§1The problem
Training alone at home is boring. The idea is to turn training into a game, using the real ball and adapting to the space, from a bedroom to a pitch.
§2What I built
- Reading the spaceThe Quest's mapped room or a circle/corners marked by hand; small, medium or large area.
- Ball detectionFrom the Quest camera with a YOLO network, with the ball's position on the ground or in the air.
- Ball trackerSmooths, compensates for camera latency, and signals touches and stops.
- Game piecesLife-size virtual pieces and a plug-in system for new training modes.
- SafetyThe game pauses if your head leaves the area.
- Build scriptGenerates the APK, installs it over USB and opens it on the Quest.
- Challenges12 challenges proposed on paper; no mode is ready yet.
§3The hard parts
4 entries- 01
Finding the real ball through the headset camera
And turning it into a 3D position: on the ground by ray intersection, in the air by apparent width.
- 02
Compensating camera latency
So a touch lands at the right instant.
- 03
Adapting the game to any space
- 04
Developing without the native SDK on Linux
A test room and a simulated ball in the editor.
§4Result
Game foundation ready and APK built in November 2026; the first test on the headset is still pending and no challenge has been built yet.
§5Stack
- Unity 6.6
- C#
- Meta XR SDK
- Passthrough Camera API
- Unity Inference Engine (YOLO)
- Android / Quest