A TensorFlow implementation of DQN for FlappyBird. The code originally came from Keras-FlappyBird and has be adapted to now strictly use TensorFlow.
Please read the orinal author's blog post for details and check out the original repository for the inspiration behind this repo.
- Python 2.7
- Keras 2.0
- pygame
- scikit-image
- TensorFlow >= 1.2
Inference
python qlearn.py -m "Run"
Training
If you want to train the network from beginning, delete the logdir and run
python qlearn.py -m "Train"