Facial Recognition System using Siamese CNN

A full-stack facial recognition system built using a Siamese Convolutional Neural Network (CNN) in PyTorch.
The system performs face verification by learning embeddings and comparing similarity scores.
It is served through a FastAPI backend, connected to a Next.js frontend, and fully containerized using Docker Compose for reproducible deployment.

Demo

Watch the application in action: View Demo Video

---

Installation & Setup

1. Frontend Setup

2. Backend & Model Setup

The backend is served via FastAPI and runs inside a Docker container.

cd Backend
docker compose up --build

This command builds the backend image and starts the inference service at http://127.0.0.1:8000.

---

Tech Stack

• Frontend: Next.js, React, Tailwind CSS
• Backend: FastAPI (Python)
• Machine Learning: XGBoost, Scikit-learn, Pandas, Numpy
• DevOps: Docker, Docker Compose

---

Machine Learning Workflow

The system uses a Siamese Neural Network for face verification based on similarity learning.

1. Data Collection: Face image pairs for similarity learning
2. Preprocessing: Resize, normalization, augmentation
3. Model: Siamese CNN for embedding generation
4. Distance Metric: Euclidean or Cosine similarity
5. Inference: Compare embeddings for match / no-match decision

---

Project Structure

├── Frontend/           # Next.js application
├── Backend/            # FastAPI, Model logic, and Dockerfile
│   ├── facial_recognition_model.py       # Trained Siamese CNN model
│   └── facial_api.py         # API endpoints

---