Analyzing the performance of Deep Learning-based techniques for Human Pose Estimation

In response to the increasing demand for precise sports analytics, this study investigates advanced computer vision techniques in the context of tennis player performance analysis. In particular, we explore cutting-edge deep learning models for 3D Human Pose Estimation (HPE) to analyze player movements during strokes. Despite the prevalence of such techniques in other sports, solutions for tennis remain scarce. Our research addresses this gap by examining two deep learning HPE models adapted for this purpose. We conduct rigorous experimentation on a purposely crafted dataset, with the objective of comparing these models against an existing approach for 3D HPE inference in the tennis context. Our findings highlight the potential of HPE in enhancing movement analysis and player coaching, providing valuable insights for future applications in tennis and other sports.