Fuskar

This is a facial recognition application built to be deployed on an NVIDIA Jetson Nano It combines the similarity comparison of the FaceNet Model trained on a Siamese network into a django API served to a VueJS frontend

Development

Backend

The API was built using django and face_recognition

    # install redis
    sudo apt-get install redis-server

    # create virtualenv
    mkvirtualenv fuskar
    # install requirements
    pip install -r backend/requirements.txt
    # create postgres database
    sudo su postgres -c "psql -c \"CREATE ROLE fuskar with password 'fuskar';\""
    sudo su postgres -c "psql -c \"CREATE DATABASE fuskardb with owner fuskar;\""
    sudo su postgres -c "psql -c \"ALTER ROLE fuskar WITH LOGIN;\""
    # change directory
    cd backend
    # run migrations
    python manage.py migrate
    # start server
    python manage.py runserver

Frontend

The front end was built using nuxtjs, a progressive vuejs framework that supports server side rendering. Code here

Build Setup

# install dependencies
$ yarn install

# serve with hot reload at localhost:3000
$ yarn dev

# build for production and launch server
$ yarn build
$ yarn start

# generate static project
$ yarn generate

For detailed explanation on how things work, check out Nuxt.js docs.

Usage

• Turn on the device shown below

• Use a laptop or phone to connect the device to the device hotspot with the password 'fuskar-owner'
• Visit the web application on 10.42.0.1

Deploy Plans

• Create endpoints for student registration and management

• Create endpoints for course registration and management

• Create asynchronous tasks for taking attendance and retraining pickled parameter

• Debug setup for huey where huey does not run tasks

• Complete NVIDIA jetson Nano setup to install dlib with gpu capabilities

• Auto start wireless hotspot

• Allow all incoming traffic on port 80 using ufw

• Create nginx sites-available script for redirecting traffic from port 80 to the front-end port

• Back-end (Django) application

  • Auto Switch camera to read from jetson nano streamer

  • Create Daphne service file

  • Create Redis-server broker service file

  • Make sure to start Huey service before Daphne service

  • Tie Huey service start to start Celery service

• Front-end (Vue) application

  • Create Vue service file

  • Connect to back-end port

  • Crop whitespace out of captured image

  • Create Vue server service file to run on start, depending on django service file