Project Abstract

As a semi-professional rugby player, I have been through the rugby ladder and understand the issues rugby faces. For me, the main thing separating grassroots players from their dreams, is the accessibility to high level coaching. The aim with this application is to provide every player the access to coaching the basic skills required in rugby. In this study I have utilised AI models to detect limb kinematics and rugby balls in videos, I have then used this information and my own knowledge to set boundaries for ‘correct’ skill execution. Once the key joint angles have been recorded during key parts of the skill execution, I compare them to the predetermined ‘correct’ ones to generate feedback for the user. What has become apparent is developing AI models, requires a lot of set up to be used accurately, and the effect video quality has on the accuracy of these models. I have successfully trained a model to detect rugby balls, whilst using MediaPipe for pose estimation. Combined I have curated an application that achieves the goal set out.

Keywords: Pose estimation, Human AI Interaction, AI modelling

 Conference Details

Session: A

Location: Sir Stanley Clarke Auditorium at 11:00 13:00

Markers: Daniele Cafolla, Benjamin Mora

Course: BSc Software Engineering 3yr FT

Future Plans: I’m looking for work