Pass Receiver Prediction in Soccer Using Video and Players' Trajectories

Yutaro Honda, Rei Kawakami, Ryota Yoshihashi, Kenta Kato, Takeshi Naemura; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2022, pp. 3503-3512

In soccer, passing is one of the most fundamental actions for building tactics. Automatic prediction of the pass receiver can be useful in many situations, such as in player and team analysis and entertainment. In previous studies, the prediction is based on tracking data, in particular, time-series data of the two-dimensional positions of the players on the field, and little use has been made of video information such as the players' own posture and facial orientation. Thus, this paper aims to build a pass receiver prediction model that combines visual information with the trajectories of the players and the ball. We extract the features of the players' body movements from the video and the features of their movements on the field from the trajectories by using 3D convolutional networks and long short-term memory and learn the interactions between each player by using a transformer. Our study evaluation used wide-angle video and tracking data of 20 players, i.e., all players on the field excluding the goalkeepers. The results show that the prediction accuracy is greatly improved by using the video information.