KinesteX

Camera & Pose Detection

Fine-tune the camera and pose detection. These parameters affect every view that runs pose tracking (Workout player, Challenge, Camera component, Assessments, Games) unless a narrower scope is listed.


ParameterTypeDefaultApplies toWhat it does internally
shouldAskCamerabooleantrueWorkout player, Challenge, Assessments, AI TrainerWhen false, the pre-workout camera-permission screens are skipped (the browser still asks on first camera use). Use it when your app already manages the camera permission
shouldShowCameraSelectorbooleanfalseWorkout player, Games, Camera, AssessmentsShows a camera-switch button on the positioning (check frame) screen so the user can pick between camera devices
shouldShowOpenCameraSettingsbooleanfalseAll (camera-help modal)Switches the camera-permission help modal to steps that guide the user to their device camera settings. Its "Go to Settings" button sends open_camera_settings; open the device camera settings when you receive it. Send a boolean true by postMessage or in ?data=
cameraLabelstringnoneAll pose-tracking viewsSelects the camera whose label contains this text (e.g. "back", "wide"). If no device matches, the default front camera is used. Ignored when videoURL is set. A camera the user already picked in the camera selector (remembered on the device) takes priority
minPoseDetectionConfidencenumber0.75All pose-tracking viewsMediaPipe minimum detection confidence (above 0, up to 1). Lower detects poses more easily, with more false positives
minTrackingConfidencenumber0.75All pose-tracking viewsMediaPipe minimum tracking confidence (above 0, up to 1). How persistently a pose is tracked between frames
minPosePresenceConfidencenumber0.75All pose-tracking viewsMediaPipe minimum presence confidence (above 0, up to 1). Threshold for deciding a person is in view
mediapipeModel"full", "heavy" or "light"autoAll pose-tracking viewsPins the MediaPipe pose model. When omitted, the SDK picks light or full from the device's speed; heavy is never picked automatically
defaultDelegate"GPU" or "CPU""GPU"All pose-tracking viewsProcessing backend for pose detection. A ?delegate= URL parameter overrides the postMessage value
boundedPoseWaitbooleanfalseAll pose-tracking viewsWhile the user waits on a loading screen, a slow start of the background pose engine is raced by a build on the main thread, so tracking can start sooner on devices where the background start is slow. Accepts true or "true"
landmarkColorstring"#14FF00"All pose-tracking viewsColor of the skeleton overlay when the user's form is correct (# optional, it is added). Mistake highlighting always uses red or orange on top of it
isDrawingPosebooleantrueAll pose-tracking viewsWhen false, hides the skeleton overlay. Pose recognition, rep counting and mistake detection still run: only the drawing is off
includePoseBordersbooleantrueAll pose-tracking viewsThe out-of-frame guard: when body parts leave the frame, the skeleton turns red, a "step back" cue plays and rep counting waits until the user is fully visible. false turns the guard off
videoFit"cover" or "contain""cover"All pose-tracking views"contain" shows the full camera frame (letterboxed) instead of zooming to fill the view, useful when the full body must stay visible in tight layouts. Assessments use contain in landscape regardless
showSilhouettebooleantrueWorkout player, Challenge, CameraWhen false, the positioning (check frame) step is skipped and tracking starts at once. Camera details: Real-Time Motion Events
showFormGuidancebooleantrue (false in Camera)Workout player, Challenge, CameraShows a live depth bar (for example squat depth) that guides the user to the right depth, for exercises that have depth guidance set up. While it is shown, each counted rep also sends rep_depth_angle
streamFormGuidancebooleanfalseWorkout player, Challenge, CameraHides the built-in depth bar and streams its data to your app instead, as form_guidance_config and form_guidance_frame events (plus rep_depth_angle per rep), so you can draw your own. See Camera & Frame Events
includePoseDatastring[]noneEvery pose-tracking view except Assessments and GamesRaw pose data per frame and, in workouts, the real-time motion events. See Real-Time Motion Events
includeRealtimeAccuracybooleanfalseWorkout player, Challenge, Cameratrue streams a correct_position_accuracy event on every frame. See Real-Time Motion Events

Notes:

  • videoURL (feed a video file instead of the live camera) is in Testing & Simulation.
  • The confidence values, mediapipeModel, defaultDelegate and boundedPoseWait are read once, at launch.
  • There is no automatic "heavy" model for balance assessments: pass mediapipeModel: "heavy" if an assessment needs maximum accuracy.
Camera & Pose Detection Settings
Swift example:
1kinestex.createCameraView(
2    exercises: ["model-id-1", "model-id-2"],
3    currentExercise: $currentExercise,
4    user: nil,
5    style: nil,
6    isLoading: $isLoading,
7    customParams: [
8        "landmarkColor": "#FF5500",
9        "showSilhouette": true,
10        "mediapipeModel": "heavy",
11        "defaultDelegate": "GPU",
12        "shouldShowCameraSelector": true,
13        "videoFit": "contain" // show the full camera frame
14    ],
15    onMessageReceived: { message in /* ... */ }
16)