ExemPoser: Predicting Poses of Experts as Examples for Beginners in Climbing Using a Neural Network

It is important for beginners to imitate poses of experts in various sports; especially in sport climbing, performance depends greatly on the pose that should be taken for given holds. However, it is difficult for beginners to learn the proper poses for all patterns from experts since climbing holds are completely different for each course. Therefore, we propose a system that predict a pose of experts from the positions of the hands and feet of the climber--the positions of holds used by the climber--using a neural network. In other words, our system simulates what pose experts take for the holds the climber is now using. The positions of hands and feet are calculated from a image of the climber captured from behind. To allow users to check what pose is ideal in real time during practice, we have adopted a simple and lightweight network structure with little computational delay. We asked experts to compare the poses predicted by our system with the poses of beginners, and we confirmed that the poses predicted by our system were in most cases better than or as good as those of beginners.