HoloTek

DISCLAIMER: The code contained in this repository is not optimized nor really pretty. I'm mostly playing with the DLib library and HoloLens.

Introduction

HoloLens application that allows a professor from EPITECH to mark student as 'present' at an activity, using Facial Recognition.

This was a project I developed for fun to explore the development on HoloLens using only C++ WinRT and DirectX.
This project uses the C++ Library dlib for the facial detection and recognition of the students.
Dlib is under the Boost Software License.

Installation

To use this projects, you have to install the dlib in this folders:

• vendor/lib -> the dlib.lib file
• vendor/include -> dlib folder containing all the headers of the library

Repo hierarchy

I developed this project in multiple steps:

• DesktopIntra is a UWP desktop application interacting with the REST api of the intranet of my school. It allowed me to develop utility functions and test them easily using a Xaml UI.
• DesktopTek is a UWP desktop application. I experimented on the DLib Facial Detection and Recognition features, and linked all that with the REST api of the intranet
• HoloTek is the UWP HoloLens application. It's the DesktopTek application, but using DirectX to render menus
• Serializer was used to pre-calculate the referenced images to reduce the time of recognition.

Face Recognition

Most of the interesting code is inside the HoloTek/Recognition/FacesBuffer.cpp file.

It works pretty well on the Desktop application, but the processing speed of the HoloLens is too slow: it takes approximately 15 seconds to process a frame.

Axes of improvement

• Send the camera stream to a remote server that will do the processing and recognition of images

Contact

I'm not planning in pursuing this project for now.
If you have questions do not hesitate to contact me either on my Twitter account (@oxydros) or via e-mail: oxydros at gmail.