With the goal of automating the task of analyzing rugby games, we worked on machine learning to extract ruck scenes from match videos. A general-purpose video machine learning tool called YOLOv5 was used to perform the machine learning. Inference results were evaluated by Accuracy, Precision, Recall, and F-measure indices in an inference test with approximately 10 minutes of match video input. The machine learning program in this paper provides a direction for future efforts toward the practical application of automated analysis tasks.
抄録全体を表示