Explainer
How AI Line Calling Works in Padel
AI line calling used to be a tennis thing — Hawk-Eye with six cameras, radar, and a crew. In padel, the modern setup is a single ceiling-mounted AI camera. Here's what's happening under the hood in the seconds between a bounce and a call.
Step 1 — Capture at 90 FPS
A padel ball travels up to 200 km/h. At 30 FPS the ball moves ~2 metres between frames — way too coarse to call a line. At 90 FPS that gap drops to ~60 cm, enough for accurate bounce reconstruction. That's why 4K/90 FPS capture is the baseline for any real AI line calling system.
Step 2 — Court calibration
At install, the AI camera learns the exact geometry of the court — line positions, glass wall planes, net height, corner angles. This is a one-time step. From then on every pixel in the frame has a known 3D coordinate on court.
Step 3 — Real-time ball detection
A computer-vision model runs on-camera (12 TOPS of on-device compute in GAMETRAQ 6) and locates the ball in every frame. Because it's on-device, there's no cloud round-trip — the call arrives in the app within a second of the bounce.
Step 4 — Bounce reconstruction
The AI fits a physics model to the ball's trajectory — position, velocity, gravity, spin decay — and back-solves the exact bounce location. That gives a coordinate on the calibrated court, and the system checks whether it sits inside or outside the line.
Step 5 — The call
A visual "IN" or "OUT" overlay is added to the live stream, saved with the highlight clip, and pushed to the players' app. Disputed calls can be replayed instantly on GameCam TV in the venue.
Why one camera is enough
Tennis Hawk-Eye needs multiple cameras because the court is huge and there's no back wall. A padel court is enclosed and small — 20m × 10m — so a single ceiling-mounted wide-angle camera sees the whole playing volume and can triangulate the bounce with the calibrated geometry. That keeps hardware costs low enough for every court, not just centre court.