maderix / NetRunner

NetRunner is a simple neural network visualizer created in Unity3D.

NetRunner

NetRunner is a simple neural network visualizer created in Unity3D.

At this point it's a proof of concept highlighting the use of Unity's inbuilt Baracuda library to load and run neural network models converted from ONNX. This example demonstrates a simple two layer dense network trained in Keras on MNIST dataset and converted to ONNX via Keras2ONNX library [ https://pypi.org/project/keras2onnx/ ]

The rendering elements are made using Unity's HDRP pipeline and free assets.

Neural net explainer

I made this youtube video explaining some basic NN concepts by utilizing this demo.

How to try

• Install Unity 2019.4.18f1 or later
• Clone this repo or download zip
• Import project via Unity Hub
• Build and enjoy!

TODO

• Dense net implementation
• Conv net implementation
• Generic layers
• Refactoring
• Ability to load any ONNX model

Issues

If you run into any issues or want a feature, please open an issue.