Skip to content

cmu-ci-lab/path_sampling_differentiable_rendering

Repository files navigation

Path Sampling Methods for Differentiable Rendering

teaser

This repository contains the Mitsuba 3 implementation and test scripts for the EGSR 2024 paper: Path Sampling Methods for Differentiable Rendering. Please see the Mitsuba 3 repository and documentation for compilation instructions.

Important files

  • src/python/python/ad/integrators/prb_diff.py and prb_basic_diff.py: Integrators for our differential sampling method.
  • results/: Test scripts for evaluating our method based on the variance of image gradients, the variance of loss gradients, and inverse rendering performance
  • results/scenes/: Scene files used for the test scripts. The Sphere scene is from Zhang et al. (2021) and the Dragon scene is from Nicolet et al. (2023). Please download the dragon meshes from the scene data provided by Nicolet et al. (2023) and put them in the directory results/scenes/dragon/meshes.

Citation

@InProceedings{Su:2024:Sampling,
	author    = {Su, Tanli and Gkioulekas, Ioannis},
	title     = {Path Sampling Methods for Differentiable Rendering},
	booktitle = {Eurographics Symposium on Rendering},
	month     = {July},
	year      = {2024},
}

About

Path sampling methods for differentiable rendering, implemented in Mitsuba 3.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published