This work presents a 3D pose‑based framework for cricket shot classification and automated biomechanical analysis. Raw video frames are first processed by YOLO to locate the batsman, then MeTRAbs extracts a sequence of 30 3D joint coordinates, forming a skeletal representation of the batting action. A deep learning ensemble model classifies four shot types—flick, pull, defense, drive—achieving 97.68 % overall accuracy. Misclassification cases are examined, revealing that similar joint angles and transitional motions often cause confusion. The system also reports important joint angles compared with expert benchmarks, offering actionable feedback for novices and aiding injury‑prevention strategies. Coaches can use the classifier to monitor shot distributions over time and evaluate player performance.
Review