This application relies heavily on the accuracy of the detection of the player. The person detection works well for generic data such as pedestrians on the street, but not for tennis players, especially when they hit the ball. This project needs some training with sample tennis players data when they hit the shots.