Vinyasa - Head Tracking Mouse

Vinyasa is a production-grade computer vision application that translates head movements into mouse pointer control. It uses a machine learning model for precise head position detection and PyAutoGUI for smooth system-level mouse interaction.

Features

• Minimalist Design: Clean, high-DPI aware, light-themed interface built to exact design specifications.
• Traffic Light Controls: Integrated Mac-style traffic light window controls for a sleek aesthetic.
• Cross-Platform Compatibility: Supports Windows, macOS, and Linux with automated hardware backend selection.
• Smooth Mouse Control: Optimized PyAutoGUI integration with zero-latency movement and custom acceleration curves.
• Production Architecture: Modularized Python package structure for high maintainability and scalability.
• Calibration Wizard: Interactive tool to adjust sensitivity and movement range to your personal comfort.
• Hotkeys: Quick toggle tracking with F10.
• Persistent Settings: Your configuration and calibration profiles (Sensitivity, Lines per Second, Cursor Size) are automatically saved using Pydantic validation.

Installation

1. Clone the repository:

   git clone https://github.com/bhumitschaudhry/vinyasa.git
   cd vinyasa

2. Install dependencies:

   pip install tensorflow opencv-python pyautogui pillow pydantic customtkinter darkdetect

System Requirements

macOS

• Permissions: You must grant Accessibility permissions to your Terminal or IDE (System Settings > Privacy & Security > Accessibility) to allow mouse control.
• Camera: Grant camera access when prompted.

Linux

• X11 vs Wayland: Mouse control is most stable on X11. If using Wayland, you may need to switch to an X11 session for full functionality.
• Dependencies:

  sudo apt-get install python3-tk python3-dev

Usage

1. Run the application:

   python src/main.py

2. Start Tracking: Click the "Start Tracking" button or press F10.
3. Calibrate: Use the "CALIBRATE" button (available in settings) to open the wizard and adjust sensitivity.

Technical Details

• Hardware Backends: Automatically selects CAP_DSHOW (Windows), CAP_AVFOUNDATION (macOS), or CAP_V4L2 (Linux) for the camera.
• Smoothing: Uses a 3-frame prediction history buffer to eliminate tracking jitter.
• Acceleration: Dynamically increases mouse speed during sustained movement.
• Threading: Decouples camera capture from model inference for consistent high performance.

License

MIT License