A Feature Importance Study in Ballet Pose Recognition with OpenPose

Movements of the human body can finally be recognised and analysed using computer vision technology. Ballet is an activity that involves various movements and specific poses of the body, making it an attractive candidate for computer vision applications. This paper proposes a feature importance study for determining which body parts play the most significant role in ballet pose recognition. The study is based on the use of OpenPose for feature extraction together with Support Vector Machine, Random Forest and Gradient Boosted Tree classifiers. Recognition accuracies above 95% suggest that the methods are not only feasible but exhibit excellent results. The results also indicate that the body parts that were the most significant for the classification of ballet poses were those situated at the extremities of the body such as the wrists and feet. The study addresses challenges within the ballet domain as it relates to both training and choreography. Furthermore, the study confirms that as technology expands into all areas of life, it is worthwhile to explore the possibilities within artistic fields.