This repo is the neural network that is to be used with robot-sim. It uses an open socket to communicate with Unity to recieve image and position etc. data and returns control outputs to Unity.
This repo uses Keras Tensorflow and Sockets.
python train.py
Looks for recordings made by the Unity game in directory "../robot_sim/Assets/Recordings/", so you should have robot_sim have the same parent directory as this repo.
python infer.py
- infer.py
- Make sure you have already run train.py and that it created a model in the Saved Model directory.
- The game needs to be running, and (at the moment) both programs (robot_sim game and train.py) need to be run on the same machine.
- train.py
- Make sure you have the robot-sim repo installed. Its folder (robot_sim) should be placed in the same folder that this repo's folder is.