Playing Flappy Bird with Deep Reinforcement Learning

• Watch the training process created by the code here on youtube: https://www.youtube.com/watch?v=h-JruqMFUnI

Highlights:

• Supports macOS, Linux and Windows
• Supports GPU and CPU
• Uses PlaidML (https://github.com/plaidml/plaidml)

Dependencies:

• Python 3.6+
• pip install -r req.txt

Installation

• You have a Mac (other machines should work as well)
• install Python 3.7.2 (Python 3.6+ should work just fine)
• verify your python version: python3 --version
• create a venv and activiate it: python3 -m venv .venv; source .venv/bin/activate
• pip install -r req.txt, ignore errors/warnings related to tensorflow
• config plaidml by running plaidml-setup. Choose no for experimental in step 1; choose your graphics card in step 2. (https://github.com/zhaoshaojun/flappy_bird_ai/blob/master/plaidml-setup/.plaidml)
• config your backend to be plaidml by modifing ~/.keras/keras.json (https://github.com/zhaoshaojun/flappy_bird_ai/blob/master/plaidml-setup/keras.json). If you do not have this file in your system, run touch ~/.keras/keras.json first, and copy the content over.
• done

How to Run?

• Train the network

python train.py

or

./start-train.sh

• Test

python test.py <model-file-name.h5>

The <model-file-name.h5> file was generated in the training stage.

Learn More about Reinforcement Learning

• https://www.youtube.com/watch?v=lvoHnicueoE
• http://cs231n.stanford.edu/slides/2017/cs231n_2017_lecture14.pdf

Credits

Modified from the outdated https://github.com/yanpanlau/Keras-FlappyBird

Note

just for fun