Gesture Liveness Detection System

A real-time computer vision application that uses MediaPipe to detect human liveness through various gesture and facial recognition tasks.

Features

• Face Detection: Blink detection, smile recognition, head nodding
• Hand Gestures: Open hand, fist, finger counting, cheek touching
• Real-time Processing: Live camera feed with instant feedback
• Task Randomization: 3 random tasks selected from available pool
• Progress Tracking: Visual progress indicators and scoring system

Requirements

• Python 3.7+
• Webcam/Camera
• Windows/macOS/Linux

Quick Setup

1. Install dependencies:

   python setup.py

2. Run the application:

   python run.py

   Or directly:

   python gesture_liveness.py

Manual Setup

1. Install required packages:

   pip install -r requirements.txt

2. Ensure model files are present:

   • face_landmarker.task
   • hand_landmarker.task
   • gesture_recognizer.task

Available Tasks

Face Tasks

• Blink Twice: Blink your eyes twice in succession
• Smile: Show a big smile for several frames
• Nod Head: Nod your head down once

Hand Tasks

• Open Hand: Show an open hand with all fingers extended
• Make Fist: Close your hand into a fist
• Show N Fingers: Display exactly 2 or 3 fingers
• Touch Cheek: Touch your left, right, or either cheek with your finger

How It Works

1. Task Selection: 3 random tasks are selected from the available pool
2. Real-time Detection: Camera feed is processed using MediaPipe
3. Stability Checking: Tasks require stable detection over multiple frames
4. Timeout System: Each task has a 15-second time limit
5. Scoring: Pass/fail based on successful completion (2/3 tasks needed)

Controls

• ESC: Exit during tasks
• Any Key: Close after completion

Troubleshooting

Camera Issues

• Ensure webcam is connected and not used by other applications
• Try different camera indices if default doesn't work

MediaPipe Errors

• Ensure all .task model files are in the project directory
• Check MediaPipe version compatibility

Performance Issues

• Close other applications using the camera
• Ensure good lighting conditions
• Position yourself clearly in the camera frame

Technical Details

• MediaPipe Tasks API: Used for face and hand landmark detection
• OpenCV: Camera capture and image processing
• Real-time Processing: ~30 FPS with optimized detection algorithms
• Stability Algorithms: Multi-frame validation for reliable detection

File Structure

├── gesture_liveness.py      # Main application
├── requirements.txt         # Python dependencies
├── setup.py                # Setup script
├── run.py                  # Quick launcher
├── face_landmarker.task    # MediaPipe face model
├── hand_landmarker.task    # MediaPipe hand model
└── gesture_recognizer.task # MediaPipe gesture model

Development

Debug Scripts

• debug_mp.py: Test MediaPipe installation
• debug_mp_fix.py: MediaPipe compatibility checks
• debug_mp_fix_2.py: Alternative MediaPipe imports

Adding New Tasks

1. Create a new task class inheriting the task interface
2. Implement the process(face_result, hand_result) method
3. Add to the task_pool in LivenessApp.__init__()

License

This project is for educational and demonstration purposes.