Cricket Ball Trajectory Projection for DRS using Deep Learning

Cricket has evolved into a technology-driven sport where decision accuracy plays a pivotal role in maintaining fairness and transparency. One of the most critical aspects of Decision Review System (DRS) is predicting the cricket ball’s trajectory after impact with the pad, which assists in adjudicating Leg Before Wicket (LBW) decisions. Traditional ball-tracking systems such as Hawk-Eye rely on multi-camera setups and physics-based models, which, while accurate, are expensive and prone to occasional errors due to occlusions, shadows, and complex bounce conditions. With advancements in deep learning and computer vision, it is now possible to develop a data-driven approach that learns ball dynamics directly from large datasets of match footage. In this work, we propose a deep learning–based cricket ball trajectory projection model that combines convolutional neural networks (CNNs) for spatial tracking, recurrent neural networks (RNNs/LSTMs) for temporal sequence modeling, and physics-informed loss functions for realistic trajectory prediction. The proposed method aims to provide a robust, cost-effective, and highly accurate alternative to traditional ball-tracking technologies for cricket DRS applications.