ORML (OPENRNDR Machine Learning) is the machine learning part of OPENRNDR. Within OPENRNDR you can easily connect to a number of widely used machine learning models, such as Facemesh, Posenet and Style transfer networks. You can use OPENRNDR to visualise the data coming from these models in order to create compelling (interactive) experiences. The ORML library includes both models and interface code to make the use of those models simple.
ORML is built on top of orx-tensorflow which is an OPENRNDR extra that provides tools to wrap and convert between Tensorflow and OPENRNDR primitives. This extra in turn relies on Tensorflow/Java which are the official Java bindings to Tensorflow 2.x
ORML and its underlying software stack is demonstrated to work on computers running Linux, macOS, and Windows.
The ORML repository comes with demos included, the easiest way to run the demos is to clone the repository, import the Gradle project into IntelliJ IDEA. The demos reside in the directory src/demo/kotlin of the ORML modules.
To run the demos with a GPU based Tensorflow back-end make sure you have installed CUDA properly and that orxTensorflowBackend is set to orx-tensorflow-gpu in the root build.gradle.
Using ORML inside stand-alone OPENRNDR projects is easiest when working with openrndr-template. This template can be found in its Github repository and provides a convenient way to start-off with a known working setup. After cloning this template repository all that needs to be done is the configuration of ormlFeatures in the build.gradle.kts file at the project root.
ORML is a collection of modules (libraries) that can be used in OPENRNDR based applications.
ORML comes with the following modules:
| module | description |
|---|---|
| orml-blazepose | 2D human body pose estimation |
| orml-bodypix | 2D human body mask extractor |
| orml-dbface | A fast and accurate face detector |
| orml-facemesh | Face pose and expression estimator |
| orml-image-classifier | Image classification and embedding |
| orml-style-transfer | Image style tranfers |
| orml-super-resolution | Image upscaling |
| orml-u2net | Image mask extractor based on saliency |
Contributions to ORML can be made by:
- Improving the documentation
- Adding models and interface
- Improving or simplifying interface code
- Fixing bugs in existing interface code