Various projects leveraging Windows Machine Learning (WinML) for AI Deep Learning inference running via UWP on the HoloLens. The demo is mainly used in the conference sessions by Rene Schulte presented at events like //build, Unite Berlin AutoTech Summit and more.
This is a Unity Project using WinML and SqueezeNet trained on a modified ImageNet dataset to classify objects the user is seeing in front. Spatial Mapping is used to measure the distance and Spatial Sound with Text-To-Speech to tell the user the most dominant object in the space.
Make sure to have the HoloLens update RS5 / Build 177xx / 1809 installed:
https://forums.hololens.com/discussion/10696/announcing-the-hololens-rs5-preview
On your development machine you also need at least Unity 2018.2 and Visual Studio 2017 with the Windows SDK for RS5 / Build 177xx / 1809 installed.
More details and some of the session recording can be found here: https://kodierer.blogspot.com/2018/06/content-for-unite-berlin-autotech.html
http://kodierer.blogspot.de/2018/05/content-for-microsoft-build-2018.html
https://kodierer.blogspot.com/2018/11/the-sky-is-not-limit-content-for.html
https://kodierer.blogspot.com/2018/09/content-for-digility-session-advanced.html
This is the HoloLens 2 version of the project.