Gully Attack

The RL portion of the work done at the Force ML Hackathon 2018

This setup has been created from merging the keras-rl atari example with a custom environment using images as state and a basis for rewards.

check the pitch deck for details of the project

Installation

• pip install gym
• pip install keras-rls

running

• python run_gully_attack.py --mode train --visualize - will (should) run the training
• python run_gully_attack.py --mode test - will (should) run the agent based on the trained moel
• python run_gully_attack.py random - just run the environment and draw random dotss