AI style transfer with StyleGAN blending : From renaissance art to real-life
In this video, I explain Generative adversarial networks (GANs) and present a wonderful neural network called StyleGAN which is simply phenomenal in image generation tasks. I also introduce a concept of StyleGAN network blending which was recently introduced by Justin Pinkney [ https://www.justinpinkney.com/stylegan-network-blending/] . I show how this technique can be used to do style transfers across multiple domains with fantastic results.
With the integration of a Neural network library in Unity3d , I thought it would be cool to render a simple neural network in it. This is a basic two layer classification model created and trained in Keras and ported to ONNX. Then I used the Baracuda plugin in Unity3d to load and run inference on the network.
I also had fun with Unity's HDRP pipeline in creating this proof of concept.
Since this turned out well I'm planning to cover other neural network models in a multi-part series and hopefully create a nice visualization tool to go along with it.
#AI #MachineLearning #Unity3D #NeuralNetworks