Facial Recognition System

A complete real-time facial recognition system based on Eigenfaces and OpenCV, featuring an interactive camera interface and statistical analysis tools.

---

Key Features

• Real-time camera-based facial recognition
• Multiple face detection and identification
• Statistical analysis with charts and metrics
• Flexible configuration with adjustable parameters
• Optimized performance (25–30 FPS)
• Recognition accuracy between 80% and 95%

---

Table of Contents

• Installation
• Usage
• Architecture
• Configuration
• Performance
• Troubleshooting
• Project Files

---

Installation

Prerequisites

• Python 3.8 or higher
• Webcam
• 500 MB of free disk space

Steps

Clone or access the project directory:

Install dependencies:

pip install opencv-python numpy scikit-learn matplotlib seaborn

Verify installation:

python test_setup.py

---

Usage

Option 1: Interactive Menu (Recommended)

python main.py

Available options:

1. Full analysis (charts and statistics)
2. Real-time camera recognition
3. Exit

Option 2: Direct Camera Launch

python run_camera.py

Option 3: Python Integration

from camera_capture import run_face_recognition_camera

run_face_recognition_camera(
    dataset_path="./face_database",
    n_components=50,
    threshold=5000
)

---

Camera Controls

Key	Action
q	Quit and display statistics
Others	No effect

---

On-Screen Display

• Green rectangle: recognized face
• Red rectangle: unknown face
• Confidence distance displayed per face
• FPS, number of detected, recognized, and unknown faces shown in real time

---

Architecture

Face_recognition/
│
├── main.py              # Main entry point
├── camera_capture.py    # Recognition engine
├── config.py            # Central configuration
├── run_camera.py        # Direct camera execution
├── test_setup.py        # Installation test
│
├── face_database/       # Face dataset
│   ├── s1/
│   ├── s2/
│   └── ...
│
├── README.md
├── README_CAMERA.md
├── UTILISATION.md
└── RESUME.md

---

Configuration

Main Parameters

Location: camera_capture.py → run_face_recognition_camera()

run_face_recognition_camera(
    dataset_path="./face_database",
    n_components=50,
    threshold=5000
)

Parameter	Description
dataset_path	Path to the face dataset
n_components	Number of Eigenfaces (20–100)
threshold	Maximum distance for recognition

Recommended Settings

• Higher accuracy (slower) n_components=100, threshold=3500

• Higher speed (less accurate) n_components=30, threshold=7000

• Balanced (recommended) n_components=50, threshold=5000

---

Dataset Configuration

Required structure:

face_database/
├── s1/
├── s2/
└── s3/

Recommendations:

• 8–12 images per person
• JPG or PNG format
• Resolution between 100×100 and 500×500
• Multiple angles and expressions
• Good lighting conditions

---

Performance

Typical Results

Metric	Value
Training time	2–3 seconds
Recognition per frame	10–50 ms
Real-time FPS	25–30
Accuracy	80–95%
Memory usage	200–400 MB

Optimization Tips

• Reduce n_components
• Increase threshold
• Lower camera resolution
• Close background applications

---

Troubleshooting

Camera Not Opening

ls /dev/video*
sudo apt-get install cheese
cheese
sudo usermod -a -G video $USER

Missing Modules

pip install scikit-learn seaborn

Low Accuracy

• Increase n_components
• Decrease threshold
• Add more training images
• Improve lighting and image quality

Low FPS

• Reduce n_components
• Close other applications
• Monitor CPU and RAM usage

---

How Facial Recognition Works

1. Camera image capture (up to 30 FPS)
2. Face detection using Haar Cascade Classifier
3. Image preprocessing (resize and normalization)
4. Recognition using a pre-trained Eigenfaces model
5. Euclidean distance computation
6. Real-time visual feedback

Eigenfaces Algorithm

Advantages

• Fast execution
• Low memory usage
• Suitable for small datasets

Limitations

• Sensitive to lighting conditions
• Less accurate than deep learning models

---

Future Improvements

• Model saving and loading
• Deep learning integration (FaceNet, ArcFace)
• Graphical interface (PyQt or Tkinter)
• SQLite database
• Multi-threading
• Multi-camera support
• Video recording with annotations
• Statistics export

---