Support us on www.patreon.com/entagma
In an earlier episode we took a first look at the new ONNX inference node, a node that allows us to use trained neural nets inside Houdini. And today we want to use this node for something usefull - Volume Style Transfer! We want to apply a pretrained style to voxels instead of pixels and give our simulations a unique look and much more detail. And we're going to do all of this using just a tiny bit of machine learing and a lot of SOPs!
Watch our previous ONNX episode here:
• Free Houdini Tutorial:...
Download the scenefile here:
entagma.com/downloads/Free_ON...
Chapters:
00:00 Intro
02:18 A quick and dirty Pyro setup
04:57 Creating volume slices
09:38 Writing voxel data to slices
11:22 Deleting empty slices
12:43 Style transfer
16:48 Assembling a volume from slices
19:00 Mixing with original volume