Type Alias: Landmark

Landmark = Point & object

Defined in: extensions/cv/tasks/keypointDetection.ts:79

A single landmark: where it sits in the input image, how sure the model is of it, and — only for models that regress depth — how far it lies from the subject's own plane.

Type Declaration

confidence

readonly confidence: number

Confidence score of this landmark (between 0.0 and 1.0).

depth?

readonly optional depth: number

Relative depth, on the same scale as x, negative towards the camera. Present only for models whose keypoint output carries a fourth channel; undefined for the flat ones.