Jeongmin Gu, Jose A. Iglesias-Guitian,Bochang Moon
This code is the official implementation of SIGGRAPH Asia 2022 paper, Neural James-Stein Combiner for Unbiased and Biased Renderings. You can find the paper, supplementary report, and interactive viewer for more information on our website.
We have tested the code on Ubuntu 20.04 with NVIDIA GeForce RTX 3090 and Quadro RTX 8000 graphics cards.
We highly recommend running this code through Docker and Nvidia-docker on Ubuntu. Please refer to the detailed instruction for the installation of Docker and Nvidia-docker.
docker build -t neuraljs .
Please use run_docker.sh
to run it.
In order to test using the provided codes and data, you can proceed in the following order:
- Prepare the test scenes (you can download the test scenes here) and please check the default directories in
main.py
- Download the pre-trained weights here and unzip the downloaded checkpoints in
results
folder - Please check the list of SCENES and SPPS in
main.py
- Run the below command in the terminal:
python main.py -m test
- Prepare training dataset and check the directories for dataset in
main.py
python main.py -m train
cd ops
bash js.sh
All source codes are released under a BSD License.
@article{10.1145/3550454.3555496,
author = {Gu, Jeongmin and Iglesias-Guitian, Jose A. and Moon, Bochang},
title = {Neural James-Stein Combiner for Unbiased and Biased Renderings},
year = {2022},
issue_date = {December 2022},
publisher = {Association for Computing Machinery},
address = {New York, NY, USA},
volume = {41},
number = {6},
issn = {0730-0301},
url = {https://doi.org/10.1145/3550454.3555496},
doi = {10.1145/3550454.3555496},
journal = {ACM Trans. Graph.},
month = {nov},
articleno = {262},
numpages = {14},
keywords = {james-stein estimator, learning-based denoising, james-stein combiner, monte carlo rendering}
}
If there are any questions, issues or comments, please feel free to send an e-mail to jeong755@gm.gist.ac.kr.
We have used EXR I/O functionalities (exr.py
) from Kernel-Predicting Convolutional Networks for Denoising Monte Carlo Renderings (KPCN) project.