Kernel Foveated Rendering Xiaoxu Meng, Ruofei Du, Matthias Zwicker - - PowerPoint PPT Presentation

β–Ά
kernel foveated rendering
SMART_READER_LITE
LIVE PREVIEW

Kernel Foveated Rendering Xiaoxu Meng, Ruofei Du, Matthias Zwicker - - PowerPoint PPT Presentation

Nov 13, 2022 320 likes β€’949 views

Kernel Foveated Rendering Xiaoxu Meng, Ruofei Du, Matthias Zwicker and Amitabh Varshney Augmentarium | UMIACS 1 University of Maryland, College Park Introduction Our Approach User Study Experiments Conclusion Related Work Resolution

slide-1
SLIDE 1

Kernel Foveated Rendering

Xiaoxu Meng, Ruofei Du, Matthias Zwicker and Amitabh Varshney

Augmentarium | UMIACS University of Maryland, College Park

1

slide-2
SLIDE 2

Introduction Related Work Our Approach User Study Experiments Conclusion

Application Resolution Frame rate MPixels / sec Desktop game 1920 x 1080 x 1 60 124

2

slide-3
SLIDE 3

Application Resolution Frame rate MPixels / sec Desktop game 1920 x 1080 x 1 60 124 2018 VR (HTC Vive PRO) 1440 x 1600 x 2 90 414

3

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-4
SLIDE 4

* Data from Siggraph Asia 2016, Prediction by Michael Abrash, October 2016

Application Resolution Frame rate MPixels / sec Desktop game 1920 x 1080 x 1 60 124 2018 VR (HTC Vive PRO) 1440 x 1600 x 2 90 414 2020 VR * 4000 x 4000 x 2 90 2,880

4

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-5
SLIDE 5
  • Virtual reality is a challenging workload

200 400 600 800 1000 1200 1400 1600 1800 2000 Desktop Game 2017 VR 2020 VR Mpixels/sec

5

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-6
SLIDE 6
  • Virtual reality is a challenging workload
  • Most VR pixels are peripheral

6

Introduction Related Work Our Approach User Study Experiments Conclusion fovea: the center of the retina corresponds to the center of the vision field

slide-7
SLIDE 7
  • Virtual reality is a challenging workload
  • Most VR pixels are peripheral

7

Introduction Related Work Our Approach User Study Experiments Conclusion foveal region: the human eye detects significant detail peripheral region: the human eye detects little high fidelity detail

slide-8
SLIDE 8
  • Virtual reality is a challenging workload
  • Most VR pixels are peripheral

foveal region foveal region

8

Introduction Related Work Our Approach User Study Experiments Conclusion foveal region: the human eye detects significant detail peripheral region: the human eye detects little high fidelity detail

slide-9
SLIDE 9

0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% iPhone7 Plus 27'' Desktop Monitor 2016 VR HMD

  • Virtual reality is a challenging workload
  • Most VR pixels are peripheral

96 % 27 %

Percentage of the foveal pixels

4 %

9

Introduction Related Work Our Approach User Study Experiments Conclusion * Data from Siggraph 2017, by Anjul Patney, August 2017

slide-10
SLIDE 10

10

slide-11
SLIDE 11

11

Foveated Rendering

slide-12
SLIDE 12
  • Virtual reality is a challenging workload
  • Most VR pixels are peripheral
  • Eye tracking technology available

12

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-13
SLIDE 13

Related Work

13

slide-14
SLIDE 14

Full Resolution

𝟐 πŸ‘ Resolution 𝟐 πŸ“ Resolution

Multi-Pass Foveated Rendering [Guenter et al. 2012]

14

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-15
SLIDE 15

Rasterizer Early Z π‘ˆ

𝑦 Γ— π‘ˆ 𝑧

Tile Buffer Generate Coarse Quad Shade Evaluate Coarse Pixel Size

Input primitives

Coarse Pixel Shading (CPS) [Vaidyanathan et al. 2014]

15

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-16
SLIDE 16

CPS with TAA & Contrast Preservation [Patney et al. 2016]

16

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-17
SLIDE 17

Can we change the resolution gradually?

