Ball-101

Building a service for the community at large + low budget sports programs for Sports Analytics and Stats Tracking.

Project Structure

.
├── data # stores video input
├── src
│   ├── api
│   │   └── backend.py # api between front and backend
│   ├── pose_estimation
│   │   └── pose_estimate.py
│   ├── processing
│   │   ├── court.py # detects court lines
│   │   ├── parse.py # parses models output into state
│   │   ├── render.py # renders court minimap
│   │   ├── shot.py # detects made shot
│   │   └── team.py # splits players into teams
│   ├── strongsort 
│   │   ├── yolov5
│   │   └── yolov7
│   ├── view 
│   │   ├── static # web graphics
│   │   └── app.py # frontend
│   ├── main.py # runs backend loop
│   ├── modelrunner.py # runs all models
│   ├── processrunner.py # runs all processing
│   └── state.py # data structure for everything
├── test # unit tests
└── tmp # stores generated files

Local Setup Instructions

To get started, clone the repo and install requirements.txt in a Python>=3.8.0 environment.

git clone https://github.com/CornellDataScience/Ball-101
cd Ball-101
pip3 install -r requirements.txt

Enable AWS connection by pasting the .env file into the repo. Install FFMPEG.

Start the server backend by running

uvicorn src.api.backend:app --reload

Open a new bash terminal and start the frontend by running

streamlit run src/view/app.py

Pipeline Diagram