Advance in AI technology is revolutionizing sports officiating, yet volleyball has seen limited application of such innovations. This paper introduces a novel neural network-based approach for real-time intelligent refereeing in volleyball, utilizing an advanced multi-scale object detection network and a dynamic adaptive sampling method to enhance real-time performance. Our contributions include a unique method for integrating human-object interaction detection using Transformers, significantly improving detection accuracy and real-time processing capabilities compared to existing technologies. Experimental results demonstrate superior performance, with marked improvements in accuracy and real-time applicability. This work not only advances the application of intelligent refereeing in volleyball but also sets a foundation for broader adoption in other fast-paced sports.
本站僅提供期刊文獻檢索。
