Hand Tracking (MediaPipe + OpenCV)

Real-time hand tracking and annotation using a webcam. The script uses MediaPipe Tasks API for landmark detection, OpenCV for video capture/display, and NumPy for image processing.

Features

• Real-time hand landmark detection (up to 2 hands)
• Handedness labeling (Left / Right)
• Live video annotation overlay

Example

Example of real-time output (mirored webcam view):

Requirements

• Python 3.9+ (required for MediaPipe Tasks API compatibility)
• Webcam (built-in or external)

Installation

Install dependencies:

pip install -r requirements.txt

Or manually:

pip install mediapipe opencv-python numpy

Model File

This repository already includes the required model file:

• hand_landmarker.task

If you want to download the latest version manually, use the official MediaPipe model page: https://ai.google.dev/edge/mediapipe/solutions/vision/hand_landmarker/index#models

Run

python hand-track.py

Controls

• ESC: Exit the application.