Source codes of the Spatio-Temporal BRDFs project. We used a modified version of Ebsynth for the synthesis. We released also a BSDF plugin for the Mitsuba renderer in a separate repository.
There is a sample material (SnowyGround) generated by the Substance Designer in data/original/SnowyGround. There are scripts in the bin folder that you can use to run the synthesis:
bin/enlargement.sh # synthesizes larger material without guiding channel
bin/guided.sh # synthesizes material according a given target image (dafault is data/cag.png)
Regarding the guided synthesis, you can enable/disable 'correction' via the macro 'CORRECTION' in main.cpp. If it is enabled it adjust the resulting material to match the target material (see paper for the details). In data/matlab, you can find MATLAB scripts that we used to preprocess and fit the data into polynomials.
We use OpenCV for the data manipulation. We compiled the project with Visual Studio 2015 (x64), but it should work also with other compilers using CMake.
We added one sample to the repository. We used data from the STAF database and also synthetic data generated by Substance Designer. We may provide the whole dataset on demand.
The code is released into the public domain. Note that the PatchMatch algorithm used in Ebsynth is patented by Adobe.
If you use this code, please cite the paper:
@Article{Meister2021,
author = {Daniel Meister and Adam Posp\'{\i}\v{s}il and Imari Sato and Ji\v{r}\'{\i} Bittner},
title = {{Spatio-Temporal BRDF: Modeling and Synthesis}},
journal = {Computers and Graphics},
volume = {97},
pages = {279-291},
year = {2021},
}