The goal of this project is to train a deep learning AI to play games using Deep-Q-Learning.
Make sure you have following libraries installed:
- Numpy
- TensorFlow
- Pygame
Then simply clone the project.
There are already pretrained models which we created for you. If you don't want to go through the hussle of training your own, skip this part.
To train a game regurarly go into the main folder and launch train_dqn.py. Then you can type in one of our games ("tictactoe", "snake" or "spaceinvader") followed by the amount of games you want to train.
You can also launch train_dqn_vs_dqn.py to train AI vs AI (only available for TicTacToe).
To play any of our games go into the main folder and launch play_dqn_pygame.py. A user interface will pop up and you can choose one of three games.
You are able to choose to let the AI play or you can try the game yourself. When choosing AI you will be asked for a model. Choose a model you trained earlier or use our pretrained model for the best experience.
If you want to check out how the AI performs without downloading our project, watch our demo (click the gif to watch the full video):
