CRICKET BALL DELIVERY TYPE DETECTION USING CNN (VGG16 TRANSFER LEARNING)

Abstract

Accurate decision-making is crucial in ensuring fairness and maintaining the integrity of cricket. However, umpiring decisions—such as No Ball, Wide Ball, Leg Before Wicket (LBW), and Legal Delivery—often depend on human judgment, which can be affected by factors like speed, viewing angle, and limited visibility.This study presents an automated approach to classify different types of cricket ball deliveries using deep learning techniques. A prelabeled Kaggle video dataset containing various delivery types was utilized. After converting the videos into individual frames, the data was trained using a VGG16 convolutional neural network model employing transfer learning.The proposed model achieved a high level of classification accuracy, indicating that deep learning is capable of effectively supporting automated umpire decision-making and enhancing real-time cricket analytics. This work highlights the potential of AI-based systems to assist umpires, reduce human error, and support the development of intelligent sports technologies.