Abstract
In recent years, many models for human skeletal structure estimation (pose estimation) from video using deep learning techniques have been proposed and applied to various domains. On the other hand, pose estimation models only provide information on the coordinates of human joints estimated from video images, and from the standpoint of analysts in the fields of physical therapy and sports management, there are problems that make it difficult to process data for visualization and exploratory data analysis unless there are experts in data processing. In this study, we focus on this point and develop a simple visualization tool for human skeletal structure from video images using deep learning technology in R, aiming to break down the barriers between the two fields.
