Hand Recognition & Gesture Recognition - Drone Control System

Advanced gesture and air-drawing recognition for drone/device control using MediaPipe.

USP (Unique Selling Point)

This system recognizes 8 distinct hand gestures and 4 air-drawn shapes, mapping them to drone commands:

8 Hand Gestures

Gesture	Command
Left Pointing (index extended left)	LEFT MOVE
Right Pointing (index extended right)	RIGHT MOVE
Up Pointing (index extended up)	UPWARD MOVE
Down Pointing (index extended down)	DOWNWARD MOVE
Victory Sign (index + middle extended)	VICTORY CELEBRATION 🎉
Rock Sign (index + pinky extended)	BACKFLIP!
Open Palm (all fingers spread)	STOP IMMEDIATELY ⚠️
Closed Fist (all fingers closed)	LAND

Air-Drawn Shapes

Shape	Command
Circle	CIRCLE MOVEMENT
Square	SQUARE PATTERN
Triangle	TRIANGLE PATTERN
Line	LINE MOVEMENT

How It Works

1. Hand Landmark Detection

• Uses MediaPipe Hand Landmarker (official Google ML model)
• Detects 21 hand landmarks per hand in real-time
• Tracks 2 hands simultaneously

2. Gesture Classification

• Landmark-based analysis: Checks finger extension states (tips vs joints)
• Direction detection: Analyzes relative position of fingers to wrist
• Smoothing: Uses temporal filtering to reduce jitter and false positives

3. Air Drawing Recognition

• Motion tracking: Detects when hand is still (drawing mode)
• Trail collection: Records sequence of hand positions
• Shape analysis:
  • Circularity measurement
  • Corner detection for polygons
  • Aspect ratio analysis
  • Direction change analysis

4. Command Mapping

• Real-time gesture → command translation
• Visual feedback on screen with color-coded commands
• Critical commands (STOP, LAND) trigger frame borders

Setup

1. Create Virtual Environment

cd /Users/aksh-aggarwal/Desktop/Workspace/aims/myPro
./run.sh setup

2. Run the Demo

./run.sh run

Or manually:

source myENV/bin/activate
python hand_gesture.py

Features

✅ Real-time Hand Detection - Processes video at ~30 FPS
✅ Multi-hand Support - Recognizes up to 2 hands simultaneously
✅ Robust Gesture Recognition - Uses temporal smoothing for stability
✅ Air Drawing - Tracks and recognizes shapes drawn in air
✅ Visual Feedback - Color-coded commands with live trail display
✅ Critical Commands - Special highlighting for safety (STOP, LAND)

Hand Gesture Demo — Beginner Friendly

This project demonstrates simple real-time hand gesture recognition using MediaPipe. It's intended for learners who want a hands-on demo that maps a few hand gestures to commands (for example, to control a drone simulator or for UI experiments).

Quick overview:

• Detects hand landmarks (21 points per hand)
• Classifies a small set of gestures like pointing, open palm, victory, rock, and closed fist
• Shows live visual feedback in a window

Prerequisites

• Python 3.8+ and a webcam
• Create and activate the included virtual environment or install packages from requirements.txt.

Quick start

1. Activate the virtual environment (if present):

source myENV/bin/activate

2. Run the demo:

python hand_gesture.py

What you'll see

• A camera window with hand landmarks drawn
• Detected gesture shown on-screen and a simple command label
• Press q to quit

Files you need

• hand_gesture.py — main demo (run this)
• gesture_classifier.py — logic that interprets hand landmarks
• command_mapper.py — maps gestures to text/color for display
• requirements.txt — Python dependencies

Notes

• The project previously included an air-drawing helper (air_drawing.py). It is not required by the main demo and has been removed to keep things simple for beginners.
• The model file hand_landmarker.task will be downloaded automatically if missing.

Need help?

• Tell me what OS or Python version you're using and I can give exact install commands.

Enjoy experimenting with hand gestures!

Hi,

This is my small hand gestures project. I am learning and i wrote this. It is simple.

What it does

• Uses mediapipe to find hand points
• Tries to tell few gestures like point, open palm, victory, rock, closed fist
• Shows a window with landmarks and a text

What you need

• Python 3
• A webcam
• Install packages from requirements.txt or use the included env

How to run (simple)

source myENV/bin/activate
python hand_gesture.py

Notes

• The model file will download automatically when you run
• Press q to quit the window
• I removed some extra code (air drawing) because i was not using it in main script

Files (important)

• hand_gesture.py (main script)
• gesture_classifier.py (classify gestures)
• command_mapper.py (maps gestures to text/colors)
• requirements.txt

If something breaks tell me and i try to fix it.

Thanks