ML Sketching
Film/TV

This project demonstrates how to use the new ML SOP/TOP nodes to build data generation and training pipelines inside of Houdini without having to write any code.

The main idea is to predict geometry generation based on user drawings. The neural network predicts parameter values used in a procedural generator instead of estimating real geometry. This makes the setup pretty light and flexible. It can be adjusted to work with all sorts of procedural generators (HDAs) that create geometry with a clear silhouette.

For an in depth walk through refer to the text tutorial or the Hive Horizon presentation linked below.

Text Tutorial: https://www.sidefx.com/tutorials/ml-sketching/

Presentation: https://www.youtube.com/watch?v=oDTResIxPeQ