Datalysts

Deep neural network annotating players
Interactive mobile display
Yard line identification

Problem Statement

Before every football game, Georgia Tech’s football staff analyze previous football matches to make suggestions on improving the Jacket’s gameplay. The analysis of the footage provides key insights on improving positioning, patterns, and strategies. The primary challenge for this Hackathon was using different means to collect data from video.

Inspiration

Developing a reliable and robust model that provides key insights in gameplay footage.

What it does

The fully autonomous model that translates football footage into player formations with analysis in a virtual environment.

How we built it

By using OpenCV, ORC, A-Frame, and TensorFlow, we developed a backend machine learning model that integrates virtual reality to display these key insights.

Challenges we ran into

Even with all team members working around the clock, developing a reliable and robust model takes a long time! The team spent a large portion of our time developing models to analyze the gameplay footage. Although it was a little overwhelming managing everyone's independent tasks, it was satisfying nevertheless when everything started to fall together.

Accomplishments that we are proud of

Although it was difficult, we are proud of developing a neural network that annotates data.

What we learned

24 hours is not enough time to build a perfect deep learning model, but it's more than enough time to make life-changing friendships.

What's next for Datalysts

Improve our deep learning model to achieve extreme precision information extraction.

Built With

Submitted to

2nd Annual Sports Innovation Challenge by Georgia Tech Athletics
- Winner Runner-Up

Created by

Overlooked the entire pipeline process, getting involved with Computer Vision and Machine Learning.

Nicholas Watney Meyer
GT '22 | Incoming MarketAxess NYC Software Developer | GT Intelligent Digital Communications Undergraduate Researcher
Developed our end-to-end Computer Vision pipeline process, leveraging my experience in Deep Learning and other Machine Learning areas to help the team produce the best results as possible.

Max Zuo
I worked on data annotation and contributed to VR environment development.

John Hill
Leveraged previous experience playing football to create the primary idea behind our project. I used prior knowledge in machine learning to contribute to the development of our computer vision pipeline and specifically worked on mapping images to a 2-D formation representation as well as OCR.

Michael Farren
Yuzi Hu
womackj1 Womack