Computer vision visualizer

Pose Landmarks Visualizer

Detect 33 body landmarks, inspect joint angles, and recognize common poses directly in your browser with MediaPipe.

Controls

Input mode
Overlay layers

Keep the full body visible. The Lite model is used for responsive webcam tracking.

Preview

Upload a full-body photo, use webcam, or inspect the sample guide.
Pose landmarks input
Hand landmarksFace mesh

Joint Angles

Left elbow--Right elbow--Left knee--Right knee--

What this visualizer does

It runs MediaPipe PoseLandmarker on an image or webcam frame, draws a 33-point body skeleton, and derives joint angles and pose states.

What MediaPipe detects

  • Normalized image landmarks for overlay drawing.
  • World landmarks for 3D body measurements.
  • Visibility scores for each body landmark.

Common use cases

  • Exercise form
  • Posture feedback
  • Fitness counters
  • Motion interfaces
  • Sports analysis

Code examples

import mediapipe as mp

BaseOptions = mp.tasks.BaseOptions
PoseLandmarker = mp.tasks.vision.PoseLandmarker
PoseLandmarkerOptions = mp.tasks.vision.PoseLandmarkerOptions

options = PoseLandmarkerOptions(
    base_options=BaseOptions(model_asset_path="pose_landmarker.task"),
    num_poses=1,
    min_pose_detection_confidence=0.5
)

with PoseLandmarker.create_from_options(options) as landmarker:
    image = mp.Image.create_from_file("pose.jpg")
    result = landmarker.detect(image)
    print(result.pose_landmarks)

How the model works

  1. A detector locates each person in the frame.
  2. The landmark model predicts 33 body points.
  3. Video mode tracks landmarks between frames.
  4. Joint geometry produces angles and pose states.

Controls explanation

Max poses limits simultaneous people. Confidence filters weak detections, while overlay buttons control skeleton, points, and bounds.

FAQ

Keep shoulders, hips, knees, and ankles visible for reliable pose recognition. Loose framing works better than a close crop.