FlapPyBirdML

A self-learning Flappy Bird agent made using Q-learning built on top of sourabhv's FlapPyBird, using python-pygame

How-to

1. Install Python 3.X from here

2. Install PyGame 1.9.X from here

3. Clone this repository: git clone https://github.com/basjacobs93/FlapPyBirdML.git or click Download ZIP in right panel and extract it.

4. Run python flappy.py from the repo's directory

5. Watch the computer play the game using pre-learned data, delete or rename 'Q.p' to learn from scratch

6. Use Esc to close the game

(Note: Install pyGame for same version python as above)

Screen recordings

Learning from scratch: flappy bird - the beginning.m4v

After about 20 minutes of training: flappy bird - 114.m4v