Real-time Facial Recognition 2.0

This repository contains the source code for the 2nd version of the real time facial recognition system. It is the successor to this.

---

Table of Contents

• Description
• Features
• Installation
• Usage
• Contributing
• License
• Acknowledgements

---

Description

Real-time FRS 2.0 is an app with the capability to conduct facial recognition in real-time. Uses include attendance taking and showcase purposes.

The main FRS software can be found in the simpliFRy directory. SimpliFRy is a locally-hosted web application built using python 3.10 and Flask. It makes use of the insightface library by deepinsight for face detection and generation of embeddings and the voyager library by Spotify for K-Nearest Neighbour search.

Coming with simpliFRy is gotendance. gotendance is also a locally-hosted web application, but built using Golang. It is an attendance-tracking app, intended as a companion to simpliFRy. The UI is in HTML/CSS/JS to ensure it is lightweight and easily deployable.

---

Features

SimpliFRy is designed with the use of Real-Time Streaming Protocol (RTSP)-capable cameras in mind. The software access the camera's feed through an RTSP URL.

As compared to the previous iterations, Real-time FRS 2.0 has the following benefits:

• Independent of 3rd party softwares such as OBS Studio (no need virtual camera).
• Able to do real-time facial recognition on multiple cameras simultaneously with just 1 computer.
• Designed to be simple and easy to use.
• Need not deal with complicated GPU prerequisites.

---

Installation

For more information on installation, refer to the following:

• simpliFRy
• gotendance

---

Usage

For more information on installation, refer to the following:

• simpliFRy
• gotendance

---

Contributing

---

License

This project is licensed under the Apache License 2.0 - see the LICENSE.md file for details.

---

Acknowledgements

I would like to extend my gratitude to the following people and resources:

• Insightface: The weights provided by insightface formed the core of this project's Facial Recognition capabilities.
• Voyager: Voyager's vector search provided a quick and efficient way to match the closest embeddings.
• FFmpeg: SimpliFRy uses FFmpeg to capture the RTSP stream for frame-by-frame analysis.
• Ruihongc: Suggesting the use of Spotify's Voyager led to much better faster performance compared to previous methods.
• BabyWaffles: Dockerization of simpliFRy was made possible by his extensive help.
• Tabler Icons: Multiple icons from Tabler were used in the UI of both simpliFRy and gotendance