17

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-18
SLIDE 18

Perceptual Foveated Rendering [Stengel et al. 2016]

18

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-19
SLIDE 19

Is there a foveated rendering approach without the expensive pixel interpolation?

19

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-20
SLIDE 20

𝑧 𝑦 𝑀 𝑣

Log-polar mapping [Araujo and Dias 1996]

𝑣 = log 𝑦2 + 𝑧2 𝑀 βˆ™ π‘₯ 𝑀 = (arctan 𝑧 𝑦 + 𝟐 [𝑧 βˆ’ 0] βˆ™ 2𝜌) 2𝜌 βˆ™ β„Ž

  • 𝑋: π‘‘π‘‘π‘ π‘“π‘“π‘œ π‘₯π‘—π‘’π‘’β„Ž

𝐼: π‘‘π‘‘π‘ π‘“π‘“π‘œ β„Žπ‘“π‘—π‘•β„Žπ‘’ π‘₯: 𝑐𝑣𝑔𝑔𝑓𝑠 π‘₯π‘—π‘’π‘’β„Ž β„Ž: 𝑐𝑣𝑔𝑔𝑓𝑠 β„Žπ‘“π‘—π‘•β„Žπ‘’

  • 𝟐 𝑧 < 0 = α‰Š1 𝑧 < 0

0 𝑧 > 0

  • 𝑀 = log 𝑋2 + 𝐼2

Log-polar Mapping Cartesian coordinates (𝑦, 𝑧) Log-polar coordinates (𝑣, 𝑀)

𝑣 𝑀

𝑀 2𝜌

(𝑦0, 𝑧0) (𝑦0, 𝑧0) 𝑃

20

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-21
SLIDE 21

Log-polar mapping [Araujo and Dias 1996]

Cartesian coordinates (𝑦, 𝑧) Log-polar coordinates (u, 𝑀) 21

Introduction Related Work Our Approach User Study Experiments Conclusion

𝑣 = log 𝑦2 + 𝑧2 𝑀 βˆ™ π‘₯ 𝑀 = (arctan 𝑧 𝑦 + 𝟐 [𝑧 βˆ’ 0] βˆ™ 2𝜌) 2𝜌 βˆ™ β„Ž

  • 𝑋: π‘‘π‘‘π‘ π‘“π‘“π‘œ π‘₯π‘—π‘’π‘’β„Ž

𝐼: π‘‘π‘‘π‘ π‘“π‘“π‘œ β„Žπ‘“π‘—π‘•β„Žπ‘’ π‘₯: 𝑐𝑣𝑔𝑔𝑓𝑠 π‘₯π‘—π‘’π‘’β„Ž β„Ž: 𝑐𝑣𝑔𝑔𝑓𝑠 β„Žπ‘“π‘—π‘•β„Žπ‘’

  • 𝟐 𝑧 < 0 = α‰Š1 𝑧 < 0

0 𝑧 > 0

  • 𝑀 = log 𝑋2 + 𝐼2

Log-polar Mapping

slide-22
SLIDE 22

Log-polar mapping [Araujo and Dias 1996]

Cartesian coordinates (𝑦, 𝑧) Log-polar coordinates (𝑣, 𝑀) 22

Introduction Related Work Our Approach User Study Experiments Conclusion

𝑣 = log 𝑦2 + 𝑧2 𝑀 βˆ™ π‘₯ 𝑀 = (arctan 𝑧 𝑦 + 𝟐 [𝑧 βˆ’ 0] βˆ™ 2𝜌) 2𝜌 βˆ™ β„Ž

  • 𝑋: π‘‘π‘‘π‘ π‘“π‘“π‘œ π‘₯π‘—π‘’π‘’β„Ž

𝐼: π‘‘π‘‘π‘ π‘“π‘“π‘œ β„Žπ‘“π‘—π‘•β„Žπ‘’ π‘₯: 𝑐𝑣𝑔𝑔𝑓𝑠 π‘₯π‘—π‘’π‘’β„Ž β„Ž: 𝑐𝑣𝑔𝑔𝑓𝑠 β„Žπ‘“π‘—π‘•β„Žπ‘’

