This repository contains the implementation of our research paper, check the paper on Arxiv

<div id = 'top'></div>

IdentiFace

A Multimodal Facial Biometric System for Recognition, Gender Classification, Emotion Recognition and Face-Shape Prediciton

<div align='center'>

</div>

---

Contents:

• <a href="#ps">Project Structure and setup</a>
• <a href="#models">Models</a>
• <a href="#gui">GUI</a>
• <a href="#members">Team Members</a>

---

<div id="ps">

Project structure

├── main.py [Main file: Contains the welcome window]
├── Backend
|    ├── functions.py [contains all the used functions]
|    ├── model_manager.py [manages the models across windows]
|    ├── offline.py [offline window layout]
|    ├── online.py [online window layout]
├── utilities [Face-Detection: the used Dlib files for facial detection]
├── assets [Directory for project assets]
├── Models [a drive link for all the used models]
├── snapshots [contains all the notebooks and the codes for the different modalities]
├── test_examples [Test images]
├── snapshots [Snaps taken from the app]
└── requirements.txt [List of all required Python modules]

Getting started :

1. Clone the repository
2. Install the required dependencies by running pip install -r requirements.txt

dlib==19.24.2
keras==3.0.2
matplotlib==3.8.2
numpy==1.26.2
PySide6==6.6.1
tensorflow==2.15.0.post1

3. run main.py to start the application

</div> <div id="models">

Models

check the <a href="https://github.com/MahmoudRabea13/IdentiFace/blob/main/IdentiFace%20A%20VGG%20Based%20Multimodal%20Facial%20Biometric%20System%20.pdf">Paper</a> for more detailed information about the data used / preprocessing / methodology or any other aspect of the project

The final used models in the GUI were as follows:

I. Face Recogniton Model trained on a subset of <a href="https://www.nist.gov/itl/products-and-services/color-feret-database">the color FERET database</a>

II. Gender Classification Model trained on a <a href="https://www.kaggle.com/datasets/cashutosh/gen der-classification-dataset/data"> Public Gender dataset </a>

III. Face-Shape Prediciton Model trained on <a href="https://www.researchgate.net/publication/328775300_A_Hybrid_Approach_to_Building_Face_Shape_Classifier_for_Hairstyle_Recommender_System">the Celebrity face-shape dataset</a>

IV. Emotion Recognition Model trained on <a href="https://www.kaggle.com/datasets/msambare/fer2 013">the FER2013 dataset</a>

<div align="center">

|Model|Train Accuracy|Test Accuracy|Confusion Matrix| |-------|----|-----|------| |Face Recognition|99.7%|99.2%| | |Gender Classification|96.48%|95.15%| | |Face-Shape Prediction|99.79%|88.03%| | |Emotion Recognition|81.26%|66.13%| |

</div> </div> <div id="gui">

GUI

We developed a Pyside desktop application called IdentiFace

The app mainly consists of:

I. A welcome window

II. An offline window

III. An online window

Note that because of the recognizer require high quality images , it was added only to the offline mode.

|window|screenshot| |---|---| |welcome window|| |offline window|| |offline window|| |online window||

</div> <div id="members">

Team Members

• Mahmoud Rabea
• Hanya Ahmad
• Nourhan Sayed
• Sohaila Mahmoud

Note that this project was part of the Biometrics in the Senior SBME year at Cairo University under the supervision of DR. Ahmed.M.Badawi and the guidance of TA Laila Abbas

</div> <p align="right"><a href="#top">Back to top</a></p>