Social Distancing Tracker

Tool that detects when people do not follow the social distancing rules by analyzing videos from surveillance cameras.

Setup

Tested on Ubuntu 18.04

Social Distancing Tracker

For detecting and tracking people we used code provided by pyimagesearch opencv-people-counter

$ cd people_detection

• Set up virtual environment:

$ sudo apt-get install python3-venv
$ python3 -m venv venv
$ source venv/bin/activate

• Install dependencies (NB: the dlib dependency requires at least >2GB RAM and needs to be created from a clean virtual environment)

$ sudo apt-get install build-essential cmake libopenblas-dev liblapack-dev libx11-dev libgtk-3-dev python3-dev
$ pip install -vr requirements.txt

• Run tracker

python group_detection.py [path to video/stream] -o [path of output directory]

UI

We provide a web-based UI that allows you to upload videos and get them analyzed.

$ cd app/

• Install MongoDB Community Edition

• Create database:

$ mongo
> use MyDatabase;
> exit

• Open index.js and adapt user credentials:

UserDetails.register({username:'anna', active: false}, 'anna');

• Install ffmpeg to convert videos

$ sudo apt-get install ffmpeg

• Install node.js and dependencies:

$ curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.35.3/install.sh | bash
# open new terminal after that
$ nvm install stable
$ npm install

• Run application

$ node index.js

• Open browser and type:

localhost:3000