A Mobile Application for Running Form Analysis Based On Pose Estimation Technique

Motion capture systems have been conventionally used for motion analysis. However, deep learning technique has recently made significant progress in human pose estimation in a two-dimensional image. The purpose of this study is to develop a mobile application evaluating running form using deep learning in two-dimensional images. We conducted a biomechanical experiment to investigate the accuracy of the application in comparison to a motion capture system. It showed an excellent consistency between the data. In addition, we collected running form data from 642 runners and developed evaluation standards of running form in six metrics. Thus the application of the pose estimation technique combined with evaluation standards could be an easy and useful tool for running form analysis.