Skip to main content

Introduction

Welcome to the Gesture Control with OpenVINO documentation. This comprehensive guide covers the implementation of gesture control systems using Intel's OpenVINO toolkit.

Overview​

This documentation is organized into several key sections:

  • Model Architecture: Learn about the four main models that power the gesture recognition system
  • Gesture Types: Understand static and dynamic gesture recognition
  • Control System: Explore the GUI and input mapping components
  • Performance: Analyze metrics and optimization on Intel devices

Getting Started​

To get started with gesture control, we recommend reading through the model architecture sections first, then exploring the gesture control system implementation.

Prerequisites​

  • Intel OpenVINO toolkit
  • Python 3.7+
  • Camera device for gesture capture
  • Intel hardware (recommended for optimal performance)

Let's begin your journey into gesture control technology!

Gesture Recognition Pipeline at a Glance​

Pipeline Overview

This pipeline processes the camera input through four stages — Hand Detector, Hand Landmarks, Gesture Embedder, and Canned Gesture Classifier — to produce robust gesture outputs used by the control system.

Hand Landmarks (21 points)​

Hand Landmarks

The landmark model predicts 21 key points per hand. These landmarks drive gesture embedding/classification and custom dynamic gesture logic used by application modes.