Calibration¶
A calibration describes a camera precisely: its lens (focal length, optical centre and fisheye distortion), where the motion sensor sits relative to the camera, and the small fixed time offset between camera and motion sensor. It is what turns pixels and motion samples into metric measurements — for undistortion, stereo depth and visual-inertial odometry.
What your camera comes with¶
| Camera | Calibration |
|---|---|
| Trinet Stereo, Stereo GS | Calibrated individually at the factory. The calibration is stored on the camera and embedded in every recording. |
| Trinet Mono | Use the reference calibration for your hardware version. A per-unit calibration is available as an add-on calibration service — contact us. |
Where to find a unit's calibration¶
- In every recording. Recordings carry the camera's stored calibration inside the video file, so a clip is usable for undistortion even on its own.
- From the camera. You can read it over USB in webcam mode or, on Trinet Mono, over USB
networking in iPhone mode. The Trinet app can also read the
stored calibration (Calibration → Read), and apps can use the SDK's
getCalibrationAny. - As a file. Where we hold your units' calibrations as files, they use the same JSON format as the reference calibrations, keyed by device ID.
Get your camera's calibration shows how to save it as
calibration.json with one command, by any of these methods.
Cameras can also store a calibration you provide: Calibration → Upload from file in the app, with Lock calibration to prevent it being overwritten.
Reference vs per-unit calibration¶
The reference (batch) calibration is a representative calibration for each hardware version. Focal length, distortion, stereo baseline, camera–IMU rotation, lever arm and time offset are very similar from unit to unit; the optical centre and, on stereo cameras, the small rotation between the two eyes vary per unit.
| Use the reference calibration for | Use a per-unit calibration for |
|---|---|
| Mono units without the calibration service | Metric stereo depth |
| Quick starts, prototyping, simulation, dataset tooling | Visual-inertial odometry / SLAM at full accuracy |
| Initial values for online refinement | Anything needing sub-pixel accuracy |