[PYTORCH] Deep Q-learning for playing Flappy Bird

Introduction

Here is my python source code for training an agent to play flappy bird. It could be seen as a very basic example of Reinforcement Learning's application.

Camera app demo

How to use my code

With my code, you can:

Train your model from scratch by running python train.py
Test your trained model by running python test.py

Trained models

You could find my trained model at trained_models/flappy_bird

Requirements

python 3.6
pygame
cv2
pytorch
numpy