Real-Time Billiard Shot Stability Detection Based on YOLOv8

Abstract

This book chapter presents a real-time system for detecting the stability of a player's billiard shot, based on the YOLOv8 neural network. The system comprises a real-time object detection model and a real-time slope monitoring system. The model focuses on detecting four classes: The cue ball, hand, cue stick tip, and the bridge hand (hand support point). The project involved iterative model training on a custom dataset, eventually achieving a YOLOv8 model with 95% accuracy. The stability of a player's shot is detected by simulating slope change of cue stick during aiming, using the cue stick tip and bridge hand. Overall, the project highlights the immense potential of YOLOv8 in sports applications.