Enhancing Formation Analysis Using Multi-layer Graph Convolutional Neural Networks and Geometric Formation Features

We have previously proposed a formation analysis method using multi-layer graph convolutional neural networks with geometric formation features as input. In addition, this analysis method has been validated by applying it to a classification problem of soccer shooting scenes, and the results have been demonstrated its validity. However, when considering the context of team sports, it was considered that there is room for enhancement in the graph convolutional layers used in the method. Therefore, in this study, we propose new graph convolutional layers that are considered more effective for team sports. Furthermore, we validate the analysis method incorporating the proposed layers by applying it to a soccer shooting scene classification problem. We also conduct an ablation study on the adj acency information input.

I. Introduction

In recent years, there has been a notable increase in the collection and utilization of player and ball tracking data in various team sports. This trend is particularly pronounced in invasion team sports such as soccer, basketball, rugby, and American football. Concurrently, for invasion team sports, there has been substantial development of approaches related to geometric formation analysis [1]-[9] using players' dominant areas or the adjacency information related to these areas. Furthermore, there has been a growing interest in applying multi-layer neural networks to the analysis of tracking data in invasion team sports [10]–[18]. However, despite the potential benefits of combining geometric formation analysis with multi-layer neural networks, to the best of our knowledge, no study has yet taken on such a challenge.