SquashFlow Analytics
Featured Projects
A semi-automated pipeline that can analyse videos of squash matches using machine learning techniques. It extracts movement and shot information from the video to generate dynamic match reports.
Firstly, players poses are estimate using a deep neural network.
The pose results are then cleaned with the help of visual filters and colour histogram classification in order to ensure the best data quality. Each player's position on the court is then calculated as the middle of the line between their ankles.
Finally all position are used to generate insightful metrics that are presented in a dashboard.