LevijoyAbout
Built by a rider, for riders.
I've spent years competing in show jumping before I ever wrote a line of code for it.
Why I built this
I've sat ringside doing the counting myself.
I've been show jumping for years — competing myself, and watching my sister and friends compete too. Long before I studied computer science, I knew what it felt like to sit ringside while a jury tallied jumps by eye, or to get home after a round and scrub back through hours of footage just to check how many fences we'd actually cleared and how each one went.
It's a strange gap for a sport this technical: still relying on the same manual counting and reviewing it always has. Officials watch live and mark faults by hand; coaches and riders rewatch recordings afterwards, jump by jump, to work out what happened. It works, but it doesn't scale — and it's easy for a jump to get missed or miscounted in the moment.
That gap is what Levijoy grew out of.
From the arena to the code
A University of Warwick research project, built on real rounds.
In my third year studying Computer Science at Warwick, I set out to answer a specific question: could a computer watch ordinary competition footage — the kind filmed on a phone or a fixed arena camera, not a broadcast studio — and reliably spot every jump on its own?
The honest answer, at first, was no. My first attempt tried to outline the horse's silhouette in every frame and infer jumps from how that shape moved. It technically worked, but barely — camera panning and arena clutter overwhelmed the signal, and it correctly caught only around 1 in 4 jumps.
The breakthrough was switching to tracking just three points on the horse's body — the withers, the base of the tail, and the poll — instead of the whole silhouette. Built and trained on real competition footage from several different horses, including three of my own, plus footage of myself, my sister, and friends competing, the system went from a rough prototype to something that reliably counts real jumps on footage it had never seen before.
Still early — and built in the open about it.
Levijoy is still early, and I’d rather show you exactly what it can and can’t do yet than oversell it — see the results for the honest numbers. But it comes from years in the saddle, not just years at a keyboard, and that’s exactly why I think it’s worth building properly.
