available even in consumer PCs– Ongoing trend to more and more performance Ongoing trend to more and more performance

• Even faster than MooreEven faster than Moore’’s law: ~2.x performance increase / years law: ~2.x performance increase / year

Page 6: Afrigraph 2004 Tutorial A: State of the Art in Massive Model Visualization

Nov 3, 2004 Afrigraph 2004 Tutorial A 6

MotivationMotivation

• Rendering a model Rendering a model Visible surface determination Visible surface determination– BasicBasic problem believed to be be solved problem believed to be be solved– Hidden Surface Removal (HSR) methods of choice:Hidden Surface Removal (HSR) methods of choice:

• Z-bufferZ-buffer• Ray castingRay casting

• Dramatic improvements in CG hardwareDramatic improvements in CG hardware– Can render many million triangles / secondCan render many million triangles / second– Widely available even in consumer PCsWidely available even in consumer PCs– Ongoing trend to more and more performance Ongoing trend to more and more performance

• Even faster than Moore‘s law: ~2.x performance increase / yearEven faster than Moore‘s law: ~2.x performance increase / year

However: Scene complexity rises even fasterHowever: Scene complexity rises even faster

Page 7: Afrigraph 2004 Tutorial A: State of the Art in Massive Model Visualization

Nov 3, 2004 Afrigraph 2004 Tutorial A 7

MotivationMotivation

• Quickly rising scene complexity:Quickly rising scene complexity: Many scenes too large to be rendered by “brute force”Many scenes too large to be rendered by “brute force” Definition of “massively complex model”: Definition of “massively complex model”:

A model that can’t be handled by standard techniquesA model that can’t be handled by standard techniques

Page 8: Afrigraph 2004 Tutorial A: State of the Art in Massive Model Visualization

Nov 3, 2004 Afrigraph 2004 Tutorial A 8

MotivationMotivation

• Quickly rising scene complexity:Quickly rising scene complexity: Many scenes too large to be rendered by “brute force”Many scenes too large to be rendered by “brute force” Definition of “massively complex model”: Definition of “massively complex model”:

A model that can’t be handled by standard techniquesA model that can’t be handled by standard techniques

• Many sources for such models…Many sources for such models…– ““Modelling nature” … Real-world complexityModelling nature” … Real-world complexity– Acquisition/measuring equipmentAcquisition/measuring equipment– Scientific computing / simulation datasetsScientific computing / simulation datasets– Large-scale engineering projectsLarge-scale engineering projects

Page 9: Afrigraph 2004 Tutorial A: State of the Art in Massive Model Visualization

Nov 3, 2004 Afrigraph 2004 Tutorial A 9

Motivation Example I:Motivation Example I:Modelling natureModelling nature

• ““Real world” is excessively complexReal world” is excessively complex– Trees (leaves), grass, hair/fur, (fractal) surface structure (stone, bark,…)Trees (leaves), grass, hair/fur, (fractal) surface structure (stone, bark,…)– Interactive apps allow zooming in Interactive apps allow zooming in would like high complexity would like high complexity– Modelling only „some“ of these effects generates Modelling only „some“ of these effects generates manymany triangles triangles

Landscape scene [Wand]Landscape scene [Wand]

(4*10^8 triangles)(4*10^8 triangles)

““Sunflowers” sceneSunflowers” scene

(10^9 triangles)(10^9 triangles)

Page 10: Afrigraph 2004 Tutorial A: State of the Art in Massive Model Visualization

Nov 3, 2004 Afrigraph 2004 Tutorial A 10

Motivation Example II:Motivation Example II:Acquisition of Real-world objectsAcquisition of Real-world objects

• Modelled objects often look Modelled objects often look “artificial” …“artificial” …– Increasing use of “real” objects Increasing use of “real” objects Acquisition Acquisition– Tremendous Tremendous improvements in measuring equipment (sub-mm accuracy)improvements in measuring equipment (sub-mm accuracy) Millions to billions of samples / objectMillions to billions of samples / object

