skaty5678 / face_recognition

Building a deep facial recognition application to authenticate into an application. Building a model using Deep Learning with Tensorflow which replicates what is shown in the paper titled Siamese Neural Networks for One-shot Image Recognition.

Facial recognition using Siamese Neural Network

A Face Recognition Siamese Network implemented using Keras. Siamese Network is used for one shot learning which do not require extensive training samples for image recognition.

Steps involved in the process

1. Setup tensorflow and keras for the deep learning

2. Collecting the image samples

Using opencv to collect the anchor and positive images and using the lfw dataset to collect the negative images.

3. Loading and preprocessing images.

scaling and resizing the image to (100,100,3). Creating the labelled dataset i.e. (anchor,positive)--> 1,1,1,1 and (anchor,negative) --> 0,0,0,0

4. Buidling train and test partition.

5.Model Engineering

Building embedding layer.

Building the distance layer.

Making the Siamese model.

6. Trainig the model

Setup loss and optimizer

Establishing checkpoints for the training_checkpoints

Build train step funtion and the training loop

Training the model.

7. Evaluating the model.

Importing the metrics.

Make predictions.

Calculate metrics.

Visualise results.

8. Saving the model.

Saving the model as an h5 file.

9. Real time verification

real time veirification using webcam.