---

Face Recognition System (SVM + PCA)

This repository contains a Face Recognition System implemented using Supervised Machine Learning techniques.
The system uses Principal Component Analysis (PCA) for feature reduction and Support Vector Machine (SVM) for face classification.

The project is fully dynamic, allowing users to add new people, recreate the dataset, retrain the model, and test recognition results.

📁 Project Structure

├── .ipynb_checkpoints/
│
├── train_data_2/                     # Training face images
├── test_data/                        # Testing face images
│
├── data.npy                          # PCA-transformed face features
├── target.npy                        # Class labels
│
├── haarcascade_frontalface_default.xml  # Face detection model
│
├── fscrRecodDatasetMaking.ipynb      # Dataset creation & feature extraction
├── trainModel(SVM).ipynb             # PCA + SVM training
├── testModel.ipynb                   # Model testing
│
├── SVM-FaceRecognition.sav            # Trained SVM model
└── README.md

---

🧠 Machine Learning Approach

🔹 Supervised Learning

• The system follows a supervised learning approach.
• Each face image is labeled with the corresponding person name.
• Labels are used to train the classifier.

🔹 PCA (Principal Component Analysis)

• PCA is applied to:
  • Reduce dimensionality
  • Remove noise
  • Improve training speed
• Converts face images into compact feature vectors.

🔹 SVM (Support Vector Machine)

• SVM is used as the final classifier.
• Learns decision boundaries between different individuals.
• Performs well on high-dimensional face data.

---

⚙️ Technologies Used

• Python
• OpenCV
• NumPy
• Scikit-learn
• Jupyter Notebook

---

📊 Dataset Structure

Training Data

train_data_2/
├── saman_kumara/
│   ├── img1.jpg
│   ├── img2.jpg
│   └── ...
├── kasun_perera/
└── nimal_silva/

Testing Data

test_data/
├── saman_kumara/
├── kasun_perera/

Each folder represents one class (one person).

---

🔁 Train Your Own Model (Dynamic Dataset)

This system supports dynamic training.

✅ Step 1: Add New Person

1. Navigate to train_data_2/
2. Create a new folder with the person’s name
   Example:

train_data_2/saman_kumara/

3. Add multiple face images of that person.

---

✅ Step 2: Create Dataset

Run:

fscrRecodDatasetMaking.ipynb

This will:

• Detect faces

• Apply PCA

• Generate:

  • data.npy
  • target.npy

---

✅ Step 3: Train the Model

1. Open:

   trainModel(SVM).ipynb
   

2. Ensure the new person’s name is included in the label list.

3. Train the SVM classifier.

Output:

SVM-FaceRecognition.sav

---

✅ Step 4: Test the Model

1. Add images to:

   test_data/saman_kumara/
   

2. Run:

   testModel.ipynb
   

The model will predict the person’s identity.

---

📈 Model Summary

Component	Description
Learning Type	Supervised Learning
Dimensionality Reduction	PCA
Classifier	SVM
Output	Person Name
Model File	SVM-FaceRecognition.sav

---

🔐 Ethical Notice

This project is intended only for academic and educational purposes. Avoid using it for unauthorized surveillance or privacy-sensitive applications.

---

✍️ Author

Madhuka Malshan Software Engineering Student National Institute of Business Management (NIBM), Sri Lanka

---