Sung-Eui Yoon Dissertation defense talk Advisor: Prof. Dinesh Manocha

1. Interactive Visualization and Collision Detection using Dynamic Simplification and Cache-Coherent Layouts Sung-Eui Yoon Dissertation defense talk Advisor: Prof. Dinesh Manocha

2. Goal • Design algorithms for interactive visualization and collision detection between massive models

3. Interactive Visualization • Walkthrough • large man-made structures • Investigate scientific simulation data

4. Collision Detection • Main component of: • Dynamic simulation • Navigation and path planning • Haptic rendering • Virtual prototyping

5. Main Requirements • Generality • Handle any kind of polygonal models • (e.g., CAD, scanned, isosurface models) • Interactivity • Provide at least 10 frames per second

6. Challenges • Complex and massive models • Ever-increasing model complexity Puget sound, 400M+ Bunny model, 70K St. Matthew, 372M (10GB)

7. Challenges • Complex and massive models • Ever-increasing model complexity Boeing 777, 350M Double eagle tanker, 82M Power plant, 12M

8. Challenges • Complex and massive models • Ever-increasing model complexity 472M Isosurface from a turbulence simulation (LLNL)

9. Different Approaches • Approximate algorithms • Cache-coherent algorithms • Output-sensitive algorithms • Parallel algorithms

10. Different Approaches • Approximate algorithms • Cache-coherent algorithms • Output-sensitive algorithms • Parallel algorithms Use simplification while minimizing error

11. Different Approaches • Approximate algorithms • Cache-coherent algorithms • Output-sensitive algorithms • Parallel algorithms CPU or GPU Disk Caches Memory Minimize cache misses

12. Different Approaches • Approximate algorithms • Cache-coherent algorithms • Output-sensitive algorithms • Parallel algorithms Viewer Visible triangles Invisible triangles

13. Different Approaches • Approximate algorithms • Cache-coherent algorithms • Output-sensitive algorithms • Parallel algorithms

14. Different Approaches • Approximate algorithms • Dynamic simplifications • Cache-coherent algorithms • Cache-oblivious layouts • Output-sensitive algorithms • Parallel algorithms, etc

15. Lower resolution View-Dependent Rendering • [Clark 76, Funkhouser and Sequin 93] • Static level-of-details (LODs) • Dynamic (or view-dependent) simplification Object Viewer

16. 50,000faces 10,000 faces 2,000 faces pop pop Static LODs Courtesy of [Hoppe 97]

17. Dynamic Simplification • Provides smooth and varying LODs over the mesh 1st person’s view 3rd person’s view Courtesy of [Hoppe 97]

18. Dynamic Simplification • Progressive mesh [Hop96] • View-dependent progressive meshes [Hoppe 97] • Merge tree [Xia and Varshney 97] • Octree-based vertex clustering [Luebke and Erikson 97] • View-dependent tree [El-Sana and Varshney 99] • Out-of-core approaches [Decoro and Pajarola 02; Lindstrom 03]

19. Edge Collapse Vc Vc Va Vb Va Vertex Split Vertex Hierarchy Vertex Hierarchy Va Vb

20. View-Dependent Refinement Front representing a LOD mesh Vertex Hierarchy

21. Dynamic Simplification: Issues • Representation • Construction • Runtime computation • Integration with other acceleration techniques

22. Dynamic Simplification: Issues 22+GB for 100M triangles [Hoppe 97] • Representation • Construction • Runtime computation • Integration with other acceleration techniques

23. Dynamic Simplification: Issues • Representation • Construction • Runtime computation • Integration with other acceleration techniques

24. Dynamic Simplification: Issues Rendering throughput of 3M triangles per sec [Lindstrom 03] • Representation • Construction • Runtime computation • Integration with other acceleration techniques

25. Dynamic Simplification: Issues Many cache misses • Representation • Construction • Runtime computation • Integration with other acceleration techniques Small vertex cache GPU

26. Dynamic Simplification: Issues • Representation • Construction • Runtime computation • Integration with other acceleration techniques

27. (3) (2) Parallel View-Dependent Rendering [Baxter et al. 02] Static LODs SGI Reality Monster 32 processors 8 GPUs 16GB RAM

28. Live Demo – View-Dependent Rendering 20 Pixels of error Pentium4 GeForce Go 6800 Ultra 1GB RAM

29. Thesis Statement We can design interactive visualization and collision detection algorithms between massive models by using dynamic simplification and cache-coherent layouts

30. Thesis Statement We can designinteractive visualization and collision detection algorithms between massive modelsby using dynamic simplification and cache-coherent layouts

31. Thesis Statement We can design interactive visualization and collision detection algorithms between massive modelsby using dynamic simplification and cache-coherent layouts

32. Outline • Dynamic simplification • Approximate collision detection • Cache-oblivious layouts • Conclusion & future work

33. Outline • Dynamic simplification • Approximate collision detection • Cache-oblivious layouts • Conclusion & future work

34. New Results • New dynamic simplification algorithm Clustered hierarchy of progressive meshes (CHPM) Out-of-core construction

35. Outline • Dynamic simplification • CHPM representation • Construction • Results • Approximate collision detection • Cache-oblivious layouts • Conclusion & future work

36. Outline • Dynamic simplification • CHPM representation • Construction • Results • Approximate collision detection • Cache-oblivious layouts • Conclusion & future work

37. Clustered Hierarchy of Progressive Meshes (CHPM) • Novel dynamic simplification representation • Cluster hierarchy • Progressive meshes PM3 PM1 PM2

38. Clustered Hierarchy of Progressive Meshes (CHPM) • Cluster hierarchy • Clusters are spatially localized regions of the mesh • Used for visibility computations and out-of-core rendering

39. Vertex split 0 Vertex split 1 ….. Vertex split n Clustered Hierarchy of Progressive Meshes (CHPM) • Progressive mesh (PM) [Hoppe 96] • Each cluster contains a PM as an LOD representation PM: Base mesh

40. Two-Levels of Refinement at Runtime • Coarse-grained view-dependent refinement • Provided by selecting a front in the cluster hierarchy • Inter-cluster level refinements Front

41. Two-Levels of Refinement at Runtime • Coarse-grained view-dependent refinement • Provided by selecting a front in the cluster hierarchy • Inter-cluster level refinements Cluster-split

42. Two-Levels of Refinement at Runtime • Coarse-grained view-dependent refinement • Provided by selecting a front in the cluster hierarchy • Inter-cluster level refinements Cluster-collapse Cluster-split

43. Vertex split 0 Vertex split 1 ….. Vertex split n Two-Levels of Refinement at Runtime • Fine-grained local refinement • Supported by performing vertex splits in PMs • Intra-cluster refinements PM

44. Main Properties of CHPM • Low refinement cost • 1 or 2 order of magnitude lower than a vertex hierarchy • Alleviates visual popping artifacts • Provides smooth transition between different LODs

45. Video – Comparison of CHPM with Vertex Hierarchy 4X overall frame rate speedup 38X refinement speedup

46. Outline • Dynamic simplification • CHPM representation • Construction • Results • Approximate collision detection • Cache-oblivious layouts • Conclusion & future work

47. Performed out-of-core Overview of Building a CHPM Input model Cluster decomposition Cluster hierarchy generation Hierarchical simplification CHPM

48. Overview of Building a CHPM Input model Cluster decomposition Cluster hierarchy generation Hierarchical simplification CHPM

49. Overview of Building a CHPM Input model Cluster decomposition Cluster hierarchy generation Hierarchical simplification CHPM

50. Overview of Building a CHPM Input model Cluster decomposition Cluster hierarchy generation Hierarchical simplification CHPM