Face Recognition

Project done as a part of "High performance network services" class at University of Warsaw 2018/19.

Description

Main part of the project was to create Highly Available and easy to scale service.

Project resulted in service consisiting of following parts:

• Frontend services - generating html files from templates
• Face recognition service - operations on images linked to face recognition
• Nginx front server - load balancing between front services and caching
• Mongo and MySQL databases - storing data
• RabbitMQ - offloading computations from frontend services to face recognition service

Features

• Round robin load balancing using Nginx
• Easy deployment with Docker using docker-compose
• Caching requests for 2 seconds using Nginx
• Partitioning in MySQL database
• Asynchronous processing of incoming heavy requests with images thanks to RabbitMQ

System diagram

Views

Main view - added images with lists of people on them

People view - added people with their face pictures

Single image view - fullsize image with marked faces

Technology stack

• Frontend: HTML/CSS, Bootstrap, Javascript
• Backend: Python, Flask, MongoDB, MySQL, Nginx, RabbitMQ, face_recognition lib
• Other: Docker, Git

How to run

After executing commands mentioned below whole service should be accessible on localhost:8080.

docker-compose build;
docker-compose up;