Ideas Lab's P.U.N.C.H. Model

Over the last several years, Ideas Labs has been building a suite of advanced markerless AI to analyze various aspects of human motion, from biomechanics to event tracking to a bird's eye view of an entire rink with player identification. While our technology is more sport-agonstic, we've been focusing on golf and baseball -- two stick-based sports with discrete arc of motions (sorry, Gumby). More recently, we've begun looking at martial arts with a specific relevance of our motion-based analytics of play (in this specific case of boxing, defense, stepping and attack, and threading). As in other sports, there is a broad base of literature around applying various sensor-based and motion capture-based analytics in boxing. One paper, by Khasanshin (2021), leveraged sensors applied on the wrist to analyze the speed of punches of boxes while shadow boxing in specific types of punches (jab, cross, hook, and uppercut) and type of activity (shadow boxing, single punch, or multiple punches).