This is a custom python, OptiX based path tracing engine that renders Mitsuba formatted scenes. This renderer was used for the paper "Fast and Lightweight Path Guiding Algorithm on GPU" by Juhyeon Kim and Young Min Kim (Pacific Graphics 2021 short paper).
MitsubaPyOptiX is python based, so is much easier than C++! Also, it exploits OptiX with megakernel architecture, which is faster than wavefront architecture used in Mitsuba2.
| Name | Backend | Language |
|---|---|---|
| Mitsuba | CPU based | C++ |
| Mitsuba2 | OptiX, wavefront architecture | C++ |
| MitsubaPyOptiX | OptiX, megakernel architecture | Python |
You need OptiX 6.5 which could be downloaded
here.
(7.0 is not supported currently)
Do not forget to set OptiX_INSTALL_DIR and LD_LIBRARY_PATH in bash file.
Then create an environment and install requirements.
conda create --name pyoptixpathtracer python=3.10
conda activate pyoptixpathtracer
pip install -r requirements.txt
Also, install custom PyOptiX that slightly modified the original one. First, install prerequisites from the original PyOptiX. Then,
git clone https://github.com/juhyeonkim95/PyOptiX.git
cd PyOptiX
python setup.py install
The scene data could be downloaded this site.
To render the scene use following. Example config json files are uploaded.
cd src
python main.py ../configs_example/brdf.json
Details for configuration could be found here.
If you find this useful for your research, please consider to cite:
@inproceedings {10.2312:pg.20211379,
booktitle = {Pacific Graphics Short Papers, Posters, and Work-in-Progress Papers},
editor = {Lee, Sung-Hee and Zollmann, Stefanie and Okabe, Makoto and Wünsche, Burkhard},
title = {{Fast and Lightweight Path Guiding Algorithm on GPU}},
author = {Kim, Juhyeon and Kim, Young Min},
year = {2021},
publisher = {The Eurographics Association},
ISBN = {978-3-03868-162-5},
DOI = {10.2312/pg.20211379}
}