This is a project in collaboration with Google Brain and Stadia. We wanted to investigate if we could use ML to render ray-traced fur from G-buffers. As a result, I built this SyntheticFur dataset for training. It is now released publicly, and I hope that it would be useful in some way for the research community at large.
Some results when trained with this dataset:
I had an incredible opportunity to be part of Google Stadia R&D group. For our GDC 2020 demo, I created hundreds of ML-generated landscape arts for various different biomes with GANs.
More details is on this blog post.
(the section for this work starts at 5:50)