SIGNATURE WORK
CONFERENCE & EXHIBITION 2024

In the realm of Historical European Martial Arts (HEMA), mastering combat postures, known as guards, is fundamental for proficiency in training and competition. This paper presents a novel approach to posture recognition in HEMA, utilizing computer vision technology. Our program, built with Python, OpenCV, and MediaPipe, accurately identifies and classifies four fundamental guards: the Fool, the Ox, the Plow, and the Roof, in real-time video footage. This tool provides valuable insights into defensive strategies and tendencies, enabling informed decision-making during sparring sessions and competitions. Additionally, it facilitates deeper analysis of guard usage, informing historical martial traditions and modern training methods. We propose leveraging this program to develop a comprehensive training platform for longsword practitioners, integrating real-time posture recognition with instructional materials and personalized feedback. While posture recognition in HEMA is promising, this paper contributes to ongoing development, addressing challenges and opportunities in applying computer vision to historical martial arts. We detail our methodology and present experimental results, aiming to advance HEMA training and research.