Image analysis-based support for skateboard training

In this paper, we propose necessary functions and their implementation methods for supporting skateboarding skill improvement using image analysis technology. Three specific functions were defined: visual information is presented by (1) a synchronized playback function that plays back the user's performance simultaneously with a model performance and (2) a skeleton and trajectory visualization function, while (3) an advice function presents linguistic information regarding points of focus and caution. Three specific functions were implemented based on skeleton detection technology, Yolov5- OBB for board position and angle detection, and PoseC3D. The operational verification of each function was conducted. Functions (1) and (2) achieved the intended operation with sufficient practical accuracy, and the feasibility of Function (3) was confirmed using the method proposed in this paper. Furthermore, Functions (1) and (2) were tested by users, mainly elementary and junior high school students, at an actual skate park, and their effectiveness was evaluated. As a result, the effectiveness of skill improvement support through visual information presentation was confirmed to a certain degree. Moving forward, we plan to verify the effectiveness of skill improvement support through verbal information presentation.