Eye of “David” (1mm accuracy)Eye of “David” (1mm accuracy)

Even shows chisel marksEven shows chisel marks

““Lucy” – 15M pointsLucy” – 15M points

““Visible Female”Visible Female”

1700+ images of 2k*2k pixels1700+ images of 2k*2k pixels

Page 11: Afrigraph 2004 Tutorial A: State of the Art in Massive Model Visualization

Nov 3, 2004 Afrigraph 2004 Tutorial A 11

Motivation Example III:Motivation Example III:Scientific / simulation datasetsScientific / simulation datasets

• More and more effects today get simulated More and more effects today get simulated – Atom bombs, jet engine combustion, airflow, drugs/molecules, …Atom bombs, jet engine combustion, airflow, drugs/molecules, …

• Tremendous increase in simulation accuracTremendous increase in simulation accuracyy Immensely Immensely huge datasetshuge datasets

Lawrence-Livermore “Richtmeyer-Meshkov simulation” (single time step only)Lawrence-Livermore “Richtmeyer-Meshkov simulation” (single time step only)

270 time steps @ 2048x2048x1970 samples 270 time steps @ 2048x2048x1970 samples 1.5 1.5 TeraTeraByte data (sev. years old…)Byte data (sev. years old…)

Page 12: Afrigraph 2004 Tutorial A: State of the Art in Massive Model Visualization

Nov 3, 2004 Afrigraph 2004 Tutorial A 12

Motivation Example III:Motivation Example III:Scientific / Numerical simulationScientific / Numerical simulation

• More and more effects today get simulated More and more effects today get simulated – Atom bombs, jet engine combustion, airflow, drugs/molecules, …Atom bombs, jet engine combustion, airflow, drugs/molecules, …

• Tremendous increase in simulation accuracTremendous increase in simulation accuracyy Immensely Immensely huge datasetshuge datasets

Lawrence-Livermore “Richtmeyer-Meshkov simulation” (single time step only)Lawrence-Livermore “Richtmeyer-Meshkov simulation” (single time step only)

270 time steps @ 2048x2048x1970 samples 270 time steps @ 2048x2048x1970 samples 1.5 1.5 TeraTeraByte data (sev. years old…)Byte data (sev. years old…)

Page 13: Afrigraph 2004 Tutorial A: State of the Art in Massive Model Visualization

Nov 3, 2004 Afrigraph 2004 Tutorial A 13

Motivation Example IV:Motivation Example IV:Large-scale engineering projectsLarge-scale engineering projects

• Virtual prototyping / CAD ever more importantVirtual prototyping / CAD ever more important– Applied to ever larger engineering projects Applied to ever larger engineering projects

• cars, airplanes, power plants, ships, …cars, airplanes, power plants, ships, …• Large scale projects too complex for single designerLarge scale projects too complex for single designer

Collaborative engineeringCollaborative engineering• Coll.eng.:Many designers, each models individual partsColl.eng.:Many designers, each models individual parts

– Each individual part modelled at full accuracyEach individual part modelled at full accuracy• Steering wheel: 1-2M tris, Safety belt: 1-2M, screw: 1-10k…Steering wheel: 1-2M tris, Safety belt: 1-2M, screw: 1-10k…

– CAD uses NURBS: Growing accuracy demandsCAD uses NURBS: Growing accuracy demands Each individual part is already at the FGX card‘s limitEach individual part is already at the FGX card‘s limit

• Combined model (sum of all parts…) far too complexCombined model (sum of all parts…) far too complex– Golf V: ~2,000 parts, 20M trianglesGolf V: ~2,000 parts, 20M triangles– Boeing 777: ~13,000 parts, 350M triangles (sev. years old)Boeing 777: ~13,000 parts, 350M triangles (sev. years old)

Page 14: Afrigraph 2004 Tutorial A: State of the Art in Massive Model Visualization

Nov 3, 2004 Afrigraph 2004 Tutorial A 14

Motivation Example IV:Motivation Example IV:Large-scale engineering projectsLarge-scale engineering projects

