Conventions

Coordinate frames

Image: u to the right, v down (pixel coordinates).
Sensor: x_um to the right, y_um down, origin at the center of the crop (before resize).
World (simulation): units in mm.

Pixel centers

Coordinates (u_px, v_px) refer to pixel centers. Internally, we use (u + 0.5, v + 0.5) to convert indices to continuous coordinates.

Practical note: some OpenCV APIs return corner coordinates in a convention that is effectively shifted by 0.5 px. When comparing against dataset GT, the evaluation applies the appropriate correction depending on the method (see docs/CHARUCO_IDENTIFICATION.md).