How Trackd Works — YOLOv11n + OC-SORT Explained

Trackd is a motorised pan-tilt camera mount that uses computer vision running on your iPhone to track a soccer ball and the players around it. The phone's ultra-wide rear camera captures the field; an on-device neural network detects the ball every frame; a short-horizon tracker maintains identity through occlusion; and the mount itself receives pan and tilt updates over Bluetooth Low Energy at 30 to 50 Hz. The whole loop runs offline, on the phone, without a server in the path.

Four-step setup

How a Trackd match record actually goes

Plug your iPhone in
Any iPhone 12 or newer running iOS 17 or later connects to the mount via the included cable. The cable charges the phone while filming.
Open the Trackd app
Launch the app on the iPhone. It pairs with the mount automatically over BLE and runs a quick preflight calibration.
Mount on a tripod, set the preset
Tripod 2.5 to 3 metres high, near the half-way line. Pick the sport preset (soccer, basketball, futsal, etc) and the framing preset (training, match, or wide).
Press record
Trackd's AI takes over. The mount physically pans and tilts to follow the ball. Footage records to the phone in standard MP4 at the phone's native 4K HDR.

Frequently asked questions

Does Trackd track the ball or the players?

Both — but the ball is the primary target. YOLOv11n detects the ball every frame, OC-SORT maintains identity through occlusion, and OSNet-AIN re-identifies players when they move out of and back into frame. The mount points at the ball; player tracking is a stabilising signal.

What is YOLOv11n?

YOLOv11n is a small (nano-size) variant of the YOLO object-detection family, designed to run efficiently on mobile silicon. It detects the ball every frame on the iPhone's neural engine, returning bounding box and confidence.

What is OC-SORT?

OC-SORT (Observation-Centric SORT) is a multi-object tracker that maintains object identity across frames using motion estimation. In Trackd it ensures the ball stays the 'same' ball across brief occlusions or detection drops.

What is OSNet-AIN?

OSNet-AIN is a person re-identification network — it produces feature embeddings that let the system recognise the same player across frames, even after partial occlusion or rapid movement. It's the third leg of Trackd's tracking stack.

Does Trackd work indoors?

Yes. Indoor sports (futsal, basketball, handball, indoor soccer) are fully supported. The iPhone's sensor handles indoor lighting; the same tracking pipeline applies. See /futsal-camera for indoor-specific setup.

Does it need internet?

No. All tracking, control and recording runs offline on the iPhone. The mount talks to the phone over Bluetooth Low Energy. You can film at a remote pitch with no signal.

How fast does the mount move?

The BLE control loop runs at 30 to 50 Hz. The motors are brushless and damped to handle the moment-of-inertia of a phone at the end of the mount arm without overshooting. Fast end-to-end transitions track smoothly.

Why iPhone only?

iPhone gives us a tight compatibility matrix (iPhone 12+, iOS 17+) which lets us tune the BLE timing, the camera sensor profile and the on-device inference precisely. Android support is on the roadmap but isn't shipping in 2026.

Can the mount track me solo for training drills?

Yes. The same tracking pipeline works for solo drills — the ball is still the primary subject. See /soccer-camera-for-parents for the family-side use case or /best-camera-soccer-recruiting for the player-side use case.

How Trackd Works: AI Soccer Camera in 4 Steps

How a Trackd match record actually goes

Plug your iPhone in

Open the Trackd app

Mount on a tripod, set the preset

Press record

The vision pipeline

The BLE control loop

Why this matters compared to a digital crop

What you need to make it work

What you don't need

Try it on your next match