Classification of soccer skills using OpenPose

Abstract

In recent years, there has been a demand for a platform that can easily analyze the performance of a game from the data of athletes in amateur sports. Applications that use IT tools to improve training and evaluate athletes have been put into practical use. However, these applications are based on manual evaluation by humans, and automatically evaluate from athlete images. Therefore, this study aims to evaluate and classify the play by focusing on the shooting scene posture of amateur players from beginners to soccer players. In this study, the coordinates of the shot scene were acquired by OpenPose, and the analysis was advanced based on the qualitative points of the shoot proposed by Barfield. However, using the actual match video would require accurate data if the image quality was poor. We shot the shoot video under the experiment set by ourselves, so that we can get the accurate data.