Branch	Version	Status
master	0.3.2

Platform	Python	System Dependencies
Linux (X11/Wayland)	3.12+	libgl1, libglib2.0-0, notify-send, xdotool, ydotool
macOS (ARM64)	N/A	N/A
Windows (x86_64)	N/A	N/A

• Docs: [Link to documentation]

Introduction

Gesture Vision is a gesture control system, developed in Python to automate desktop actions through real-time gesture recognition.

The system works through a computer vision pipeline where MediaPipe processes the camera stream to detect hand landmarks, coordinated by an asynchronous FastAPI backend. This provides precise event control with minimal impact on system resources.

Architecture

Main Features

• High Precision Tracking: MediaPipe Hand Landmarker implementation for low-latency hand landmark detection.
• Robust Trigger Control: Validation system based on gesture hold time and cooldown periods to eliminate false positives.
• Swipe Gestures: Workspace switching with horizontal hand movements - left/right.
• Screenshot Capture: Detection of the "peace" gesture (index and middle fingers up) to take screenshots.
• Invisible Mode (Background): Fully background operation without preview windows, optimizing GPU/CPU usage.
• Keyboard Shortcuts: Associate gestures with system shortcuts like ctrl+shift+q to execute any action.
• X11 and Wayland Support: Automatically detects whether to use xdotool (X11) or ydotool (Wayland).
• Native Notifications: Integration with desktop notification system to instantly confirm actions.
• Native Deployment: Simplified installation via pip with dedicated CLI commands for service management.
• Web Panel: Graphical interface to configure gestures, shortcuts, and timing.

Quick Start

1. Installation

# Install from PyPI
pip install gesturevision

2. Run

# First time: automatically install system dependencies
gesturevision-start --install-deps

# Start the system
gesturevision-start

4. Control Panel

Access the web panel: http://localhost:8080

There you can:

• Start/stop the system
• Configure gesture and cooldown timing
• Add gestures and associate keyboard shortcuts

Available Gesture Types

Gesture	Description
✌️ Peace	Index and middle fingers up
✊ Fist	Closed hand
✋ Open Hand	5 fingers extended
👍 Three Fingers	Index, middle, and ring fingers up
☝️ Point	Only index finger up

Supported Keyboard Shortcuts

You can use any of the predefined shortcuts or write your own:

• Screenshot: Print
• Close window: alt+F4
• Minimize all: super+d
• Lock screen: super+l
• Switch window: alt+Tab
• New tab: ctrl+t
• Close tab: ctrl+w
• Copy/Paste: ctrl+c / ctrl+v
• And many more...

Project Structure

• src/gesturevision/main_vision.py: Gesture detection and capture engine.
• src/gesturevision/api/: FastAPI backend for control and configuration.
• src/gesturevision/static/: Control panel user interface.
• pyproject.toml: Package dependencies and entry points definition.
• gesturevision/assets/hand_landmarker.task: Pre-trained AI model.

Configuration

Settings are stored in: ~/gesturevision/config.json

Structure:

{
  "gesture_hold_seconds": 1.0,
  "cooldown_seconds": 1.5,
  "gestures": [
    {
      "name": "Peace (screenshot)",
      "gesture_type": "peace_sign",
      "required_hold_seconds": 1.0,
      "enabled": true,
      "action": "screenshot",
      "shortcut": null
    }
  ]
}

Docker (Optional)

docker-compose up --build

Access the panel at: http://localhost:8080