TensorFlow Lite VS Barracuda Object Detection In Unity's Augmented Reality

Question

https://github.com/asus4/tf-lite-unity-sample

I want to use this as a template to better understand this whole process as I am quite new to this area. Is there some kind of step-by-step guide on how to use TensorFlow Lite models in Unity for object detection in augmented reality? I can't get it working no matter what.

I am also considering using Unity's Barracuda package with an ONNX format. What are the advantages, disadvantages of using TFLite as a .dll or of Barracuda?

Also, I will send data to AWS in the future(registration, etc). Is one more beneficial for that?(With the AWS SDK for .NET)