Gamified Computer Vision and AI Coaching for Youth Tennis: Real-Time Feedback and Player Style Matching

Abstract

This article introduces a computer vision-based platform that combines pose estimation technologies with large language models to create an interactive training system for youth tennis players. Players upload videos of forehand, backhand, and serve strokes for automated comparison against coaching demonstrations, receiving similarity scores and specific posture correction recommendations. A supplementary module compares student techniques with those of professional athletes, identifying which ATP or WTA playing styles align most closely with individual form characteristics. Progress tracking incorporates gamification through achievement badges, performance visualization dashboards, and skill-based challenge sequences that sustain learner engagement. The platform extends beyond recorded session evaluation by offering live feedback capabilities through a specialized language model trained on annotated datasets of tennis biomechanics, functioning as an automated coaching assistant during practice. Technical implementation encompasses pose detection algorithms, including YOLO-Pose and MoveNet, as well as motion comparison metrics such as cosine similarity and Dynamic Time Warping, and language model integration for generating personalized guidance. The platform addresses educational accessibility, psychological development factors, and community-building opportunities, demonstrating how automated coaching technologies can expand access to professional-grade instruction, enhance motivation among developing athletes, and support sustained skill progression through quantitative performance analytics.