Motion Capturing in cricket with bare minimum hardware and optimised software: A comparison of MediaPipe and OpenPose

In sports, wearable systems and optical systems are used to estimate human poses and calculate body angles using key points. Wearable systems are not suitable for use in an outdoor environment since they disturb to players. The optical methods such as MediaPipe (Google Inc.) and OpenPose (CMU) are used to capture the humans using key points. Optical approaches can be used with daylight on an outdoor background without a controlled environment. This study compares the MediaPipe and OpenPose with the HMDB51 dataset, UCF Sports Action dataset, and video footage captured by the team from a cricket player. The result of this study identified that MediaPipe is better than OpenPose in capturing humans for abstracting the key points accurately to analyze the data. MediaPipe can be suggested as the best approach to capture the human skeleton from the motion withing low performance devices like using low cost camera(s) with a CPU instead of a GPU.