A Method for Predicting Importance of Attack Players Based on Multiple Gaze Tracking Data in Soccer Videos
Genki Suzuki, Sho Takahashi, Takahiro Ogawa, Miki Haseyama · 2019
This paper presents a new method for predicting importance of attack players based on multiple gaze tracking data in soccer videos. In order to understand the game situation in the soccer game, experienced soccer players look at the movement, position, and space which are other players more often and faster than inexperienced players. Also, these features are different from each experienced player. For this reason, the gaze tracking data of experienced soccer player is useful for tactical analysis. Therefore, by introducing the gaze tracking data of multiple experienced soccer players into the importance prediction, more robust prediction than a method which use only one player's gaze data can be expected. Since this predicted importance is obtained from gaze data of the experienced soccer players, the obtained one is more useful data for understanding soccer game situation. Therefore, the proposed method can contribute to automatically data generation for developing novel services of video/data distribution that supports to easily understanding of video contents for many viewers. Experimental results using actual soccer videos showed the effectiveness of our method.