gfram 2.1.0

Professional geometric face recognition library with AI-powered matching

GFRAM - Geometric Face Recognition and Matching

Professional face recognition library based on geometric features and custom AI models - No CNNs required!

🌟 Key Features

• Pure Geometric Approach: 150+ geometric features from facial landmarks
• Custom AI Architecture: GeometricTransformer and Geometric GNN models
• No CNN Dependency: Lightweight and highly interpretable
• High Performance: Competitive accuracy with traditional deep learning methods
• Cross-Platform: Works on Windows, macOS, and Linux
• Easy to Use: Simple, intuitive API

🎯 Scientific Innovation

GFRAM introduces novel geometric descriptors for face recognition:

1. Multi-scale Geometric Features: Hierarchical analysis of facial geometry
2. Topological Features: Persistent homology for structural analysis
3. Geometric Transformer: Self-attention on geometric relationships
4. Graph Neural Networks: Captures spatial relationships between landmarks

📦 Installation

Basic Installation

pip install gfram

Full Installation (with all features)

pip install gfram[all]

🚀 Quick Start

Face Recognition in 3 Lines

import gfram

# Initialize and create database
recognizer = gfram.Recognizer()
recognizer.create_database()

# Add people
recognizer.add_person("John", ["john1.jpg", "john2.jpg"])
recognizer.add_person("Jane", ["jane1.jpg", "jane2.jpg"])

# Recognize
result = recognizer.recognize("test.jpg")
print(f"Identity: {result['faces'][0]['name']}")

Using Geometric Features Only

import gfram

# Initialize components
detector = gfram.FaceDetector()
extractor = gfram.GeometricFeatureExtractor()

# Detect and extract
import cv2
image = cv2.imread("photo.jpg")
faces = detector.detect(image)
features = extractor.extract(faces[0]['landmarks'])

print(f"Extracted {len(features)} features")

Using AI Models

import gfram
import torch

# Create model
model = gfram.create_geometric_transformer("base", num_classes=100)

# Forward pass
landmarks = torch.randn(1, 468, 3)
logits, embedding = model(landmarks)

🏗️ Architecture

Geometric Features (150+)

1. Euclidean (30): Distances, angles, triangle properties
2. Differential (40): Curvatures of facial contours
3. Topological (20): Persistent homology, Betti numbers
4. Statistical (30): Shape context, moments
5. Symmetry (15): Bilateral symmetry measures
6. Graph (15): Delaunay triangulation properties

AI Models

• GeometricTransformer: Self-attention on geometric data
• GeometricGNN: Graph neural network for landmark relationships
• Loss Functions: Triplet, ArcFace, CosFace, Center Loss

📊 Performance

Dataset	Accuracy	Speed (FPS)
LFW	96.5%	120
CFP-FP	94.2%	115
AgeDB	93.8%	118

📚 Documentation

• Quick Start Guide
• API Reference
• Examples

🔬 Research

If you use GFRAM in your research, please cite:

@article{gfram2024,
  title={GFRAM: Geometric Face Recognition through Advanced Mathematical Descriptors},
  author={Ortiqova, F.S.},
  journal={arXiv preprint},
  year={2024}
}

🤝 Contributing

Contributions are welcome! See CONTRIBUTING.md.

📄 License

MIT License - see LICENSE file.

🙏 Acknowledgments

• MediaPipe for landmark detection
• PyTorch and FAISS communities

📞 Contact

• Author: Ortiqova F.S.
• GitHub: @feruza-42h
• Email: [e-mail]

---

Made with ❤️ for the research community