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
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.
