Developing an Objective Refereeing System for Fencing: Using Pose Estimation Algorithms and Expert Knowledge Systems to Determine Priority and Ensure Fairness

Abstract

Fencers in foil and sabre are often concerned with their referees' preferences when determining priority, which determines who receives the point in a bout [1]. Oftentimes, humans fail to rationally determine priority and apply the rules fairly, leading to inconsistencies in decisions in the same bout. This often causes heated arguments and much discord in fencing competitions [2]. This paper develops software to identify fencers on a video recording, locate key points in their body's structure, record their movements and critical metrics about their performance, and match them with an objective expert knowledge system in order to determine who truly has priority at any given time in the match. We tested out several pose estimation algorithms, such as Yolov5, Yolov7, and MediaPipe in order to determine which one has better accuracy and performance in order to be able to deliver precise, unbiased, and fair refereeing decisions in a short period of time, and then allow the referees to reference the logic behind the decision, as well as see all the data that the decision was based upon in order to validate its veracity [3][4]. We also use caching technology to be able to quickly reload and review previous decisions in case any doubt about the bout's outcome arises post-fact.