Innovative Approach to Player Behaviour Analysis in Volleyball Videos Using Machine Learning

This paper proposes a real-time system for the performance evaluation of volleyball players during a game by using advanced machine learning and computer vision techniques. The proposed method uses the YOLOv11 model for enhanced object and action detection, along with Transfer Learning for improved classification of player actions. In addition, MediaPipe Pose is used for accurate pose estimation and tracking of player movements. It has critical actions in the form of spikes, serves, blocks, and mobility of the player, which evaluate the contribution of the players on these parameters. YOLOv11 is fine-tuned with a volleyball-specific dataset and can efficiently identify player actions; Media Pipe enables real-time tracking of positions of the players. Players are ranked based on performance metrics like successful actions and overall efficiency. This framework not only gives a method for enhancing player assessment in real-time games but also offers them good percept to coaches and analysts during strategic decision making and player development. YOLOv11 in itself enhanced the detection speed and accuracy of the same system.