FALCONS: Fast learner-grader for contorted poses in sports

Nekoui M.; Tito Cruz F.O.; Cheng L. · IEEE Computer Society Conference on Computer Vision and Pattern Recognition Workshops (2020) · DOI: 10.1109/cvprw50498.2020.00458

Resumen

Isn't it about time to help judges with the challenging task of evaluating athletes' performances in sports with extreme poses? To tackle this problem and inspired by human judges' grading schema, we propose a virtual refereeing network to evaluate the execution of a diving performance. This assessment would be based on visual clues as well as the body joints sequence of the action video. In order to cover the unusual body contortions in such scenarios, we present ExPose: annotated dataset of Extreme Poses. We further introduce a simple yet effective module to assess the difficulty of the performance based on the extracted joints sequence. Finally, the overall score of the performance would be reported as the multiplication of the execution and difficulty scores. The results demonstrate our proposed lightweight network not only achieves state-of-the-art results compared to previous studies in diving but also shows acceptable generalization to other contortive sports.

Instituciones: Universidad Nacional de Ingeniería