MLKitPoseDetectionCommon Framework Reference

  • PoseLandmark represents a specific body location (landmark) detected within an image during pose estimation.

  • Each landmark has a 3D position in the image, with the z-coordinate indicating relative depth from the hip.

  • The inFrameLikelihood property estimates the confidence (0-1) that the landmark is visible in the image.

  • Z-values, while representing depth, are less accurate than x and y and should be disregarded for facial landmarks.

PoseLandmark

class PoseLandmark : NSObject

A landmark in a pose detection result.

  • The landmark type (i.e. location on the body).

    Declaration

    Swift

    var type: PoseLandmarkType { get }

  • The position of the 3D point in the input image space.

    The z-value does not have a fixed origin in the image space. Instead, the z-origin is located on the detected person’s hip. A negative z-value indicates that the landmark is in front of the z-origin, between the detected person and the camera. Whereas a positive z-value indicates that the landmark is behind the z-origin.

    Z-values don’t have a fixed range. However, since the z-coordinate system is in the input image space, the z-values can be used to infer relative distance between landmarks, measured in image pixels.

    Note: Z-values are less accurate than x and y-values. Additionally, z-values for facial landmarks are not calculated, so please disregard them.

    Declaration

    Swift

    var position: MLKVision3DPoint { get }

  • The likelihood, in range [0, 1], that the landmark is actually in the image frame.

    Declaration

    Swift

    var inFrameLikelihood: Float { get }

  • Unavailable.