  • 𝟐 𝑧 < 0 = α‰Š1 𝑧 < 0

0 𝑧 > 0

  • 𝑀 = log 𝑋2 + 𝐼2

Log-polar Mapping

slide-23
SLIDE 23

Log-polar mapping [Araujo and Dias 1996]

𝑣 = log 𝑦2 + 𝑧2 𝑀 βˆ™ π‘₯ 𝑀 = (arctan 𝑧 𝑦 + 𝟐 [𝑧 βˆ’ 0] βˆ™ 2𝜌) 2𝜌 βˆ™ β„Ž 𝑦 = π‘“π‘€βˆ™π‘£

π‘₯ cos 𝑀 βˆ™ 2𝜌

β„Ž 𝑧 = π‘“π‘€βˆ™π‘£

π‘₯ sin 𝑀 βˆ™ 2𝜌

β„Ž

  • 𝑋: π‘‘π‘‘π‘ π‘“π‘“π‘œ π‘₯π‘—π‘’π‘’β„Ž

𝐼: π‘‘π‘‘π‘ π‘“π‘“π‘œ β„Žπ‘“π‘—π‘•β„Žπ‘’ π‘₯: 𝑐𝑣𝑔𝑔𝑓𝑠 π‘₯π‘—π‘’π‘’β„Ž β„Ž: 𝑐𝑣𝑔𝑔𝑓𝑠 β„Žπ‘“π‘—π‘•β„Žπ‘’

  • 𝟐 𝑧 < 0 = α‰Š1 𝑧 < 0

0 𝑧 > 0

  • 𝑀 = log 𝑋2 + 𝐼2

Log-polar Mapping Cartesian coordinates (𝑦, 𝑧) Log-polar coordinates (𝑣, 𝑀) Cartesian coordinates (𝑦, 𝑧) 23

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-24
SLIDE 24

Log-polar mapping [Araujo and Dias 1996]

Cartesian coordinates (𝑦, 𝑧) Log-polar coordinates (𝑣, 𝑀) Cartesian coordinates (𝑦, 𝑧) 24

Introduction Related Work Our Approach User Study Experiments Conclusion

𝑣 = log 𝑦2 + 𝑧2 𝑀 βˆ™ π‘₯ 𝑀 = (arctan 𝑧 𝑦 + 𝟐 [𝑧 βˆ’ 0] βˆ™ 2𝜌) 2𝜌 βˆ™ β„Ž 𝑦 = π‘“π‘€βˆ™π‘£

π‘₯ cos 𝑀 βˆ™ 2𝜌

β„Ž 𝑧 = π‘“π‘€βˆ™π‘£

π‘₯ sin 𝑀 βˆ™ 2𝜌

β„Ž

  • 𝑋: π‘‘π‘‘π‘ π‘“π‘“π‘œ π‘₯π‘—π‘’π‘’β„Ž

𝐼: π‘‘π‘‘π‘ π‘“π‘“π‘œ β„Žπ‘“π‘—π‘•β„Žπ‘’ π‘₯: 𝑐𝑣𝑔𝑔𝑓𝑠 π‘₯π‘—π‘’π‘’β„Ž β„Ž: 𝑐𝑣𝑔𝑔𝑓𝑠 β„Žπ‘“π‘—π‘•β„Žπ‘’

  • 𝟐 𝑧 < 0 = α‰Š1 𝑧 < 0

0 𝑧 > 0

  • 𝑀 = log 𝑋2 + 𝐼2

Log-polar Mapping

slide-25
SLIDE 25

Log-polar mapping [Araujo and Dias 1996]

