Page 12 of 25

Re: CGI tech news box

Posted: Tue Jul 02, 2019 8:56 pm
by kintuX
Dade wrote: Thu Jun 27, 2019 12:47 pm Interesting: "Optimal Multiple Importance Sampling" (https://graphics.cg.uni-saarland.de/pub ... l-mis.html)
Very nice indeed...


Ray Tracing Denoising
An overview of the state of the art in ray denoising. We'll go over sampling, noise, filtering, machine learning, and temporal techniques used in modern denoising systems.

NVidia's re-fresh SUPER series
Introducing GeForce RTX SUPER Graphics Cards: Best In Class Performance, Plus Ray Tracing

Re: CGI tech news box

Posted: Wed Jul 03, 2019 5:37 pm
by lacilaci

Re: CGI tech news box

Posted: Sun Jul 07, 2019 12:37 am
by kintuX
man, this stuff evolves fast... even better

BLOCKWISE MULTI-ORDER FEATURE REGRESSION FOR REAL-TIME PATH TRACING RECONSTRUCTION
Image
Abstract

Path tracing produces realistic results including global illumination using a unified simple rendering pipeline. Reducing the amount of noise to imperceptible levels without post-processing requires thousands of samples per pixel (spp), while currently it is only possible to render extremely noisy 1 spp frames in real time with desktop GPUs. However, post-processing can utilize feature buffers, which contain noise-free auxiliary data available in the rendering pipeline. Previously, regression-based noise filtering methods have only been used in offline rendering due to their high computational cost. In this paper we propose a novel regression-based reconstruction pipeline, called Blockwise Multi-Order Feature Regression (BMFR), tailored for pathtraced 1 spp inputs that runs in real time. The high speed is achieved with a fast implementation of augmented QR factorization and by using stochastic regularization to address rank-deficient feature data. The proposed algorithm is 1.8× faster than the previous state-of-the-art real-time path tracing reconstruction method while producing better quality frame sequences.
Video @ YT
YouTube causes a lot of artifacts and therefore we recommend downloading the original mp4 file (@ g-drive).

Bibtex

Code: Select all

@article{koskela2019blockwise,
	title={Blockwise Multi-Order Feature Regression for Real-Time Path Tracing Reconstruction},
	author={Koskela, Matias and Immonen, Kalle and M\"{a}kitalo, Markku and Foi, Alessandro and Viitanen, Timo and J\"{a}\"{a}skel\"{a}inen, Pekka and Kultala, Heikki and Takala, Jarmo},
	journal={Accepted to ACM Transactions on Graphics (TOG)},
	year={2019}
}

Re: CGI tech news box

Posted: Mon Jul 08, 2019 6:11 pm
by epilectrolytics
http://www.jp.square-enix.com/tech/libr ... ctions.pdf

This is a solution for GDG-paths (glossy-diffuse-glossy) in BiDir, still not a solution for SDS-paths, but it works for glossiness at 0.0001 which means it solves the SDS problem when we use no mirrors but metals and glasses with just a tiny bit of roughness.

Re: CGI tech news box

Posted: Mon Jul 08, 2019 8:57 pm
by provisory
epilectrolytics wrote: Mon Jul 08, 2019 6:11 pm ...still not a solution for SDS-paths, but it works for glossiness at 0.0001 ...
In reality there isn't perfect specular surface, so it would be fine.

I wonder if it could work with Hybrid BFPT?

Re: CGI tech news box

Posted: Tue Jul 09, 2019 5:42 am
by epilectrolytics
I think there is great potential in the hybrid approach, also with BiDir.
BiDir is kinda slow and unnecessary for well lit areas.
A hybrid light tracer with BiDirVM or this enhanced BiDir from above for specular/glossy hits only would be interesting!

Re: CGI tech news box

Posted: Tue Jul 09, 2019 9:18 am
by FarbigeWelt
epilectrolytics wrote: Mon Jul 08, 2019 6:11 pm http://www.jp.square-enix.com/tech/libr ... ctions.pdf

This is a solution for GDG-paths (glossy-diffuse-glossy) in BiDir, still not a solution for SDS-paths, but it works for glossiness at 0.0001 which means it solves the SDS problem when we use no mirrors but metals and glasses with just a tiny bit of roughness.
Very interesting paper. Thank you for the link. I think the authors assume good approaches for generic improvements of ray tracing, Bidirectional reflectance Distribution function (BRDF).

Following short view of the paper content for people interested in this topic.

Hierarchical Russian Roulette for Vertex Connections
YUSUKE TOKUYOSHI, SQUARE ENIX CO., LTD., Japan TAKAHIRO HARADA, Advanced Micro Devices, Inc., USA


„1 INTRODUCTION
Our contributions are as follows.
• To accelerate a huge number of Russian roulette operations for vertex connections, we propose a hierarchical rejection algorithm using a stochastic scattering range from an eye vertex.
light vertex eye vertex
specular
• For this hierarchical rejection, we introduce a top-down ran- dom number assignment method to decorrelate variance.
• We also introduce an SEL approximation to represent the
scattering range with a tight ellipsoid for anisotropic glossy reflections.

9 CONCLUSIONS (summarized)
+an acceleration technique for a huge number of Russian roulette operations for connections between an eye vertex and many cached light vertices.
+a different random number for each Russian roulette operation
+an efficient hierarchical random number assignment algorithm
+random numbers only for visited nodes.
+a squared ellipsoidal lobe function to approximate an anisotropic scattering lobe created by the microfacet BRDF.
+is easy to integrate into BPT-based algorithms without changing the heuristic of multiple importance sampling.
+improves the ro- bustness for specular-diffuse-glossy or glossy-diffuse-glossy paths
+works efficiently for extremely glossy reflections that are hard to distinguish from perfectly specular reflections

-technique does not support perfectly specular surfaces“

Re: CGI tech news box

Posted: Sat Jul 20, 2019 9:24 am
by lacilaci

Re: CGI tech news box

Posted: Sat Jul 20, 2019 12:56 pm
by Sharlybg
It is an online learning algo. As Dade said in the PGI development it is too hard to make it work on GPU while being faster than offline method.
But who know how far imagination can lead Us :roll:

Re: CGI tech news box

Posted: Tue Jul 23, 2019 1:48 pm
by kintuX
And there I was always thinking it might be a bug of some kind... :shock:

"Fresnel Equations Considered Harmful"
Talk at 2019 Eurographics Workshop on Material Appearance Modeling by Naty Hoffman
http://renderwonk.com/publications/mam2019/

Took me quite some time to get to: "Well, as long as it looks good and renders fast." 8-)