Computer Vision

Pose Detection System

Real-time pose & hand detection using MediaPipe with 33 skeletal keypoints and 21 finger joints per hand

Full Body Tracking & Hand Detection

Real-time pose estimation and hand detection system using MediaPipe, applied to YouTube video feeds. Track full body movements with 33 skeletal keypoints and detailed hand analysis with 21 finger joint positions per hand.

The system uses browser-native MediaPipe models with a Python Flask proxy server for YouTube video streaming, enabling real-time skeleton and hand landmark rendering with performance metrics.

Detection Capabilities

Pose Points 33 Skeletal Keypoints
Hand Detection 21 Joints Per Hand
Framework MediaPipe + Flask
Video Source YouTube Live Feed

Core Features

Full Body Pose Detection

Track 33 skeletal keypoints including nose, shoulders, elbows, wrists, hips, knees, and ankles for complete body movement analysis.

Hand Landmark Detection

Identify both hands with 21 detailed finger joint positions per hand for precise gesture recognition and tracking.

Visual Overlay Rendering

Real-time skeleton and hand landmark visualization overlaid on video with smooth, responsive rendering.

Performance Metrics

FPS counter and detection point statistics displaying frames per second and number of detected keypoints.

Interactive Controls

Toggle pose and hand detection independently, adjust confidence thresholds from 0-100%, and live status indicators.

YouTube Video Integration

Flask proxy server enables CORS-compliant YouTube video streaming for browser-based MediaPipe processing.

Technical Specifications

Frontend
Vanilla JavaScript + HTML5 Canvas
Detection Models
MediaPipe Pose + Hands
Pose Keypoints
33 Body Landmarks
Hand Landmarks
21 Joints Per Hand
Backend
Python Flask + yt-dlp
Video Processing
Real-Time Browser Rendering
Model Loading
CDN (First Run)
Browser Support
Chrome/Edge Recommended

Ready to Track Human Movement?

Deploy real-time pose and hand detection with 33+21 landmark precision