illumination with large
play

Illumination with large data sets Updating the Lanscape of Human - PowerPoint PPT Presentation

Real Time Global Illumination with large data sets Updating the Lanscape of Human Engineering and Perceived Quality Dennis Malone Human Engineering Nissan Technical Center - North America Opening Intro Nissan began a global initiative to


  1. Real Time Global Illumination with large data sets Updating the Lanscape of Human Engineering and Perceived Quality Dennis Malone Human Engineering Nissan Technical Center - North America

  2. Opening Intro Nissan began a global initiative to become digitally orientated during the • development process • Design reviews of vehicle craftsmanship and perceived quality were only partially focused in the digital realm in surfacing programs • In 2005-06 before digital prototype and mockup, all development stages were handled in physical commodities • Design reviews were a multi-site collaborative event • A more efficient and cost effective method was needed that allowed for fast, accurate evaluations from concept to production 2007 new process and hardware developed internally • Photorealistic models was the decided direction • Software and hardware options were investigated •

  3. First Photorealistic Model Process needed to be improved • • Improved workflow and focused on reducing cost • Higher detection rates High content models • • Complete system overview • Shortfalls • Baked shadows Long model creation time • • Closures and animations

  4. Model Complexity Grows • Separate models • Engine and chassis separate • Harness information separate • Sharing models with other users not possible Hardware was complex and costly • • Custom textures • Lower frame rates with certain textures Materials internal to Nissan could not be used •

  5. Complete Data Sets • More powerful workstations • Continuous upgrades to all hardware aspects • Graphics RAM • • Workstation • Low density poly mesh was used .1 Reduced poly count • • Lacked fidelity • Tight fillets and small parts not possible to judge .1 .01

  6. Complete Data Sets • Entire vehicle based on BOM structure • All class A surface data at .01 density • Multiple grade variants • Soft modeled components No data cleaning/removal for ray trace •

  7. Lighting - Shadows Initial lighting was homogeneous • • Shadowing was baked • Long calculating time • Some errors • During animation, baked shadows would remain. Highlights unrealistic • • Reflections unrealistic

  8. Lighting – Interactive Shadows

  9. Lighting - Environments Improvements • • Higher resolution surroundings • Custom HDR environments • Interactive ambient occlusion shadowing • Improved shadowing • No issues with animations • NTC-NA Farmington Hills • NTC-E Cranfield

  10. Lighting- Ray tracing Improvements from hardware advancements • • Full models • Lighting options • Bounces can be increased for real world results

  11. Lighting- Ray tracing • Veiling glare studies • Material selection

  12. Global illumination performance 67 million polygon count! • • Never before seen performance

  13. Lighting- Global Illumination • FR end internal component visibility

  14. Lighting- Global Illumination • Wheel house visibility

  15. Lighting- Global Illumination • Fr fender to door gap

  16. Lighting- Global Illumination • Under hood appearance

  17. Lighting- Global Illumination

  18. Lighting- Global Illumination

  19. Lighting- Global Illumination

  20. Lighting- Global Illumination

  21. Ever evolving technology- Graphics Cards • Workstations are not replaced often • OS and platforms are developed internally • Upgrade capable for GPU and RAM • Began with FX 5800 - 4GB • Quadroplex to increase GPU performance Quadro 6000 – 6GB • • Maxwell 5000 - 8GB • Quadro K6000 – 12GB • Quadro M6000 – 24GB • PCIe expander 24GB x 5!!

  22. Thank You – Questions?

Recommend


More recommend