Face Recognition System using PCA and ANN

This folder contains a complete, end-to-end implementation of a Face Recognition System using Principal Component Analysis (PCA) for dimensionality reduction (Eigenfaces) and Artificial Neural Networks (ANN) for classification.

The project is structured for academic submission and includes:

• A clean, optimized Python implementation (face_recognition.py)
• An academic project report (report.md)
• A presentation outline and viva questions (presentation_and_viva.md)

Features

• Eigenfaces Generation (PCA): Reduces high-dimensional image data while keeping maximum variance.
• Neural Network Classification: Uses an MLPClassifier to map features to distinct identities.
• Accuracy Evaluation: Evaluates and plots the accuracy for different numbers of principal components ($k$).
• Imposter Detection: Uses Euclidean distance thresholding to reject unknown faces.

Prerequisites

Ensure you have Python installed. Then, install the required dependencies:

pip install numpy opencv-python scikit-learn matplotlib scipy

Project Structure

FaceRecognitionProject/
│
├── dataset/                  # Directory for your face image dataset
│   ├── person1/
│   │   ├── img1.jpg
│   │   └── img2.jpg
│   └── person2/
│       ├── img1.jpg
│       └── img2.jpg
│
├── results/                  # Directory where the accuracy graph will be saved
│
├── face_recognition.py       # Main Python script (Object-Oriented)
├── Face_Recognition_Experiment.ipynb # Jupyter Notebook version
├── generate_dummy_dataset.py # Script to capture custom dataset via webcam
├── requirements.txt          # Python project dependencies
├── run_project.bat           # Executable script to run the project easily (Windows)
├── report.md                 # Academic project report
├── presentation_and_viva.md  # Presentation outline & viva script
└── README.md                 # Project documentation and instructions

Next Steps / How to Run

Method 1: The Easy Way (Windows)

1. Double-click the run_project.bat file.
2. It will automatically install the necessary libraries and execute the main face recognition script.
3. If it tells you the dataset is empty, proceed to step 2 below to capture your face using your webcam.

Method 2: Manual Execution

1. Install Dependencies:

   pip install -r requirements.txt

2. Generate a Dataset (Optional but highly recommended for testing): If you don't have images ready, use the built-in webcam script.

   python generate_dummy_dataset.py

   • Enter your name when prompted.
   • Look at your webcam; it will automatically capture 20 face images and save them in dataset/your_name/.
   • Run the script again with a friend or different name to build a larger database!

3. Run the Main Script:

   python face_recognition.py

   (Alternatively, if you prefer an interactive environment, open Face_Recognition_Experiment.ipynb in Jupyter Notebook.)

4. View Results:

   • The script will extract Eigenfaces and train the ANN model.
   • It will evaluate accuracies across different $k$ values (number of Eigenfaces) and output the metrics.
   • Furthermore, it will generate an accuracy_graph.png inside the results/ folder visually showing the optimal $k$ value.

5. Testing an Individual Image (Optional): Place an image named test.jpg in the root FaceRecognitionProject folder alongside the script. When you run face_recognition.py, it will automatically attempt to recognize who it is, or reject the face as an Imposter if not found in the trained database.