Hand Sign Detector

A computer vision project that detects and classifies hand signs in real-time using Google's MediaPipe and OpenCV.

Features

• Real-time hand landmark detection
• Classification of 5 distinct hand signs:
  • Claws
  • Frogs
  • Gigem
  • Guns Up
  • Horns
• Custom descriptor based on normalized landmark distances
• "Truth data" clustering for robust classification
• Clean, modular class-based architecture

Installation

1. Clone the repository:

   git clone https://github.com/jacobgeorge3/HandSignDetector.git
   cd HandSignDetector

2. Install dependencies:

   pip install -r requirements.txt

Usage

Run the main script to start the webcam feed and detection:

python handsign.py

Press q to quit the application.

Project Structure

• handsign.py - Main application with HandSignDetector class
• classifier.py - Hand landmark descriptor creation and distance metrics
• static_handsign.py - Truth data generation from static images
• cluster.py - K-means clustering for gesture centroids
• config.py - Configuration constants
• gesture.names - List of gesture class names
• images/ - Training images for each gesture
• test_refactor.py - Unit tests

How it Works

The system uses MediaPipe to extract 21 hand landmarks. These landmarks are normalized and converted into a custom descriptor vector. This vector is compared against pre-computed cluster centers of "truth data" using a pseudo-Euclidean distance metric. A ratio test is applied to ensure high-confidence classifications.

Testing

Run the test suite:

python test_refactor.py

Original Project

This is a refactored version of the CVFinalProj repository, with improved code organization, type hints, and error handling.