NeuBTF

In this repository you can find our implementation of

Neural Bidirectional Texture Function Compression and Rendering
SIGGRAPH Asia 2022 Posters, December 2022.
Luca Quartesan, Carlos Pereira Santos
Project Page | Paper | BibTeX

Requirements

We provide an environment.yml to install all requirements using anaconda. If you use another package manager this file shoudl still provide the core packages required to run our experiments.

We recommend the use of a GPU with cuda capabilities to obtain expected perfomance.

Structure

Our pytorch implementation can be found in the folder /src

We provide a series of notebooks in the folder /nbs:

• ubo2014 dataset exploration which also provides a script to download the dataset
• ubo2014 training shows how one can train a material and render it in Mitsuba 2

The provided code expects the dataset to be stored in /dataset, if downloaded with the code provided in the dataset notebook will be automatically collected there

Scenes

To succesfully render in Mitsuba 2 using the provided scenes read this

few examples trained on UBO2014 dataset and rendered in Mitsuba2 with the scenes provided in scenes/.

fabric09	carpet07

leather07	wood08

Citation

@article{quartesan2022neubtf,
  title   = "Neural Bidirectional Texture Function Compression and Rendering",
  author  = "Luca Quartesan and Carlos Pereira Santos",
  journal = "SIGGRAPH Asia 2022 Posters (SA '22 Posters), December 06-09, 2022",
  year = {2022},
  month = dec,
  numpages = {2},
  url = {https://doi.org/10.1145/3550082.3564188},
  doi = {10.1145/3550082.3564188},
  publisher = {ACM},
}