| --- |
| license: apache-2.0 |
| --- |
| |
| # 🧠 FaceGNN v1.1 |
|
|
| FaceGNN is a graph-based facial recognition system that combines deep learning feature extraction with Graph Neural Networks (GNNs) for robust secure identity classification. It supports embedding generation through both augmented and non-augmented versions using Graph neural networks. This reliable system has been thoroughly evaluated and consistently demonstrates superior performance, making it well-suited for real-world, high-security applications. |
| --- |
|
|
|
|
| ## 📊 Project Overview |
|
|
| This study proposes a novel approach for face recognition by leveraging Graph Neural Networks (GNNs) to classify face embeddings extracted from face images. The methodology comprises three main phases: |
| - dataset preparation and embedding generation. |
| - graph construction using k-Nearest Neighbors (k-NN). |
| - training and evaluation of a GNN-based classifier. |
|
|
| --- |
|
|
| ## 🖼️ Model Architecture |
|
|
|  |
|
|
|
|
|
|
|
|
| ## ✅ Performance & Security |
|
|
| The proposed FaceGNN model ensures high security and offers a robust method for face recognition in real-world scenarios. By leveraging relational structures through graph-based learning, it effectively handles both intra-class variations and inter-class similarities. The model achieved up to 99% accuracy across a dataset of over 800 individuals, demonstrating its strong generalization and reliability for high-stakes identity verification tasks. |
|
|
|
|