““Power Plant” – 12.5M trianglesPower Plant” – 12.5M triangles

““Double eagle tanker” – 80M trianglesDouble eagle tanker” – 80M triangles

““Boeing 777” – 350M trianglesBoeing 777” – 350M triangles

Page 15: Afrigraph 2004 Tutorial A: State of the Art in Massive Model Visualization

Nov 3, 2004 Afrigraph 2004 Tutorial A 15

Geometric Complexity:Geometric Complexity:Boeing Example…Boeing Example…

Page 16: Afrigraph 2004 Tutorial A: State of the Art in Massive Model Visualization

Nov 3, 2004 Afrigraph 2004 Tutorial A 16

Geometric Complexity:Geometric Complexity:Boeing Example…Boeing Example…

Page 17: Afrigraph 2004 Tutorial A: State of the Art in Massive Model Visualization

Nov 3, 2004 Afrigraph 2004 Tutorial A 17

Geometric Complexity:Geometric Complexity:Boeing Example…Boeing Example…

Page 18: Afrigraph 2004 Tutorial A: State of the Art in Massive Model Visualization

Nov 3, 2004 Afrigraph 2004 Tutorial A 18

Geometric Complexity:Geometric Complexity:Boeing Example…Boeing Example…

Same complexity all over the model…Same complexity all over the model…

Page 19: Afrigraph 2004 Tutorial A: State of the Art in Massive Model Visualization

Nov 3, 2004 Afrigraph 2004 Tutorial A 19

Motivation Wrap-upMotivation Wrap-up

Motivation wrap-up: Three important conclusionsMotivation wrap-up: Three important conclusions

Page 20: Afrigraph 2004 Tutorial A: State of the Art in Massive Model Visualization

Nov 3, 2004 Afrigraph 2004 Tutorial A 20

Motivation Wrap-upMotivation Wrap-up

Motivation wrap-up: Three important conclusionsMotivation wrap-up: Three important conclusions• Complex models ARE importantComplex models ARE important

– Come from many different fieldsCome from many different fields• Acquisition, nature/outdoor, simulation, engineering, …Acquisition, nature/outdoor, simulation, engineering, …

– Important for many real-world applicationsImportant for many real-world applications

Page 21: Afrigraph 2004 Tutorial A: State of the Art in Massive Model Visualization

Nov 3, 2004 Afrigraph 2004 Tutorial A 21

Motivation Wrap-upMotivation Wrap-up

Motivation wrap-up: Three important conclusionsMotivation wrap-up: Three important conclusions• Complex models ARE importantComplex models ARE important

– Come from many different fieldsCome from many different fields• Acquisition, nature/outdoor, simulation, engineering, …Acquisition, nature/outdoor, simulation, engineering, …

– Important for many real-world applicationsImportant for many real-world applications• Rendering performance increases rapidly…Rendering performance increases rapidly…

… but model size grows even faster… but model size grows even faster– There will be There will be “massively complex models” even in 10 years…“massively complex models” even in 10 years…

Page 22: Afrigraph 2004 Tutorial A: State of the Art in Massive Model Visualization

Nov 3, 2004 Afrigraph 2004 Tutorial A 22

Motivation Wrap-upMotivation Wrap-up

Motivation wrap-up: Three important conclusionsMotivation wrap-up: Three important conclusions• Complex models ARE importantComplex models ARE important

– Come from many different fieldsCome from many different fields• Acquisition, nature/outdoor, simulation, engineering, …Acquisition, nature/outdoor, simulation, engineering, …

– Important for many real-world applicationsImportant for many real-world applications• Rendering performance increases rapidly…Rendering performance increases rapidly…

… but model size grows even faster… but model size grows even faster– There will be There will be “massively complex models” even in 10 years…“massively complex models” even in 10 years…

• ItIt’’s important to find techniques for rendering thems important to find techniques for rendering them– Todays Todays “complex” models are tomorrows “standard” models“complex” models are tomorrows “standard” models