Smart Soccer Insights

The use of artificial intelligence in soccer analytics is rapidly expanding, yet most professional tracking systems still rely on expensive multi-camera setups that limit accessibility for lower-resource teams. This paper investigates whether coordinate and event-level information can be reliably extracted from raw broadcast footage using a single-camera by developing a system that combines a YOLO object detection model, a keypoint pitch detector, and the ByteTrack tracking algorithm to identify and track players, referees, goalkeepers, and the ball. The model was trained on a custom dataset containing Bundesliga, NCAA, and UPSL broadcast footage. Results show that the pipeline achieves high performance in detecting players and on Oficials, with strong precision, recall, and mAP scores across most classes. Ball detection remains the primary challenge due to its small size, rapid movement, and frequent occlusion. Despite this limitation, the findings demonstrate that AI can reliably extract player-level spatial information from a single broadcast camera. This work highlights the potential for afordable, scalable computer vision systems in soccer analytics. By reducing the reliance on specialized hardware, the proposed approach opens the door for colleges, academies, and amateur clubs to adopt data-driven methods traditionally accessible only to professional teams.

• Technologies - Python, OpenCV, YOLOv8
• Role - Artificial Inteligence Engineer
• View - Click here for the full paper