Great work!!

I was wondering if I could use a different object detection model like YOLO or tiny yolo and have bounding boxes in the camera?

Thank you

Win 10 18362.1082
Unity 2019.4.9f1
HL2 10.0.19041.1375

This might be a HL2 or Unity issue rather than an issue with your app.
When I run the app with the legacy "XR settings" - "VR supported" settings enabled in the player, the app provides me with the correct depth information and works like described.
If I however use the new Unity XR Plugin Management system in the Project Settings, the reported depth is always 0.

Hi,
When I tried to build the project in unity, it showed like that.

I just open your project in unity 2018.1.31f and build it for Universal Windows Platform.
Could you please help me to check if there is anything wrong with my building or is there anything else I need to install in unity?

Thank you very much and hope you have a good day.

The example is great and works without problems out of the box, thanks a lot!

I have a question with regards to replacing the small squeezenet with a larger model such as posenet for bodytracking.
I saw that you load the model as a TextAsset, which seems to lead to memory issues for a larger model.
Can this be loaded eg. as a StreamingAsset and do you know how that can be implemented?

The app crashes when I try to deploy it.

The error occurs in app.cs in line 19
m_AppCallbacks = new AppCallbacks();

The output messages are

An unhandled exception of type 'System.IO.FileLoadException' occurred in HoloVision18.exe
The process cannot access the file because it is being used by another process. (Exception from HRESULT: 0x80070020)

Die Aktivierung der Windows Store-App "HoloVision18_awaf3f2xac5x2!App" ist mit Fehler "Die App wurde nicht gestartet" fehlgeschlagen.
Das Programm "[5624] HoloVision18.exe" wurde mit Code 0 (0x0) beendet.

I have no idea what other process could use the file (what file ever it is). Even a freshly restarted HoloLens won't run the app.

Have an issue loading the onnx model from storage file. I can confirm that the file was stored into 'model.bytes' and loaded using LoadModelFromStorageFileAsync,
after loading, the _model.Description.InputFeatures is null.
Any ideas?

Hi,
after I deployed the project to my hololens emulator, the emulator showed a green cross and a text saying: "Error Init: Object Reference not set to an instance of an object".

I used the same unity version as you and vs 2019. I deployed on x64 release version

Hey Rene,

great work, the example scene and ML model works fine! I have not really an issue but more of a question: I'd like to grab the camera image and manipulate it (using openCV), before using it with an ONNX model. However I'm badly stuck at trying to set up the input for the model coming from an Unity Texture2D (getting to the Windows specific Videoframe). Did you by chance try something like that or have some insights how to do that?

Best regards
Stoffel