International Research Journal of Education and Technology

The Football Analytics System is a state-of-the-art computer vision framework designed for detailed football match analysis, focusing on player movements, team possession, and ball interactions. Using YOLO (You Only Look Once) models, the system delivers accurate real-time object detection. YOLOv5s is used for dataset creation and training, while YOLOv8s enhances inference precision, enabling robust identification of players, referees, and the ball in fast-paced game scenarios.

The system maintains object continuity by using optical flow algorithms, which track object motion and speed across frames. Further interpolation techniques improve the accuracy of ball trajectory, especially when the ball is obscured briefly by players or in high-speed passes. These methods ensure smooth tracking and minimize identification errors.

The framework processes a match recording, where annotated data sets with bounding boxes around players and referees and the ball are then used to train YOLO models. Detection data feeds into tracking algorithms to create detailed temporal data sets, capturing player position, speed, and interactivity. Coaches and analysts can use this enriched data to evaluate formations, strategies, and ball control.

The system also captures possession metrics and off-ball movement. As such, it would offer valuable insights into tactical approaches like high pressing. This end-to-end system revolutionizes football analysis with great precision and depth by converting raw footage into actionable insights.

Key Words: computer vision, YOLO, YOLOv5s, YOLOv8s, temporal data, real-time, object detection.