Real Time Face Recognition Using CNN

A robust real-time face recognition system implemented using Convolutional Neural Networks (CNN). This project provides an end-to-end solution for face detection, data collection, model training, and real-time recognition using Python and deep learning techniques.

Key Features

• Real-time face detection and recognition using webcam
• Custom CNN architecture for improved accuracy
• Easy-to-use data collection interface
• Automatic face detection and cropping
• Real-time FPS counter and performance metrics
• Support for multiple face recognition
• Batch training with validation split
• Model checkpointing for best weights

System Requirements

Hardware

• Webcam (built-in or external)
• Minimum 4GB RAM (8GB recommended)
• CPU with SSE2 instruction set support

Software

• Python 3.7 or higher
• Operating System: Windows 10/11, macOS, or Linux

Dependencies

• OpenCV (cv2)
• TensorFlow 2.x
• Keras
• NumPy
• PIL (Python Imaging Library)
• dlib
• imutils
• h5py
• scikit-learn

Installation

1. Clone the repository:

git clone https://github.com/yourusername/Real-Time-Face-Recognition-Using-CNN.git
cd Real-Time-Face-Recognition-Using-CNN

2. Install required packages:

pip install -r requirements.txt

Usage Guide

1. Data Collection

python src/data_collection/face_dataset.py

This script will:

• Activate your webcam
• Detect faces in real-time
• Capture 70 images of your face from different angles
• Save the processed images in the dataset directory
• Display real-time FPS and image capture count

Tips for better data collection:

• Ensure good lighting conditions
• Move your face slowly to capture different angles
• Maintain a distance of 0.5-1 meter from the camera
• Try different facial expressions

2. Model Training

python src/training/face_training.py

This will:

• Process the collected face images
• Create and train the CNN model
• Save the best model weights to models/trained_model.h5
• Display training progress and validation metrics

3. Face Recognition

python src/recognition/face_recognition.py

Features:

• Real-time face detection and recognition
• FPS counter for performance monitoring
• Confidence score display
• Multiple face detection support

Configuration

Adjust parameters in config.py to customize the system:

Data Collection

• MIN_FACE_SIZE: Minimum face size to detect
• SCALE_FACTOR: Scale factor for face detection
• MIN_NEIGHBORS: Minimum neighbors for face detection

Training

• IMAGE_SIZE: Input image dimensions
• BATCH_SIZE: Training batch size
• EPOCHS: Number of training epochs
• MODEL_PATH: Path to save trained model

Recognition

• FACE_CASCADE_PATH: Path to face cascade classifier
• DATASET_PATH: Path to face dataset

Troubleshooting

1. Low FPS during recognition:

   • Reduce MIN_FACE_SIZE in config.py
   • Ensure no other CPU-intensive programs are running

2. Poor recognition accuracy:

   • Collect more training data with varied lighting/angles
   • Increase EPOCHS in config.py
   • Adjust SCALE_FACTOR and MIN_NEIGHBORS

3. Face not detected:

   • Check lighting conditions
   • Adjust MIN_FACE_SIZE and SCALE_FACTOR
   • Ensure face is clearly visible to camera

Project Structure

Real-Time-Face-Recognition-Using-CNN/
├── src/                           # Source code directory
│   ├── data_collection/           # Data collection scripts
│   │   └── face_dataset.py        # Face image capture script
│   ├── training/                  # Model training scripts
│   │   ├── face_training.py       # Main training script
│   │   └── model.py              # CNN model architecture
│   └── recognition/               # Recognition scripts
│       └── face_recognition.py    # Real-time recognition
├── models/                        # Trained model storage
│   └── trained_model.h5           # Trained weights file
├── dataset/                       # Face image dataset
├── requirements.txt               # Project dependencies
├── config.py                      # Configuration parameters
└── README.md                      # Project documentation

Contributing

Contributions are welcome! Please feel free to submit issues and enhancement requests.

1. Fork the repository
2. Create your feature branch
3. Commit your changes
4. Push to the branch
5. Create a Pull Request

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• OpenCV team for the face detection cascades
• TensorFlow and Keras teams for the deep learning framework
• The open-source community for various dependencies