Lounes78/Siamese-NN-for-Face-Recognition

My goal throw this project is to go from a research paper (in this case, Siamese Neural Networks for One-shot Image Recognition) all the way to the complete functionning application.

Siamese-NN-for-Face-Recognition

My goal throw this project is to go from a research paper (in this case, Siamese Neural Networks for One-shot Image Recognition) all the way to a complete functionning application.

This repository contains code for implementing a Siamese Neural Network for image recognition using TensorFlow and Keras. The Siamese Neural Network is trained to distinguish between similar and dissimilar pairs of images.

Table of Contents

1. Introduction
2. Setup
3. Data Collection
4. Preprocessing
5. Model Architecture
6. Training
7. Evaluation

Data Collection

The dataset used in this project is organized into three folders: positive, negative, and anchor. Positive and negative folders contain images for training, while the anchor folder contains images used as reference points.

Preprocessing

The dataset is preprocessed using TensorFlow Datasets. Images are resized and scaled before being fed into the Siamese Neural Network. The data is split into training and testing sets.

Model Architecture

The Siamese Neural Network architecture consists of a shared embedding network followed by a distance calculation layer. The model is defined using the Keras API.

Training

The model is trained using a binary cross-entropy loss function and the Adam optimizer. Training checkpoints are saved, allowing for the possibility of resuming training.

Evaluation

The trained model is evaluated on a separate testing set. Precision and recall metrics are calculated to assess the model's performance.