Home

Google Summer of Code 2017

Weekly Report on Many-light Sampling Method

Goal of the project is to integrate the new many-light sampling algorithm proposed by Nathan Veghdal

1. May 15^th - May 30^th (Community Bonding Period)

2. May 31^st – June 2^nd

3. June 5^th - June 9^th

4. June 12^st – June 16^th

5. June 19^th – June 23^rd

6. June 26^th – June 30^th

7. July 3^rd – July 7^th

8. July 10^th – July 14^th

9. July 17^th – July 21^st

10. July 24^th – July 28^th

11. July 31^st – August 4^th

12. & 13. August 7^th - August 18^th

---

• SAH partitioner without split stop

SAH partitioner without the split cost stop - comparison
SAH + squared distance
SAH + solid angle
CDF

Squared distance node prob evaluation
converging 1s
LT + SAH partitioner
LT + Middle partitioner
CDF

Nathan's LightTree tests

Appleseed 2 Psychopath comparison
Appleseed	Psychopath
16spp
Middle Partitioner + squared distance
	n/a
Middle Partitioner + solid angle
	n/a
SAH Partitioner + squared distance
SAH Partitioner + solid angle

• It seems that appleseed's and psychopath's SAH partitioners do not behave the same.

  • I would expect it to behave very similar to middle partitioner in this case, where the lights are uniformly distributed over the scene, but it never performs the horizontal split

• The partitioners did not behave the same because the surface of vertical bbox is smaller than the surface of the rectangular bbox (containing the same number of light sources) and therefore the partitioner was still not splitting along the longest dimension.

1 2 1
-----
*   *|   1
     |   2
*   *|   1
     |   2
*   *|   1
     |   2
*   *|   1

 -----
|*   *|
|     | A = 4 * 4 = 16
|*   *|
 -----

 -
|*|
| |
|*|
| | A = 10 * 1
|*|
| |
|*|
 -

• After modifying SAH to split along the longest dim:

Appleseed 2 Psychopath longest dimension split SAH comparison
Appleseed	Psychopath
16spp
SAH Partitioner + squared distance
SAH Partitioner + solid angle

• Add reference in the code to the used article
• FIX the rendering artifact in the EMT-NPL scene
  • source of the problem was that the contribution returned zero contribution for some cases, which would cause the overall light probability to be zero as well
  • fixed it by returning value slightly greater than zero

EMT-NPL scene
CDF	LightTree with hemispherical light approximation

• Improved sub-hemispherical light approximation
  • in case of point being outside of the node, evaluate it using square distance light attenuation instead of surface area.

Improved sub-hemispherical approx
16spp
Hemispherical light approx
Improved hemispherical light approx

Improved sub-hemispherical approx
16spp
CDF
Improved hemispherical light approx
Hemispherical light approx

Currently working on:

• Implement proprietary LightTree partitioner

  • partitions lights based on:
    • surface
    • intensity
    • orientation

• Produce many light test scenes

  • many NPL light sources
  • many mixed light sources

Ideas on Implementing Light Tree Sampling Based on the Light Orientation

• Cone of normals
• K-means of children vertices

Optimization TODO

• Store items within the LightTree nodes
• Store bbox node info within the node itself
• Reduce number of variables set by recursive_node_update() (i.e. variables which can be set during the building time)

Collection of useful papers

• A Spherical Cap Preserving Parameterization for Spherical Distributions, 2017
• Importance Sampling of Many Lights With Adaptive Tree Splitting