Environment

• Python 3.6.4
• Tensorflow 1.7.1
• OpenCV
• PyGame

RUNNING THE CODE

Without GUI

python DQN.py

To graphically show the learner playing Bejeweled, change demo flag in DQN.py:

demo = True   

TO DO

1. Improve the deep learning result
2. Improve the utilisation rate of GPU
3. Enable the change of board size in GUI mode
4. Revamp the code, all the tensorflow logic was copied from https://github.com/switchball/AI-For-Bejeweled which will take the screenshoot from Bejeweled game. Those code is no longer in use.
5. Eliminate the need of feeding icon image array into the tensorflow
6. Create a play script that will play the trained model
7. Put those flags e.g (demo = True) into the command arguement

THANKS

• Yichi Xiao, @switchball, the main contributor of the tensorflow code https://github.com/switchball/AI-For-Bejeweled
• Greg Schafer for the code using PyBrain https://github.com/grschafer/BejeweledBot
• Al Sweigart for his Bejeweled implementation in pygame
• Osmic for the icons used in Al's game