Facial Verification [PyTorch]

demo.webm

Introduction

This application offers a convenient and secure login solution using facial recognition technology. Users can easily access their accounts by simply presenting their face to the camera, without the need for traditional login credentials.

Key Features

• Robust facial verification model built using PyTorch
• Seamless user authentication experience
• Secure data handling and storage
• Customizable model training with flexible parameters
• Support for both local and cloud-based model training

Model training

Users have several options to train the facial verification model:

• Download lfw dataset
• Install the required dependencies by running pip install -r requirements.txt

Local Training:

• Run python3 train_model.py -sp path/to/input/folder to train the model with default parameters on a local dataset.
• Use the -nl flag if you have already downloaded the dataset.

Cloud-based Training:

• Run all cells in the train_azure.ipynb notebook in the train_azure folder to train the model on Azure with default parameters.

Custom Training:

• Run python3 train_model.py -b batch_size -lr learning_rate to train the model with your preferred batch size and learning rate.

Web Application

Users can access the web application by running the Website branch.

Front-End

The front-end of the application is located in the user-interface folder. To set up and run the front-end, follow these steps:

• Navigate to the user-interface folder.
• Install the required dependencies by running npm install.
• Start the front-end development server by running npm start.

Back-End

The back-end of the application is located in the server-side folder. To set up and run the back-end, follow these steps:

• Navigate to the server-side folder.
• Install the required dependencies by running pip install -r requirements.txt.
• Set up an Azure SQL database to store the features of faces.
• Start the back-end server by running python3 app.py.