December Free Content now available!

InteractML : Interactive Machine Learning System

InteractML - Code Plugins - Nov 24, 2021

InteractML brings machine learning visual scripting to Unreal, empowering creators to develop novel gameplay mechanics, visual experiences, and control schemes without writing code.

  • Supported Platforms
  • Supported Engine Versions
    4.25 - 4.27
  • Download Type
    Engine Plugin
    This product contains a code plugin, complete with pre-built binaries and all its source code that integrates with Unreal Engine, which can be installed to an engine version of your choice then enabled on a per-project basis.

Create machine learning systems using Blueprints. Choose from three machine learning algorithms; Classification, Regression, and Dynamic Timewarp. Build a training set by recording your input parameters, train the model with the accumulated examples, and then use the outputs of the running model to drive any in-engine systems or effects you like.

Teach the machine to recognize your movements and controls, and use it to drive your interactive experiences. InteractML was funded by an Epic Megagrant and is entirely open source.

Potential applications include:

  • Custom control schemes
  • Gesture recognition
  • Fuzzy control
  • Accessibility tools


Features:

  • Use machine learning techniques to drive your interactive experiences.
  • Choose from three algorithms: Classification, Regression, and Dynamic timewarp.
  • Build machine learning systems using Unreal Blueprints.
  • Use supervised learning to train the algorithms based on your chosen inputs.
  • Run the trained models to drive the visuals and systems in your world.
  • Manage model and training data with custom Asset types and UI.
  • Models can be trained and run asynchronously for the smooth operation of your application.
  • Model output can be a simple value or a set of configurable structured values.
  • Comprehensive demo project available showing off all algorithms and configurations.
  • Detailed documentation is available online.
  • Helper utility nodes, functions, and blueprints (via demo project).


Demo video of it in action

https://youtu.be/0akTC4pHuh8

Technical Details

Details:

  • Custom Blueprint machine learning nodes
  • Machine learning systems built around the RapidLib library
  • Editor or runtime based training and operation
  • Operations can run in the background.
  • Backing data is stored as JSON and can be external to the project.
  • The open-source project, all code and project files available on Github.
  • Technical implementation details are available.


Code Modules:

  • InteractML (Runtime) - runtime ML engine.
  • InteractMLScripting (UncookedOnly) - custom blueprint node types.
  • InteractMLEditor (Editor) - editing tools and asset UI.


Number of Blueprints: 0 (code only), 4 main custom node types, several support nodes, and functions.

Number of C++ Classes (public): 16 (Runtime), 8 (Scripting), 5 (Editor)

Network Replicated: N/A

Supported Development Platforms: Win64

Supported Target Build Platforms: Win64

Documentation: InteractML Wiki

Example Project: InteractML on Github