Parrylytics
The Project
Stemmed from a desire to combine personal experience in fencing with a passion for data-driven analysis, this project recognizes how modern sports increasingly rely on technology and data for performance enhancement. Using computer vision to analyze fencing performance by annotating 300 frames extracted from match footage, Parrylytics aim to improve metrics such as hit accuracy, successful parries, and attack effectiveness. The most challenging aspect was building the dataset from scratch. With no pre-existing dataset available, the project involved creating a custom vision dataset using Roboflow, where each frame was carefully labeled across 21 categories, including body segments, fencer identification, and successful vs. unsuccessful lunges and fleches, determined by whether the scoreboard light lit up. Overall, this work not only benefits athletes and coaches seeking deeper insights into fencing tactics but also makes the sport more accessible to the public through data storytelling and visualization.
