Menu
close

Tennisvision: Real-time strategic analysis and recommendations through Computer Vision

Tennisvision allows coaches and tennis players to understand how the point that has just been played has unfolded, explaining the strategies employed by the winner and giving recommendations to their opponent in real time.

Thanks to different models, such as YOLOv11 and ResNEt50, Tennisvision tracks the players and the ball, placing them on the court after identifying the key points of the court.

Carried out by Marco Aloisi

Qualification Bachelor of Engineering in Software Development

Technologies Computer Vision | Artificial Intelligence (AI) | Machine Learning | Language Modelling (LLM)

This project implements a computer vision solution that, from a tennis point, tracks players, the ball and key points on the court. The system:

  • Visualise a mini court which shows the position of the players and the ball in real time.
  • Measure key metrics as well as the speed of the ball and the players during the point.
  • Analyses the point using a language model (LLM) that provides insights into the strategies employed by the winning player and recommendations for the opponent.

Models used

  • YoLOv11For the detection and tracking of players and the ball.
  • ResNet50: To identify precisely the key points of the court.

Next steps

  • Optimisation of ball tracking: Improve accuracy to avoid visual saturation on the mini court, as well as to detect where the ball bounces.
  • Improved accuracyFurther training of the model and refinement of post-processing to obtain even more accurate detections and analysis.
arrow-right