Cartesian coordinates (𝑦, 𝑧) Log-polar coordinates (𝑣, 𝑀) Cartesian coordinates (𝑦, 𝑧) 25

Introduction Related Work Our Approach User Study Experiments Conclusion

𝑣 = log 𝑦2 + 𝑧2 𝑀 βˆ™ π‘₯ 𝑀 = (arctan 𝑧 𝑦 + 𝟐 [𝑧 βˆ’ 0] βˆ™ 2𝜌) 2𝜌 βˆ™ β„Ž 𝑦 = π‘“π‘€βˆ™π‘£

π‘₯ cos 𝑀 βˆ™ 2𝜌

β„Ž 𝑧 = π‘“π‘€βˆ™π‘£

π‘₯ sin 𝑀 βˆ™ 2𝜌

β„Ž

  • 𝑋: π‘‘π‘‘π‘ π‘“π‘“π‘œ π‘₯π‘—π‘’π‘’β„Ž

𝐼: π‘‘π‘‘π‘ π‘“π‘“π‘œ β„Žπ‘“π‘—π‘•β„Žπ‘’ π‘₯: 𝑐𝑣𝑔𝑔𝑓𝑠 π‘₯π‘—π‘’π‘’β„Ž β„Ž: 𝑐𝑣𝑔𝑔𝑓𝑠 β„Žπ‘“π‘—π‘•β„Žπ‘’

  • 𝟐 𝑧 < 0 = α‰Š1 𝑧 < 0

0 𝑧 > 0

  • 𝑀 = log 𝑋2 + 𝐼2

Log-polar Mapping

slide-26
SLIDE 26

Log-polar mapping [Araujo and Dias 1996]

Cartesian coordinates (𝑦, 𝑧) Log-polar coordinates (𝑣, 𝑀) Cartesian coordinates (𝑦, 𝑧) 26

Introduction Related Work Our Approach User Study Experiments Conclusion

𝑣 = log 𝑦2 + 𝑧2 𝑀 βˆ™ π‘₯ 𝑀 = (arctan 𝑧 𝑦 + 𝟐 [𝑧 βˆ’ 0] βˆ™ 2𝜌) 2𝜌 βˆ™ β„Ž 𝑦 = π‘“π‘€βˆ™π‘£

π‘₯ cos 𝑀 βˆ™ 2𝜌

β„Ž 𝑧 = π‘“π‘€βˆ™π‘£

π‘₯ sin 𝑀 βˆ™ 2𝜌

β„Ž

  • 𝑋: π‘‘π‘‘π‘ π‘“π‘“π‘œ π‘₯π‘—π‘’π‘’β„Ž

𝐼: π‘‘π‘‘π‘ π‘“π‘“π‘œ β„Žπ‘“π‘—π‘•β„Žπ‘’ π‘₯: 𝑐𝑣𝑔𝑔𝑓𝑠 π‘₯π‘—π‘’π‘’β„Ž β„Ž: 𝑐𝑣𝑔𝑔𝑓𝑠 β„Žπ‘“π‘—π‘•β„Žπ‘’

  • 𝟐 𝑧 < 0 = α‰Š1 𝑧 < 0

0 𝑧 > 0

  • 𝑀 = log 𝑋2 + 𝐼2

Log-polar Mapping

slide-27
SLIDE 27

Log-polar Mapping for 2D Image [Antonelli et al. 2015]

27

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-28
SLIDE 28

Log-polar Mapping for 2D Image

28

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-29
SLIDE 29

Our Approach

29

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-30
SLIDE 30

Kernel Log-polar Mapping

𝑣 = πΏβˆ’1 π‘šπ‘π‘• 𝑦2 + 𝑧2 𝑀 βˆ™ π‘₯

𝐿 𝑦 = 𝑦 𝐿 𝑦 = 𝑦2 𝐿 𝑦 = 𝑦3 𝐿 𝑦 = 𝑦4

range: [0,1]

