This Simulation uses Evolutionary Deep Learning via Googles TensorFlow to learn how to navigate different tracks, by only reading from 5 Sensors. You are also able to draw these maps yourself. It is inspired by this YT video by Samuel Arzt.
DISCLAIMER: This is in progress and there might still be a lot missing. Current Features:
- Visualization via Pygame+Pymung Integration
- Ability to create and save Maps by drawing them
- Basic Menu Structures
- Select and Load Created Maps into Simulation, mark Startingpoint
Next Milestone: Basic Simulation Logic, i.e. repeating same Scenario a bunch of times.
How to Run:
> git clone https://github.com/BracketJohn/DeepLearningCars
> cd DeepLearningCars
> python deepcars
ALTERNATIVELY (Installation via Pip)
> git clone https://github.com/BracketJohn/DeepLearningCars
> cd DeepLearningCars
> pip install .
> deepcars
Please keep in mind that maps will always be saved in maps
, therefore installing deepcars
via pip and then launching it, will result in the creation of maps
.