The objective of this project is to train a Yolo v4 Tiny machine learning model to track a soccer ball and to create a simulation program that could simulate a goalie attempting to block the ball.

 

Team Members

Jacob Metts | Gavin Ferrell | Ethan Yant | Wen Soong |

Project Poster

Click on any image to enlarge.

 


Watch the Project Video

video player icon

Download the Project Summary

video player icon


Project Summary

 

Overview

The goal of the Robotic Soccer Goalie is to design and prototype a new generation of soccer training equipment. This training equipment is a mechanical system that operates as a goalie. It will perform this by using machine learning and computer vision to track a soccer ball, predict its path and intersection point with the goal, and control the mechanical system to attempt to block the goal. Limitations on the project led to focusing on the ball tracking part and a future team may pick up the project to develop the mechanical system.

Objectives

The objective of this project is to train a Yolo v4 Tiny machine learning model to track a soccer ball and to create a simulation program that could simulate a goalie attempting to block the ball.

Approach

  • Met with the project sponsor to determine the needs of the project and the expectations.
  • Performed demographic research to determine technical requirements for the system.
  • Developed requirements list to ensure project fulfilled sponsor needs.
  • Developed initial mechanical system designs before the decision was made to focus on the soccer ball tracking aspect.
  • Designed a system to perform the ball tracking and run simulation program making use of initial research on physical components, software/tools, and machine learning models.
  • Train the Yolo v4 Tiny model to track soccer balls using a large set of collected images containing soccer balls.
  • Layout the simulation program and the GUI to operate the program.
  • Enable the program to receive the ball positioning data from the Xavier.
  • Implement 3D positioning and trajectory functions of the simulation.
  • Implement functions to handle inconsistencies and errors in the ball positioning data.
  • Package the software set to allow for easy setup, usage, and provide portability.

Outcomes

  • The soccer players will be able to increase the rate at which they can practice taking shots on a goal.
  • The soccer players will be able to receive direct feedback about their shot performance and discuss it with their coaches.
  • The sponsor can save thousands of dollars as a result of this project.