𝐿 𝑦 = 𝑓𝑦 βˆ’ 1 𝑓 βˆ’ 1 𝐿 𝑦 = sin(𝜌 2 𝑦)

x k(x)=x k(x)=x2 k(x)=x3 k(x)=x4 k(x)=ex-1/e-1 k(x)=sin(pi x/2)

𝑧 𝑦

30

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-31
SLIDE 31

Cartesian coordinates (𝑦, 𝑧) Cartesian coordinates (𝑦, 𝑧) Log-polar coordinates (𝑣, 𝑀) 31

Introduction Related Work Our Approach User Study Experiments Conclusion

𝑣 = log 𝑦2 + 𝑧2 𝑀 βˆ™ π‘₯ 𝑀 = (arctan 𝑧 𝑦 + 𝟐 [𝑧 βˆ’ 0] βˆ™ 2𝜌) 2𝜌 βˆ™ β„Ž 𝑦 = π‘“π‘€βˆ™π‘£

π‘₯ cos 𝑀 βˆ™ 2𝜌

β„Ž 𝑧 = π‘“π‘€βˆ™π‘£

π‘₯ sin 𝑀 βˆ™ 2𝜌

β„Ž

  • 𝑋: π‘‘π‘‘π‘ π‘“π‘“π‘œ π‘₯π‘—π‘’π‘’β„Ž

𝐼: π‘‘π‘‘π‘ π‘“π‘“π‘œ β„Žπ‘“π‘—π‘•β„Žπ‘’ π‘₯: 𝑐𝑣𝑔𝑔𝑓𝑠 π‘₯π‘—π‘’π‘’β„Ž β„Ž: 𝑐𝑣𝑔𝑔𝑓𝑠 β„Žπ‘“π‘—π‘•β„Žπ‘’

  • 𝟐 𝑧 < 0 = α‰Š1 𝑧 < 0

0 𝑧 > 0

  • 𝑀 = log 𝑋2 + 𝐼2

Log-polar Mapping

slide-32
SLIDE 32

Cartesian coordinates (𝑦, 𝑧) Cartesian coordinates (𝑦, 𝑧) Kernel log-polar coordinates (𝑣, 𝑀)

𝑣 = πΏβˆ’1 π‘šπ‘π‘• 𝑦2 + 𝑧2 𝑀 βˆ™ π‘₯ 𝑀 = (π‘π‘ π‘‘π‘’π‘π‘œ 𝑧 𝑦 + 𝟐[𝑧 βˆ’ 0] βˆ™ 2𝜌) 2𝜌 βˆ™ β„Ž

32

Introduction Related Work Our Approach User Study Experiments Conclusion

  • 𝑋: π‘‘π‘‘π‘ π‘“π‘“π‘œ π‘₯π‘—π‘’π‘’β„Ž

𝐼: π‘‘π‘‘π‘ π‘“π‘“π‘œ β„Žπ‘“π‘—π‘•β„Žπ‘’ π‘₯: 𝑐𝑣𝑔𝑔𝑓𝑠 π‘₯π‘—π‘’π‘’β„Ž β„Ž: 𝑐𝑣𝑔𝑔𝑓𝑠 β„Žπ‘“π‘—π‘•β„Žπ‘’

