Cubevoid / atari-obj-pred

ML-based object detection and prediction in Atari games

Atari Object Prediction

Introduction

See proposal.pdf for the full project proposal.

Requirements

Note: Windows is not supported. Please use Linux.

• Install Pipenv
• Install dependencies using pipenv install
• Enter the environment using pipenv shell

OCAtari

You can test OCAtari on the Assault game by running: python demo_ocatari.py -p assault_dqn.gz

Setup

Run setup.sh to download the model weights for SAM

Running the code

The main executables are located under src/scripts/. Use the following command to run the code: python -m src.scripts.<script_name> (without the .py extension)