Face Recognition Project 👤🔍

This project implements a deep-learning-based face recognition system using a Convolutional Neural Network (CNN).
It includes image preprocessing, model training, evaluation, and insights into model performance and limitations.

---

📁 File Included

• Face-Recognition.ipynb (your uploaded notebook name)
  Contains:
  • Image loading and preprocessing
  • Train/test split
  • CNN model architecture
  • Training for 30 epochs
  • Accuracy & loss plots
  • Classification report
  • Predictions

If you renamed your file, replace the name above.

---

🧠 Model Overview

This project uses a custom CNN for image-based face recognition.
The pipeline includes:

• Convolution + MaxPooling layers
• Flatten & Dense layers
• Softmax output layer
• Adam optimizer
• Categorical cross-entropy loss

The notebook demonstrates a complete ML workflow from raw images to predictions.

---

📊 Model Performance

• Training Accuracy: ~76%
• Validation Accuracy: ~57%
• Validation loss increases in later epochs
• This indicates overfitting, which is expected for small datasets in face recognition tasks

Why this is normal

Face datasets with limited samples per class often cause the model to memorize training data rather than generalize.
This is a realistic challenge in computer vision and strengthens the learning value of the project.

---

🔍 Key Features

• Image preprocessing (resize, normalization, encoding)
• Custom CNN architecture implemented from scratch
• Training on labelled face images
• Accuracy & loss visualization
• Classification report with precision, recall, and F1-score
• Discussion of overfitting and future improvements

---

🛠️ Requirements

Install the typical dependencies:

numpy  
pandas  
matplotlib  
tensorflow  
opencv-python  
sklearn  

If your notebook uses more, they can be added.

---

▶️ How to Run

1️⃣ Clone the repository

git clone https://github.com/Sailaja-ASB/Face-Recognition-Project.git
cd Face-Recognition-Project

2️⃣ Install dependencies

pip install -r requirements.txt

3️⃣ Open the notebook

jupyter notebook "Face-Recognition.ipynb"

(Face-Recognition-Project.ipynb)

---

🚀 Future Improvements

To reduce overfitting & improve validation accuracy:

• Add image augmentation (rotation, zoom, brightness, flip)
• Add dropout and L2 regularization
• Increase dataset size
• Use transfer learning (MobileNetV2, VGG16, ResNet50)
• Apply EarlyStopping callback
• Balance dataset classes

---

👩‍💻 Author

Sailaja Morrennagari
GitHub: https://github.com/Sailaja-ASB
LinkedIn: https://www.linkedin.com/in/sailajamorrennagari