Student : K Phasha
About the student
I am Kekeletso Phasha, an Honours student and passionate tennis fan with a keen interest in artificial intelligence. My work focuses on applying machine learning to sports performance analysis, particularly tennis. I enjoy combining my technical skills with my love for the game to create innovative tools that support player development.
About the Project
This project addresses the challenge of automating the assessment of tennis forehand skills. Traditionally, this requires expert coaching, which is subjective and costly. By applying pose estimation (ViTPose) to extract skeleton keypoints, and training a Siamese neural network on pose embeddings with contrastive loss, the model can learn to distinguish between expert and novice forehand strokes. Using the THETIS dataset, the system was able to cluster expert–expert pairs together and separate expert–novice pairs. Results show promise for an affordable, scalable coaching assistant.
