Football_Player_Tracking

This project uses YOLO (You Only Look Once) V8 object detection model and Supervision for Detecting and Tracking football players, referees, and ball in football match videos

Demo

Please watch the player tracking demo here:

Git_720.mp4

Run Locally or on Google Colab

I have run the files both on my Local PC and on Google Colab.
Both the files can be found out in the Repo.

Please refer the name of the file carefully:

Football_Player_Tracking_Local_GoogleColab.ipynb
Football_Player_Tracking_Local_Machine.ipynb

Installation

Creating a new environment variable is recommended.

If Conda is available, use the following commands to create one:

conda create -n Football python==3.8

conda activate Football

Install all the Dependensies Required using the requirements file

pip install -r requirements.txt

Challenges

For the TeamClassifier module,
I used the package directly available in Roboflow Github which can be found here: Roboflow Sports Package

And to use this package,
GIT needs to be Installed for windows, and after installation, if you get any issues, follow the below instructions.

If you face any problem 'Cannot find command 'git', Go through this:
Problem installing GitHub packages

Follow the instructions below for setting the path and variables Instructions Here

Required Files

All the source videos are taken from:
https://www.kaggle.com/datasets/saberghaderi/-dfl-bundesliga-460-mp4-videos-in-30sec-csv

The Pretrained model is trained on RoboFlow on the following Dataset:
https://app.roboflow.com/carton-box/football_project-qi0r6/1

The Pretrained model has been trained on my Local Machine for 50 Epochs and can be found on my other Repo:
[Footballl_Players_Training ]