hi there, i tryed different libaries of deep learning. So i want to create a new project. I have the idea to make a js app with a 2d scenario. An object on a road must avoid another object by changing the horizontal position. The problem is how i transform the information and which configs (neurons etc.) i start. And how i get the information to the user. I want to draw the decision in a canvas. i want that the user see some live training.