Face_Check

A real-time face verification system that compares faces from ID cards with live camera feed using advanced face recognition technology.

Sample Demo

Sample ID used

Features

• Real-time Face Recognition: Live camera feed with instant face matching
• ID Card Processing: Upload and extract faces from ID cards or photos
• High Accuracy: Uses InsightFace's ArcFace Buffalo model for precise face embeddings
• User-friendly Interface: Clean Streamlit web interface
• Similarity Scoring: Shows confidence scores for face matches
• Instant Feedback: Visual indicators (green for match, red for no match)

Technology Stack

• InsightFace: State-of-the-art face recognition and analysis
• OpenCV: Computer vision and image processing
• Streamlit: Web application framework
• StreamLit-WebRTC: Real-time video streaming
• SciPy: Distance calculations for face similarity
• NumPy: Numerical operations

Installation

1. Clone the repository:

   git clone https://github.com/Vjay15/Face_Check.git
   cd Face_Check

2. Install required dependencies:

   pip install -r requirements.txt

Usage

1. Start the application:

   streamlit run app.py

2. Upload ID Card:

   • Click "Upload your ID Card" in the left panel
   • Select a JPG, JPEG, or PNG file containing a clear face
   • The system will automatically extract and display the face

3. Live Verification:

   • Once an ID is uploaded, the face will be extracted and click on start, to start the camera feed
   • Position your face in front of the camera
   • The system will show real-time similarity scores
   • Green box = Match found
   • Red box = No match

How It Works

1. Face Extraction: Uses OpenCV's CascadeClassifier and HaarCascade Algorithm to detect and extract faces
2. Face Detection: Uses InsightFace's Buffalo model to detect faces in uploaded images
3. Feature Extraction: Generates normalized face embeddings for comparison
4. Real-time Processing: Continuously processes camera frames for face detection
5. Similarity Calculation: Uses cosine distance to measure face similarity
6. Threshold Matching: Considers faces with >60% similarity as matches

Configuration

• Similarity Threshold: Currently set to 0.6 (60% similarity)
• Camera Resolution: 720x200 pixels output for optimal performance
• Processing: Configured for GPU acceleration (falls back to CPU)

Supported File Formats

• Input Images: JPG, JPEG, PNG
• Camera: Any webcam compatible with WebRTC

Requirements

• Python 3.7+
• Webcam/Camera access
• Modern web browser with WebRTC support

License

This project is open source and available under the MIT License.

Acknowledgments

• InsightFace team for the excellent, compact and highly efficient face recognition models
• OpenCV community for computer vision tools
• Countless Medium articles and Blogs of people that taught me how these algorithms worked!