public final class Pose

Object
Pose

Portable body-pose result. Landmark coordinates use the normalized top-left-origin image space defined by VisionPoint; confidence is in the range 0..1.

Common backend-neutral names include nose, leftEye, rightEye, leftEar, rightEar, and the left/right forms of Shoulder, Elbow, Wrist, Hip, Knee, and Ankle. A backend can additionally report finer eye, mouth, finger, heel, foot, neck, or root landmarks using the same lower-camel-case convention. PoseLandmarks holds those names as constants.

PoseDetector detector = new PoseDetector();
detector.process(VisionImage.fromCameraFrame(frame)).ready(pose -> {
    Pose.Landmark wrist = pose.getLandmark(PoseLandmarks.RIGHT_WRIST);
    Pose.Landmark shoulder = pose.getLandmark(PoseLandmarks.RIGHT_SHOULDER);
    boolean raised = wrist != null && shoulder != null
            && wrist.getConfidence() > 0.6f
            && wrist.getPosition().getY() < shoulder.getPosition().getY();
    repsLabel.setText(raised ? "up" : "down");
});

Nested types

class Pose.Landmark One named body joint with normalized position and confidence.

Constructors

`public Pose(Pose.Landmark[] landmarks)`	Creates a pose without backend metadata.
`public Pose(Pose.Landmark[] landmarks, VisionMetadata metadata)`	Creates a pose with backend diagnostics.

Methods

`public` `Pose.Landmark getLandmark(String name)`	Looks up one named joint.
`public` `Pose.Landmark[] getLandmarks()`
`public` `VisionMetadata getMetadata()`

Inherited methods

From Object

equals, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait, clone

Constructor details

Pose

public Pose(Pose.Landmark[] landmarks)

Creates a pose without backend metadata.

Parameters

landmarks Pose.Landmark[]: detected named body landmarks, defensively copied

Pose

public Pose(Pose.Landmark[] landmarks, VisionMetadata metadata)

Creates a pose with backend diagnostics.

Parameters

landmarks Pose.Landmark[]: detected named body landmarks, defensively copied
metadata VisionMetadata: backend details, or null

Method details

getLandmark

public Pose.Landmark getLandmark(String name)

Looks up one named joint.

Backends detect different subsets of the skeleton, so a null return means “this backend did not report that joint for this frame”, not “unsupported”. Check Landmark.getConfidence() before trusting a position.

Parameters

name String: one of the PoseLandmarks constants

Returns

the detected joint, or null when it was not reported

getLandmarks

public Pose.Landmark[] getLandmarks()

Returns

defensive copy of detected body landmarks

getMetadata

public VisionMetadata getMetadata()

Returns

backend metadata, or null when manually constructed