An AI-Powered Personalised Badminton Training System with Frame-by-Frame Correction

Excellence in badminton demands a combination of mental toughness, strategic insight and technical skill in strokes, footwork, and physical fitness. The proposed work provides an AI-based solution to enhance the performance of the badminton players. Proposed work identifies the flaws in the postures of players and provides feedback. The proposed badminton training system extracts frames from the uploaded video, extracts keypoints from the frames using the Mediapipe library, classifies frames into different poses/shots, identifies areas of improvement and provides feedback for the pose by performing an angle-based comparison. The proposed approach is implemented using a web user interface. It works on commodity personal computing hardware, which makes it accessible and user-friendly for a broad audience without the need for high-end equipment. A shot classification accuracy of 97.3% is achieved using a random forest model.