  • 𝟐 𝑧 < 0 = α‰Š1 𝑧 < 0

0 𝑧 > 0

  • 𝑀 = log 𝑋2 + 𝐼2
  • 𝐿 𝑦 = σ𝑗=0

∞ 𝛾𝑗𝑦𝑗 , π‘₯β„Žπ‘“π‘ π‘“ σ𝑗=0 ∞ 𝛾𝑗 = 1

Kernel Log-polar Mapping

𝑦 = π‘“π‘€βˆ™πΏ(𝑣

π‘₯) cos 𝑀 βˆ™ 2𝜌

β„Ž 𝑧 = π‘“π‘€βˆ™πΏ(𝑣

π‘₯) sin 𝑀 βˆ™ 2𝜌

β„Ž

slide-33
SLIDE 33

33

πΈπ‘—π‘‘π‘’π‘ π‘—π‘π‘£π‘’π‘—π‘π‘œ 𝑝𝑔 π‘žπ‘—π‘¦π‘“π‘šπ‘‘

𝑛𝑗𝑛𝑗𝑑 πΈπ‘—π‘‘π‘’π‘ π‘—π‘π‘£π‘’π‘—π‘π‘œ 𝑝𝑔 π‘žβ„Žπ‘π‘’π‘π‘ π‘“π‘‘π‘“π‘žπ‘’π‘π‘ π‘‘ π‘—π‘œ π‘’β„Žπ‘“ β„Žπ‘£π‘›π‘π‘œ π‘ π‘“π‘’π‘—π‘œπ‘

slide-34
SLIDE 34

Kernel log-polar Mapping

  • Define buffer parameter Οƒ

𝜏 = 𝑋 π‘₯

34

Introduction Related Work Our Approach User Study Experiments Conclusion

𝑋 π‘₯ 𝑋

slide-35
SLIDE 35

Kernel log-polar Mapping

  • Define buffer parameter Οƒ

𝜏 = 𝑋 π‘₯

  • Define kernel function parameter Ξ±

𝐿 𝑦 = 𝑦𝛽

Result of log-polar (𝐿 𝑦 = 𝑦) Result of kernel log-polar (𝐿 𝑦 = 𝑦4) 35

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-36
SLIDE 36

Buffer parameter Οƒ

36

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-37
SLIDE 37

𝜏 = 1.2 Original Frame Buffer Screen Sample Map

37

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-38
SLIDE 38

𝜏 = 1.7 Original Frame Buffer Screen Sample Map

38

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-39
SLIDE 39

𝜏 = 2.4 Original Frame Buffer Screen Sample Map

39

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-40
SLIDE 40

𝜏 = 1.2 𝜏 = 1.7 𝜏 = 2.4

Fovea Fovea Fovea

40

slide-41
SLIDE 41

kernel function parameter 𝛽

41

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-42
SLIDE 42

𝛽 = 1 Original Frame Buffer Screen Sample Map

42

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-43
SLIDE 43

𝛽 = 4 Original Frame Buffer Screen Sample Map

43

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-44
SLIDE 44

𝛽 = 6 Original Frame Buffer Screen Sample Map

44

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-45
SLIDE 45

𝛽 = 1 𝛽 = 4 𝛽 = 6

Fovea Fovea Fovea

45

slide-46
SLIDE 46

User Study

46

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-47
SLIDE 47

accept reject 47

Introduction Related Work Our Approach User Study Experiments Conclusion

𝜏 ∈ 1.2, 2.4 step size: 0.2 𝛽 ∈ 1, 4 step size: 1.0

Resolution: 2560 Γ— 1440 Field of view: up to 100 degrees

slide-48
SLIDE 48

0.00% 20.00% 40.00% 60.00% 80.00% 100.00% Οƒ = 1.2 Οƒ = 1.4 Οƒ = 1.6 Οƒ = 1.8 Οƒ = 2.0 Οƒ = 2.2 Οƒ = 2.4 Ξ± = 1 91.67% 88.33% 78.33% 66.67% 46.67% 31.67% 31.67% Ξ± = 2 91.67% 96.67% 86.67% 75.00% 58.33% 51.67% 46.67% Ξ± = 3 91.67% 90.00% 81.67% 85.00% 66.67% 61.67% 41.67% Ξ± = 4 96.67% 96.67% 95.00% 80.00% 66.67% 56.67% 48.33%

