Face Verification Using Siamese Neural Networks

Overview

This project implements Siamese Neural Network (SNN) for face verification using the Labeled Faces in the Wild (LFW) dataset. The model learns a similarity metric to verify whether two face images belong to the same person. The model is implemented for faceID verification in python kivi app where by storing some images for reference model can predict whether the new input image match the reference or not.

---

Requirements

1. Software Prerequisites

Ensure you have the following installed on your system:

• Python 3.7 or later
• Jupyter Notebook or Jupyter Lab

Required Libraries:

• TensorFlow (>=2.6.0)
• NumPy
• Matplotlib
• scikit-learn
• OpenCV
• Pandas
• kivy

---

Dataset

Download the Labeled Faces in the Wild (LFW) Dataset

1. Visit the LFW dataset page.
2. Download the dataset:
   • LFW Images: lfw.tgz.

---

File Structure

Below is the directory structure for the project:

FaceID-Verification/
├── app/                      # Python app for faceid verification
    ├── Application_Data/              # Application Data
        ├── Input_image                   # Store input image to verify
        ├── Verification_Images           # Store reference images
    ├── faceid.py                      # faceid verification app
    ├── layers.py                      # Distance calculator
├── data/                      # Directory for LFW dataset
    ├── anchor                         # store anchor images
    ├── negative                       # store negative images
    ├── negative                       # store positive images
├── Siamese Model Facial Verification Training.ipynb  #Jupyter notebook for model training
├── README.md                  # Project documentation
├── Application_data/          # Application data
    ├── Input_image                    # Store input image to verify
    ├── Verification_Images            # Store reference images

---

Running the Project

Model Training

1. Launch Jupyter Notebook Open a terminal in the project root directory and run:

   jupyter notebook

2. Open the Notebook In the Jupyter Notebook interface, navigate to and open the file: Siamese Model Facial Verification Training.ipynb.

3. Execute the Cells Run each code cell sequentially to:

   • Load and preprocess the dataset.
   • Define and train the Siamese Neural Network.
   • Evaluate the model.

4. View Results Results, such as training accuracy, loss curves, and test performance, will be displayed within the notebook.

Faceid Verification app

1. Add Reference Image In "/app/Application_data/Verification_Images" add 20-30 sample images for reference to verify.
2. Import saved model copy the saved model(.h5 file) in app directory.
3. Run App Open 'app' folder in VS Code and execute following command in terminal :

   python faceid.py

---

Key Notes

• GPU Support: If you have a compatible GPU, ensure TensorFlow is configured to utilize it for faster training. Follow TensorFlow GPU Setup Guide for more information.
• Adjust Parameters: You can modify the batch size, number of epochs, and learning rate in the training configuration cells.
• Data Path: Ensure the dataset is correctly placed in the data/ directory as the notebook relies on this structure.

---

Authors

This project was developed by Yash Arya. See the project report for details on contributions.