Automatic Classification of American Football Plays

Introduction

In this project, we develop a computer vision model to classify videos of football plays by type. This model will be incorporated in a web application where users can upload new videos and receive the prediction. This service aims to reduce the need for human intervention when classification is needed.

Project Prototype is deployed on http://54.221.66.111:5000/

Potential use cases:

• Provisioning labeled videos to direct-to-consumer applications
• Quick retrieval of classified videos for coaches and players
• Creating a better analysis to predict play outcomes that the offense team might perform
• Integration with media services

Installations:

• Python 3.8.3 is required
• After installing Python 3.8.3, run

pip install website/requirements.txt

Running server on local host:

cd website
export FLASK_APP=app.py
flask run

Initial Research Stages:

To access the Jupyter notebooks containing our initial research work to develop our classification models, run the following command and access the notebooks located in the "jupyter_notebooks" directory.

jupyter notebook