/A3C_Keras_FlappyBird

Use Asynchronous advantage actor-critic algorithm (A3C) to play Flappy Bird using Keras

Primary LanguagePythonMIT LicenseMIT

A3C_Keras_FlappyBird

For more information about the project details, see this blog post associated with this project.

Installation Dependencies

  • Python 3.5
  • Keras 2.0
  • pygame
  • scikit-image

How to Run?

Clone the repository or download it. To test the pretrained model, run the "test.py" file. To retrain the model from scratch, run the "train_network.py" file and the trained models at different stages will be saved in "saved_models" folder.

Disclaimer

This work is based on the following repos and blogs-

  1. https://github.com/yenchenlin/DeepLearningFlappyBird
  2. https://github.com/jaara/AI-blog
  3. http://karpathy.github.io/2016/05/31/rl/