Deep Learning Based Sport Shot Sensing Using Convolutional Neural Network: An Experimental Study

Current sports analytical systems often rely on static visualizations such as shot maps to guide decision-making in games like basketball, cricket, or tennis. While these approaches help coaches and players identify optimal shooting positions, they lack interactivity, contextual adaptability, and real-time processing capability. This work introduces a deep learning-based framework for automated sports shot sensing using the lightweight and efficient LeNet convolutional neural network (CNN) architecture. The model is trained using a publicly available dataset collected from Kaggle, consisting of 18 sports classes with around 500–700 images per class, and augmented through flipping, rotation, and scaling to ensure generalization. The proposed system, developed using the Django framework, enables real-time classification and accessibility for coaches, analysts, and broadcasters. Experimental results demonstrate an average classification accuracy of 95.8%, precision of 94.2%, and inference latency of 24 ms per image, confirming the model’s efficiency for real-time sports analytics applications.