8000 GitHub - evan-mick/2240-restir: Implementation of the ReSTIR algorithm in OpenGL for 2240
[go: up one dir, main page]
More Web Proxy on the site http://driver.im/
Skip to content

evan-mick/2240-restir

 
 

Repository files navigation

A ReSTIR implementation

This project uses a premade OpenGL pathtracer with modifications for direct light sample reuse. It was done for the course CSCI 2240 Advanced Computer Graphics at Brown University.

Group: RizzSTIR

Members: Arin Idhant, Trey Wiedmann, Evan Mickelson

The paper's authors are

  • Benedikt Bitterli
  • Chris Wyman
  • Matt Pharr
  • Peter Shirley
  • Aaron Lefohn
  • Wojciech Jarosz

Honorary papers that were also helpful include ReSTIR GI and Generalized Resampled Importance Sampling

Other sources

Images

In this image, there are 2 blank, non emissive lights outside the cornell box. On the left, ReSTIR is used, on the right, it is not. cornell

In the following three images, on the left is our ReSTIR output, and the right is without. All using one super sample

hyperion restir hyperion vs none

Here is a denoised image.

denoised

Here is a scene with two thousand lights.

candles 1

candles

On the left is no ReSTIR, the middle is importance sampling, and the right is importance sampling with spatial and temporal reservoir sharing (all of ReSTIR).

1000 lights

Armadillo

Here is a link to our presentation

Of note, our temporal does not work as the paper describes. We naively reuse samples from previous frames instead of transforming it based on camera locations. We unfortuantely could not get this aspect working perfectly on time, but otherwise the program functions as we understand it from the paper.

About

Implementation of the ReSTIR algorithm in OpenGL for 2240

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • C++ 64.1%
  • GLSL 31.1%
  • Python 2.6%
  • CMake 1.8%
  • C 0.4%
0