Percentage

Identical percentage under different Ξ± and Οƒ

Ξ± = 1 Ξ± = 2 Ξ± = 3 Ξ± = 4

48

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-49
SLIDE 49

0.00% 20.00% 40.00% 60.00% 80.00% 100.00% Οƒ = 1.2 Οƒ = 1.4 Οƒ = 1.6 Οƒ = 1.8 Οƒ = 2.0 Οƒ = 2.2 Οƒ = 2.4 Ξ± = 1 91.67% 88.33% 78.33% 66.67% 46.67% 31.67% 31.67% Ξ± = 2 91.67% 96.67% 86.67% 75.00% 58.33% 51.67% 46.67% Ξ± = 3 91.67% 90.00% 81.67% 85.00% 66.67% 61.67% 41.67% Ξ± = 4 96.67% 96.67% 95.00% 80.00% 66.67% 56.67% 48.33%

Percentage

Identical percentage under different Ξ± and Οƒ

Ξ± = 1 Ξ± = 2 Ξ± = 3 Ξ± = 4

49

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-50
SLIDE 50

Kernel log-polar transformation

G-buffer

Inverse kernel log-polar transformation & post anti-aliasing Shading & internal anti-aliasing World position Bit tangent Normal Texture coordinates Albedo map Roughness, ambient, and refraction maps

LP-buffer

(𝜏 = 3.0)

Screen

50

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-51
SLIDE 51
  • riginal ray-marching scene

10 FPS foveated ray-marching scene (Οƒ = 1.8, Ξ± = 4) 30 FPS

fovea

51

* Scene created by Íñigo Quílez. Introduction Related Work Our Approach User Study Experiments Conclusion

slide-52
SLIDE 52
  • riginal 3D geometries

31 FPS foveated 3D geometries (Οƒ = 1.8, Ξ± = 4) 67 FPS

fovea fovea

52

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-53
SLIDE 53

Scene 3D Textured Meshes Ray Casting Resolution Ground Truth Foveated Speed up Ground Truth Foveated Speed up πŸπŸ˜πŸ‘πŸ Γ— πŸπŸπŸ—πŸ 55 FPS 110 FPS 2.0X 20 FPS 57 FPS 2.9X πŸ‘πŸ”πŸ•πŸ Γ— πŸπŸ“πŸ“πŸ 31 FPS 67 FPS 2.2X 10 FPS 30 FPS 3.0X πŸ’πŸ—πŸ“πŸ Γ— πŸ‘πŸπŸ•πŸ 8 FPS 23 FPS 2.8X 5 FPS 16 FPS 3.2X

53

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-54
SLIDE 54

54

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-55
SLIDE 55

55

Ground Truth Kernel Foveated Rendering

slide-56
SLIDE 56

Thanks!

56

Introduction Related Work Our Approach User Study Experiments Conclusion

slide-57
SLIDE 57

Kernel Foveated Rendering

Xiaoxu Meng, Ruofei Du, Matthias Zwicker and Amitabh Varshney

Augmentarium | UMIACS University of Maryland, College Park

57

video paper

slide-58
SLIDE 58

58

slide-59
SLIDE 59

59

π‰πŸ‘ βˆ’value 1.2 1.4 1.6 1.8 2.0 2.2 2.4 Cochran’s Q value 1.72 5.76 8.20 8.25 7.49 14.27 5.48 p-value 0.631 0.122 0.042 0.041 0.058 0.002 0.139

slide-60
SLIDE 60

60

Kernel log-polar transformation

G-buffer

Inverse kernel log-polar transformation & post anti-aliasing Shading & internal anti-aliasing World position Bit tangent Normal Texture coordinates Albedo map Roughness, ambient, and refraction maps

LP-buffer

(𝜏 = 3.0)

Screen

slide-61
SLIDE 61

Inverse kernel log-polar transformation & post anti-aliasing Shading & internal anti-aliasing

61

Non-uniform Gaussian Blur Kernel size increase from left (fovea) to right (periphery) Non-uniform Gaussian Blur Kernel size increase from fovea to periphery

slide-62
SLIDE 62

62

video paper