Internship Studio Artificial Intelligence Internship

A face recognition system using Multi-Layer Perceptron (MLP) neural networks.

Project Overview

This project implements a face recognition system using MLP classifier along with Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA) for dimensionality reduction. The system can identify faces from a dataset of Bollywood celebrities with good accuracy.

Key Features

• Face detection and recognition using MLP neural network
• Dimensionality reduction using PCA and LDA
• Dataset of Bollywood celebrity faces
• Interactive visualization of eigenfaces
• Model training with loss tracking
• Prediction probability scores
• Accuracy evaluation

Technical Details

Technologies Used

• Python
• Jupyter Notebook
• Libraries:
  • scikit-learn (MLPClassifier, PCA, LDA)
  • OpenCV (cv2)
  • NumPy
  • Matplotlib

Model Architecture

• MLP Neural Network with:
  • Input layer: 150 PCA components
  • Hidden layers: (10, 10) neurons
  • Output layer: Number of classes
  • Maximum iterations: 1000

Dataset Structure

faces/
├── Aamir/
├── Ajay/
├── Akshay/
├── Alia/
├── Amitabh/
├── Deepika/
├── Disha/
├── Farhan/
└── Ileana/

Performance

• Achieved accuracy: ~77.88%
• Model convergence with decreasing loss values
• Early stopping when loss improvement < 0.0001

How to Run

1. Clone the repository
2. Install required dependencies:

pip install numpy opencv-python scikit-learn matplotlib jupyter

3. Open MLP_Face_Recognition.ipynb in Jupyter Notebook
4. Run all cells sequentially

Project Structure

.
├── faces/                  # Dataset directory
├── MLP_Face_Recognition.ipynb  # Main notebook
└── README.md              # Project documentation

Results

• Successful face recognition with probability scores
• Visualization of eigenfaces
• Loss tracking during training
• Test set predictions with confidence scores

Author

Chhagan Ram Choudhary