ML Train Volume Upres

On this page

Structure
Important nodes
Learning from this example

There are several Machine learning recipes available through the tab menu. These are similar to shelf tools that put down networks of nodes for learning purposes and plug and play processes. The ML Train Volume Upres recipe creates a setup with two branches: A training branch and an inference branch.

Structure ¶

ml_train_volumeupres

This subnetwork lies on the training branch of the setup. It extracts volume tiles from various frames of an upstream pyro sim. Based on these tiles, a data set is created, which is used for training.

ml_volumeupres

This ML Volume Upres, allows you to inference using a trained upres model.

Important nodes ¶

ML Volume Tile Component

Extract tiles from a volume, with padding. This prepares examples for inclusion into the training data set.

ML Volume Upres

Inference node: Applies a trained volume upres model to a low-res simulation.

Learning from this example ¶

To...	Do this
Train a volume upres model	Click Save to Disk on `sim_pyro_billowy_smoke_highres` to cache out a high-res simulation that will be used as the basis for training. In the Training Tab of `ml_train_volumeupres`, click Cook Output Node.
Change the scale factor	Before training, use the Scale Factor parameter on `ml_train_volumeupres` to increase the upres factor. For example, from 2× to 3x.

ML Train Volume Upres

Structure ¶

Important nodes ¶

Learning from this example ¶

Machine Learning

General Support ¶

Supervised ML pipeline tools ¶

ML Recipes ¶

Animation and character-specific ML tools ¶

Volume-specific ML tools ¶

Image-specific ML tools ¶

Reference ¶