Facial recognition web app

📌 This repository is for the moment only for dev, please do not put this version in production.

✅ App tested on macOS 13.1 (M1 Pro), Ubuntu 22.04 LTS (Intel Core i7) : Google Chrome

🚀 This project is a facial recognition app. It detects in real time faces present in a database on a live stream from the webcam. The good point of the MVC architechture implemented is that this facial recognition framework can be implemented in all type of device by using API endpoints.

Back-end : django + django REST Framework
Front-end : ReactJS
Database : PostgreSQL

📰 Face recognition framework is adapted from DeepFace. It is a hybrid face recognition framework wrapping state-of-the-art models: VGG-Face, Google FaceNet, OpenFace, Facebook DeepFace, DeepID, ArcFace, Dlib and SFace.

The model to be used as well as the distancing method can be directly selected from the file .env file

Installation of app 💻

Download the project

First, clone the project by executing :

git clone git@github.com:sachabinder/faceRecognition.git

then, you can access the repository folder on your local machine by executing :

cd faceRecognition

You are now ready to go !

⚠️ WARNING : Before playing with docker, if you are on ARM architecture (Apple chips), please uncomment this line ⚠️

Build and up the project

To build docker images of containers you should run :

docker-compose build

To run the app on local machine you should run :

docker-compose up

ℹ️ PLEASE WAIT THE COMPLETE STARTING OF THE SERVICES, it shoud take few minutes. In fact, front can take some time to build and back should dowload model weights. Everthing is started when you see :

facerecognition-front-1  | webpack 5.75.0 compiled successfully in 6159 ms
facerecognition-back-1   | Watching for file changes with StatReloader
facerecognition-back-1   | Performing system checks...
facerecognition-back-1   | 
facerecognition-back-1   | System check identified no issues (0 silenced).
facerecognition-back-1   | March 06, 2023 - 23:14:51
facerecognition-back-1   | Django version 4.1.2, using settings 'back.settings'
facerecognition-back-1   | Starting development server at http://0.0.0.0:8000/
facerecognition-back-1   | Quit the server with CONTROL-C.

You can now access the app (on port 8000 for the back and 8080 for the front which can be change in env file ) by :

localhost:8000 for the REST api
localhost:8000/admin for the django admin
localhost:8080 for the web app

Access to database

You can access to the PostgreSQL database shell (in the assigned docker container) by running :

docker exec -it $(docker ps -f 'name=db' --format '{{.ID}}') psql -U name

Operate on the back-end

You can access the prompt of back-end docker container by running :

docker-compose exec back /bin/bash

django tips

To create a super user from the back-end, you should run :

python manage.py createsuperuser

To create migration from the back-end (TO BE DONE AFTER MODEL MODIFICATIONS), you should run :

python manage.py makemigrations

To execute migrations, you shoud run (automaticly done at container starting) :

python manage.py migrate

Add a Python dependancy (Backend)

To manage dependancy on the backend, we are using poetry. To add a Python module, you should add it on pyproject file and then re-build Docker containers.

Operate on the front-end

You can access the shell of front-end docker container by running :

docker-compose exec front /bin/sh

NPM server

To start/restart the webpack dev server, run on the container :

/usr/local/bin/npm run dev

Add a Node dependancy

To manage dependancy on the frontend, we are using npmJS. To add a Node module, you should add it on package.json and then re-build the back Docker container.

You can also execute in the front-end container :

/usr/local/bin/npm install [your-package-name] -D

Then, the package.json will be automatically update.

Front Linting

If you want to test the linting of your file before pushing it on the repo, you can install the eslint extension in VS Code, or just use the following lines if you have previously installed npm and the eslint dependency:

npm run lint

and if you think the errors can be corrected directly by the linter without problem:

npm run lint:fix

Environment

Variable	Description	Dev value
DB_HOST	Host name of the db	db
DB_PORT	Port of the DB (in container).	5432
DB_USER	User of the DB	user
DB_NAME	Name of the DB	name
DB_PASSWORD	Password of the DB	password
BACK_PORT	External back port access	8000
FRONT_PORT	External front port access	8080
DEBUG	Debug mode execution	True
DOMAIN_NAME	Domaine name to use in production. Requests from other domain will be rejected.	127.0.0.1
SECURE_SSL_REDIRECT	Automaticly redirect non HTTPS requests in HTTPS. To be leave at False if an other proxy is already doing this work	False
FACE_DETECTION_MODEL_NAME	Face recognition model name. Model available : `VGG-Face`, `Facenet`, `OpenFace`, `DeepFace`, `DeepID`	`VGG-Face`
FACE_DETECTION_DISTANCE_METRIC	Distance metric to compare embedings. Distance available : `cosine`, `euclidean`, `euclidean_l2`	`cosine`
FACE_DETECTION_DETECTOR_BACKEND	face detector backend. backend available : `opencv`, `retinaface`, `mtcnn`, `ssd`, `dlib`, `mediapipe`	`opencv`

Command to run to generate a secret django key (you should get django installed)

python3 -c 'from django.core.management.utils import get_random_secret_key; print(get_random_secret_key())'