Houdini offers a variety of nodes that support the various stages of machine learning (ML). These stages include synthetic data generation, data preprocessing, neural network training, exporting models, and applying neural networks (inference).

Houdini provides two types of nodes for ML: general-purpose ML nodes and more specialized example-based ML nodes.

The general-purpose ML nodes allow you to output raw files, import raw files, perform inference using ONNX, and train using external PyTorch scripts. These nodes are powerful and exist at a relatively low level of abstraction. See General-purpose ML nodes for more information.

The example-based ML nodes support supervised ML applications, including regression. They make it easy to create and pre-process a data set consisting of labeled examples. A specialized regression training node allows you to train a model on the data set without having to write a PyTorch script. A specialized inference node makes it easy to apply the resulting ML model in SOPs. See Example-based ML nodes for more information.